Coding Techniques for Neural Network Architectures

Abstract

Methods, systems, and devices for wireless communications are described. A user equipment (UE) may receive an indication of one or more encoding operations to be used to encode a compressed data set, the one or more encoding operations including a differential encoding operation or an entropy encoding operation or both. In some examples, using a neural network, the UE may first encode the data set based on an additional encoding operation to generate a compressed data set, and then quantize the compressed data set that is encoded based on the additional encoding operation. Subsequently, after the data set has been first encoded and then quantized, the UE may further encode and compress the data set using the indication of the one or more encoding operations. The UE may then transmit the data set to a second device based on the one or more encoding operations.

Description

[Technical Field]

[0001] cross reference This patent application claims the interests of U.S. Patent Application No. 17 / 194,077, entitled "ENCODING TECHNIQUES FOR NEURAL NETWORK ARCHITECTURES," filed on March 5, 2021, and assigned to the assignee of this application.

[0002] The following concerns wireless communications, including coding techniques for neural network architectures. [Background technology]

[0003] Wireless communication systems are widely deployed to provide various types of communication content, including voice, video, packet data, messaging, and broadcast. These systems can support communication with multiple users by sharing available system resources (e.g., time, frequency, and power). Examples of such multiple access systems include fourth-generation (4G) systems such as Long Term Evolution (LTE) systems, LTE-Advanced (LTE-A) systems, or LTE-A Pro systems, and fifth-generation (5G) systems, sometimes called New Radio (NR) systems. These systems may employ techniques such as code division multiple access (CDMA), time division multiple access (TDMA), frequency division multiple access (FDMA), orthogonal FDMA (OFDMA), or discrete Fourier transform spread orthogonal frequency division multiplexing (DFT-S-OFDM). [Overview of the project] [Problems that the invention aims to solve]

[0004] A wireless multiple access communication system may include one or more base stations or one or more network access nodes, each simultaneously supporting communication for multiple communication devices, which may sometimes be known as user equipment (UEs). In some wireless communication systems, UEs may perform several channel state measurements as part of their communication with the base station. In some examples, these measurements may generate a large amount of data that will be sent to the base station to assist it in network management. [Means for solving the problem]

[0005] The techniques described relate to improved methods, systems, devices, and apparatus that support coding techniques for neural network architectures. Generally, the techniques described enable a user device (UE) to receive an indication of one or more coding operations to be used to code a compressed dataset, where one or more coding operations include differential coding operations, entropy coding operations, or both. In some examples, using a neural network, the UE may first encode the dataset based on an additional coding operation (e.g., a single-shot encoder) to produce a compressed dataset, and then quantize the compressed dataset encoded based on the additional coding operation. Subsequently, after the dataset has been initially encoded and then quantized, the UE may further encode and compress the dataset using the indication of one or more coding operations. The UE may then transmit the dataset (e.g., encoded, quantized, and compressed datasets) based on the use of one or more coding operations to a second device (e.g., a base station, an additional UE, etc.) where the information may be further encoded after the data has been quantized using a neural network. In some examples, differential coding operations may involve coding the amount of data in a compressed dataset (for example, after additional coding operations) based on previous values ​​for the amount of data, such as initial values, initial reconstructed values, previous reconstructed values, and previous values ​​for the same data from previous time instances.

[0006] A method for wireless communication in a UE is described. The method may include: receiving an indication of one or more coding operations to be used to encode a compressed dataset, wherein the one or more coding operations include a differential coding operation or an entropy coding operation or both; encoding a dataset by a neural network to produce a compressed dataset; quantizing the compressed dataset encoded by the neural network; encoding the quantized and compressed dataset based on the receipt of the indication of one or more coding operations; and transmitting the encoded, quantized and compressed dataset to a second device after encoding the compressed dataset based on one or more coding operations.

[0007] A device for wireless communication in a UE is described. The device may include a processor, memory coupled to the processor, and instructions stored in the memory. The instructions may be executable by the processor to cause the device to receive an indication of one or more coding operations to be used to encode a compressed dataset, the one or more coding operations including differential coding operations or entropy coding operations or both, to cause a neural network to encode the dataset to generate a compressed dataset, to quantize the compressed dataset encoded by the neural network, to encode the quantized and compressed dataset based on the receipt of the indication of one or more coding operations, and, after encoding the compressed dataset based on one or more coding operations, to transmit the encoded, quantized, and compressed dataset to a second device.

[0008] Another apparatus for wireless communication in a UE is described. The apparatus may include means for receiving indications of one or more coding operations to be used to encode a compressed dataset, wherein the one or more coding operations include differential coding operations or entropy coding operations or both; means for encoding a dataset by a neural network to generate a compressed dataset; means for quantizing the compressed dataset encoded by the neural network; means for encoding the quantized and compressed dataset based on the reception of indications of one or more coding operations; and means for transmitting the encoded, quantized and compressed dataset to a second device after encoding the compressed dataset based on one or more coding operations.

[0009] A non-temporary, computer-readable medium for storing code for wireless communication in a UE is described. The code may include instructions executable by a processor to receive an indication of one or more coding operations to be used to encode a compressed dataset, where one or more coding operations include differential coding operations or entropy coding operations or both, encode the dataset by a neural network to produce a compressed dataset, quantize the compressed dataset encoded by the neural network, encode the quantized and compressed dataset based on the receipt of the indication of one or more coding operations, and transmit the encoded, quantized and compressed dataset to a second device after encoding the compressed dataset based on one or more coding operations.

[0010] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, receiving an indication of one or more coding operations may include an operation, feature, means, or instruction for receiving one or more parameters corresponding to one or more coding operations, each of which corresponds to each coding operation of the one or more coding operations, and the quantized and compressed dataset may be coded based on the one or more parameters.

[0011] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, encoding a quantized and compressed dataset may include operations, features, means, or instructions for encoding a quantized and compressed dataset using differential coding operations after encoding the dataset using a neural network.

[0012] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, encoding a quantized and compressed dataset may include operations, features, means, or instructions for encoding a quantized and compressed dataset using an entropy coding operation after encoding the dataset using a neural network.

[0013] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, encoding a quantized and compressed dataset may include an operation, feature, means, or instruction for determining, after quantizing the compressed dataset to be encoded by a neural network, the difference between a first value of data in the quantized and compressed dataset in an initial time instance and a second value of data in a second time instance after the initial time instance, wherein the difference value may be determined based on one or more coding operation markings, and the quantized and compressed dataset may be encoded based on the difference value.

[0014] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, encoding a quantized and compressed dataset may include an operation, feature, means, or instruction for determining, after quantizing the compressed dataset to be encoded by a neural network, the difference between a first reconstructed value of the data in the quantized and compressed dataset in a first time instance and a second reconstructed value of the data in a second time instance following the first time instance, wherein the difference value may be determined based on one or more coding operation markings, and the quantized and compressed dataset may be encoded based on the difference value.

[0015] In some examples of the methods, apparatus, and non-temporal computer-readable media described herein, encoding a quantized and compressed dataset may include an operation, feature, means, or instruction for determining an initial reconstruction value of the data in the quantized and compressed dataset at an initial time instance value related to encoding the dataset, after quantizing the compressed dataset to be encoded by a neural network, and after quantization, determining a difference value between the initial reconstruction value of the data and an additional reconstruction value of the data at an additional time instance after the initial time instance, the difference value may be determined based on one or more coding operation markings, and the quantized and compressed dataset may be encoded based on the difference value.

[0016] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, the differential coding operation may involve coding the amount of data in a compressed dataset based on a previous value for the amount of data in that dataset.

[0017] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, the entropy coding operation may include coding a compressed dataset using one or more symbols having lengths that vary based on the probability of the symbols appearing.

[0018] A method for wireless communication in a device is described. The method includes transmitting, to a UE, an indication of one or more encoding operations for the UE to use to encode a compressed data set, where the one or more encoding operations include a differential encoding operation or an entropy encoding operation or both; receiving, from the UE, the encoded, quantized, and compressed data set after the compressed data set has been encoded based on the one or more encoding operations followed by a quantization operation based on the one or more encoding operations; decoding the encoded, quantized, and compressed data set to generate the compressed data set based on the one or more encoding operations; and decoding the compressed data set to generate a data set based on decoding the encoded, quantized, and compressed data set by a neural network.

[0019] An apparatus for wireless communication in a device is described. The apparatus can include a processor, a memory coupled to the processor, and instructions stored in the memory. The instructions can be executable by the processor to cause the apparatus to transmit, to a UE, an indication of one or more encoding operations for the UE to use to encode a compressed data set, where the one or more encoding operations include a differential encoding operation or an entropy encoding operation or both; receive, from the UE, the encoded, quantized, and compressed data set after the compressed data set has been encoded based on the one or more encoding operations followed by a quantization operation based on the one or more encoding operations; decode the encoded, quantized, and compressed data set to generate the compressed data set based on the one or more encoding operations; and decode the compressed data set to generate a data set based on decoding the encoded, quantized, and compressed data set by a neural network.

[0020] Another device for wireless communication in a device is described. The device is means for transmitting to a UE an indication of one or more encoding operations for the UE to use to encode a compressed data set, where the one or more encoding operations include a differential encoding operation or an entropy encoding operation or both; means for receiving from the UE, following a quantization operation based on the one or more encoding operations, an encoded, quantized, and compressed data set after the compressed data set has been encoded; means for decoding the encoded, quantized, and compressed data set to generate the compressed data set based on the one or more encoding operations; and means for decoding the encoded, quantized, and compressed data set to generate a data set based on the neural network decoding the encoded, quantized, and compressed data set.

[0021] A non-transitory computer-readable medium storing code for wireless communication in a device is described. The code includes instructions executable by a processor to transmit to a UE an indication of one or more encoding operations for the UE to use to encode a compressed data set, where the one or more encoding operations include a differential encoding operation or an entropy encoding operation or both; receive from the UE, following a quantization operation based on the one or more encoding operations, an encoded, quantized, and compressed data set after the compressed data set has been encoded; decode the encoded, quantized, and compressed data set to generate the compressed data set based on the one or more encoding operations; and decode the encoded, quantized, and compressed data set to generate a data set based on the neural network decoding the encoded, quantized, and compressed data set.

[0022] In some examples of methods, apparatus, and non-temporary computer-readable media described herein, transmitting an indication of one or more coding operations may include an operation, feature, means, or instruction for transmitting one or more parameters corresponding to one or more coding operations, each of which corresponds to each coding operation of the one or more coding operations, and the quantized and compressed dataset may be coded based on the one or more parameters.

[0023] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, decoding an encoded, quantized, and compressed dataset may include operations, features, means, or instructions for decoding an encoded, quantized, and compressed dataset using differential decoding operations before decoding the dataset using a neural network.

[0024] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, decoding an encoded, quantized, and compressed dataset may include operations, features, means, or instructions for decoding an encoded, quantized, and compressed dataset using an entropy decoding operation before decoding the dataset using a neural network.

[0025] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, receiving an encoded, quantized, and compressed dataset may include an operation, feature, means, or instruction for receiving an encoded, quantized, and compressed dataset that includes differential values ​​to the data in the dataset, which may be based on the initial values ​​of the data.

[0026] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, receiving an encoded, quantized, and compressed dataset may include an operation, feature, means, or instruction for receiving an encoded, quantized, and compressed dataset that includes differential values ​​to the data in the dataset, which may be based on previous reconstruction values ​​of the data.

[0027] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, receiving an encoded, quantized, and compressed dataset may include an operation, feature, means, or instruction for receiving an encoded, quantized, and compressed dataset that includes differential values ​​to the data in the dataset, which may be based on the initial reconstructed values ​​of the data.

[0028] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, the differential decoding operation may involve encoding the amount of data in a compressed dataset based on a previous value for the amount of data in that dataset.

[0029] In some examples of the methods, apparatus, and non-temporary computer-readable media described herein, the entropy decoding operation may involve encoding a compressed dataset using one or more symbols having lengths that vary based on the probability of the symbols appearing. [Brief explanation of the drawing]

[0030] [Figure 1] This figure shows an example of a wireless communication system that supports coding techniques for neural network architectures according to the embodiments of this disclosure. [Figure 2] This figure shows an example of a wireless communication system that supports coding techniques for neural network architectures according to the embodiments of this disclosure. [Figure 3A] This figure shows an example of a compression procedure that supports coding techniques for neural network architectures according to the aspects of this disclosure. [Figure 3B] This figure shows an example of a compression procedure that supports coding techniques for neural network architectures according to the aspects of this disclosure. [Figure 4] This figure shows an example of a compression configuration that supports coding techniques for neural network architectures according to the embodiments of this disclosure. [Figure 5] This figure shows an example of a compression and coding configuration that supports coding techniques for neural network architectures according to aspects of the present disclosure. [Figure 6] This figure shows an example of a compression and coding configuration that supports coding techniques for neural network architectures according to aspects of the present disclosure. [Figure 7] This figure shows an example of a process flow supporting coding techniques for neural network architectures according to the aspects of this disclosure. [Figure 8] This is a block diagram of a device supporting coding techniques for neural network architectures according to the embodiments of this disclosure. [Figure 9] This is a block diagram of a device supporting coding techniques for neural network architectures according to the embodiments of this disclosure. [Figure 10] This is a block diagram of a communications manager supporting coding techniques for a neural network architecture according to the aspects of this disclosure. [Figure 11] This is a diagram of a system including a device that supports coding techniques for neural network architectures, according to an aspect of this disclosure. [Figure 12] This is a block diagram of a device supporting coding techniques for neural network architectures according to the embodiments of this disclosure. [Figure 13] This is a block diagram of a device supporting coding techniques for neural network architectures according to the embodiments of this disclosure. [Figure 14]This is a block diagram of a communications manager supporting coding techniques for a neural network architecture according to the aspects of this disclosure. [Figure 15] This is a diagram of a system including a device that supports coding techniques for neural network architectures, according to an aspect of this disclosure. [Figure 16] This flowchart shows a method for supporting coding techniques for neural network architectures according to the aspects of this disclosure. [Figure 17] This flowchart shows a method for supporting coding techniques for neural network architectures according to the aspects of this disclosure. [Figure 18] This flowchart shows a method for supporting coding techniques for neural network architectures according to the aspects of this disclosure. [Figure 19] This flowchart shows a method for supporting coding techniques for neural network architectures according to the aspects of this disclosure. [Figure 20] This flowchart shows a method for supporting coding techniques for neural network architectures according to the aspects of this disclosure. [Modes for carrying out the invention]

[0031] In some wireless communication systems, user equipment (UE) may perform several channel state measurements as part of communicating with a base station. For example, these measurements may include per-antenna port channel and interference measurements (e.g., channel state feedback), power measurements from serving and neighboring cells, inter-radio access technology (RAT) measurements (e.g., from a Wireless Fidelity (WiFi) network), and sensor measurements. These measurements can generate a large amount of data that will be transmitted to the base station to assist it in network management. For example, a wireless communication system using an antenna panel with multiple elements may generate more measurement information than a wireless communication system using a single antenna or a smaller antenna panel with fewer elements. In some examples, the UE may use a neural network to compress the measurement results, reducing the size of the transmission carrying the results. In addition to neural network-based compression techniques, techniques for applying additional compression techniques to the measurement data are desirable.

[0032] Techniques for compressing measured data by a UE for transmission are described. A UE may measure one or more channel conditions. The UE may then compress the measured data using a neural network. The output of the neural network may be quantized to provide data that is more likely to be transmitted. In addition to neural network compression, the UE may use an additional layer of compression before sending the transmission to the base station. For example, the UE may further compress the measured data after compressing it with a neural network and before sending the transmission to the base station, using differential coding, entropy coding, or both. Differential coding may be an example of coding the difference between two measured results rather than coding the absolute value of the measured result. For example, differential coding may use the previous value, initial value, reconstructed value, initial reconstructed value, or a combination thereof to indicate the difference value. In some examples, the UE may further compress the data after the quantization step associated with neural network processing by performing differential coding, entropy coding, or both. Additionally or alternatively, differential coding may be performed at the input stage of the compression operation, where bits representing the difference between the measured results are first coded and then input to the neural network for processing. The base station can then use differential decoding, entropy decoding, or both when decoding the compressed and encoded transmission from the UE. In some examples, the base station may transmit indicators of the differential and entropy coding (e.g., and parameters for each coding) that the UE should use. Thus, the coding operation performed by the transmitting device (e.g., the UE) and the decoding operation performed by the receiving device (e.g., the base station) can be coordinated.

[0033] The aspects of this disclosure are first described in the context of wireless communication systems. In addition, the aspects of this disclosure are shown through additional wireless communication systems, compression procedures, compression configurations, compression and coding configurations, and process flows. The aspects of this disclosure are further illustrated and described by apparatus diagrams, system diagrams, and flowcharts relating to coding techniques for neural network architectures.

[0034] Figure 1 shows an example of a wireless communication system 100 supporting coding techniques for a neural network architecture according to aspects of the present disclosure. The wireless communication system 100 may include one or more base stations 105, one or more UEs 115, and a core network 130. In some examples, the wireless communication system 100 may be a Long Term Evolution (LTE) network, an LTE-Advanced (LTE-A) network, an LTE-A Pro network, or a New Radio (NR) network. In some examples, the wireless communication system 100 may support extended broadband communication, ultra-high reliability (e.g., mission-critical) communication, low-latency communication, communication with low-cost, low-complexity devices, or any combination thereof.

[0035] Base stations 105 may be distributed across a geographical area to form a wireless communication system 100 and may be devices of different forms or with different capabilities. Base stations 105 and UEs 115 may communicate wirelessly via one or more communication links 125. Each base station 105 may provide a coverage area 110 over which UEs 115 and base station 105 may establish one or more communication links 125. A coverage area 110 may be an example of a geographical area over which base stations 105 and UEs 115 may support the communication of signals according to one or more radio access technologies.

[0036] The UE115 may be distributed across the entire coverage area 110 of the wireless communication system 100, and each UE115 may be fixed, mobile, or both at different times. The UE115 may be devices of different forms or with different capabilities. Several exemplary UE115 are shown in Figure 1. The UE115 described herein may be capable of communicating with various types of devices, such as other UE115, base stations 105, or network equipment (e.g., core network nodes, relay devices, integrated access and backhaul (IAB) nodes, or other network equipment), as shown in Figure 1.

[0037] Base stations 105 can communicate with the core network 130, with each other, or both. For example, base stations 105 can interface with the core network 130 through one or more backhaul links 120 (e.g., via S1, N2, N3, or other interfaces). Base stations 105 can communicate with each other via the backhaul links 120 (e.g., via X2, Xn, or other interfaces) either directly (e.g., directly between base stations 105), indirectly (e.g., via the core network 130), or both. In some examples, the backhaul links 120 may be one or more wireless links, or may include one or more wireless links.

[0038] One or more of the base stations 105 described herein may include, or be referred to by, a base transceiver station, radio base station, access point, radio transceiver, NodeB, eNodeB (eNB), next-generation NodeB or giga-NodeB (any of which may be called gNB), Home NodeB, Home eNodeB, or other appropriate terms.

[0039] UE115 may include, or may be referred to as, a mobile device, wireless device, remote device, handheld device, or subscriber device, or any other appropriate term, and “device” may also be referred to as a unit, station, terminal, or client, among other things. UE115 may also include, or may be referred to as, a personal electronic device such as a mobile phone, personal digital assistant (PDA), tablet computer, laptop computer, or personal computer. In some examples, UE115 may also include, or may be referred to as, a wireless local loop (WLL) station, an Internet of Things (IoT) device, any Internet of Things (IoE) device, or a machine-type communications (MTC) device, among other things, which may be implemented in an appliance, or various items such as a vehicle, meter, etc.

[0040] The UE115 described herein may be capable of communicating with other UE115s that may act as relays, as shown in Figure 1, as well as with various types of devices, including, in particular, base stations 105 and network equipment, such as macro eNBs or gNBs, small cell eNBs or gNBs, or relay base stations.

[0041] UE115 and base station 105 may communicate wirelessly with each other via one or more communication links 125 over one or more carriers. The term “carrier” may refer to a set of radio frequency spectrum resources having a defined physical layer structure for supporting communication links 125. For example, a carrier used for communication link 125 may include a portion of the radio frequency spectrum band (e.g., a bandwidth portion (BWP)) operating according to one or more physical layer channels for a given radio access technology (e.g., LTE, LTE-A, LTE-A Pro, NR). Each physical layer channel may carry collected signaling (e.g., synchronization signals, system information), control signaling to coordinate operations with the carrier, user data, or other signaling. The wireless communication system 100 may support communication with UE115 using carrier aggregation or multi-carrier operation. UE115 may consist of multiple downlink component carriers and one or more uplink component carriers according to a carrier aggregation configuration. Carrier aggregation can be used with both frequency-division duplex (FDD) component carriers and time-division duplex (TDD) component carriers.

[0042] In some examples (for instance, in carrier aggregation configurations), a carrier may also have acquisition or control signaling to coordinate its operation with other carriers. A carrier may be associated with a frequency channel (e.g., an evolved universal mobile telecommunication system terrestrial radio access (E-UTRA) absolute radio frequency channel number (EARFCN)) and may be arranged according to a channel raster for discovery by the UE115. A carrier may operate in standalone mode, where initial acquisition and connection may be performed via the carrier by the UE115, or it may operate in non-standalone mode, where connection is anchored using different carriers (e.g., the same or different radio access technologies).

[0043] A communication link 125 shown in the wireless communication system 100 may include uplink transmissions from UE 115 to base station 105, or downlink transmissions from base station 105 to UE 115. The carrier may carry downlink communications or uplink communications (for example, in FDD mode), or may be configured to carry downlink communications and uplink communications (for example, in TDD mode).

[0044] A carrier may be associated with a specific bandwidth in the radio frequency spectrum, and in some examples, the carrier bandwidth may be referred to as the carrier or the “system bandwidth” of the wireless communication system 100. For example, the carrier bandwidth may be one of several determined bandwidths of a certain amount for the carrier of a particular radio access technology (e.g., 1.4, 3, 5, 10, 15, 20, 40, or 80 megahertz (MHz)). Devices of the wireless communication system 100 (e.g., base station 105, UE 115, or both) may have a hardware configuration that supports communication on a specific carrier bandwidth, or may be configurable to support communication on one carrier bandwidth out of a set of carrier bandwidths. In some examples, the wireless communication system 100 may include a base station 105 or UE 115 that supports simultaneous communication over carriers associated with multiple carrier bandwidths. In some examples, each UE 115 being served may be configured to operate on a portion (e.g., subband, BWP) or all of the carrier bandwidth.

[0045] The signal waveform transmitted on a carrier may consist of multiple subcarriers (for example, using multicarrier modulation (MCM) techniques such as orthogonal frequency division multiplexing (OFDM) or discrete Fourier transform spread OFDM (DFT-S-OFDM). In systems employing MCM techniques, a resource element may consist of one symbol period (e.g., the duration of one modulation symbol) and one subcarrier, where the symbol period and subcarrier spacing are inversely proportional. The amount of bits carried by each resource element may depend on the modulation scheme (e.g., the order of the modulation scheme, the coding rate of the modulation scheme, or both). Therefore, the more resource elements the UE115 receives, and the higher the order of the modulation scheme, the higher the data rate for the UE115 can be. Wireless communication resources may refer to a combination of radio frequency spectral resources, temporal resources, and spatial resources (e.g., spatial layers or beams), and the use of multiple spatial layers may further increase the data rate or data integrity for communication with the UE115.

[0046] One or more numerologies may be supported for a carrier, where the numerology may include a subcarrier interval (Δf) and a cyclic prefix. A carrier may be divided into one or more BWPs having the same or different numerologies. In some examples, UE115 may consist of multiple BWPs. In some examples, a single BWP for a carrier may be active at a given time, and communication for UE115 may be limited to one or more active BWPs.

[0047] The time interval for base station 105 or UE115 is, for example, T s = 1 / (Δf max ·N f It can refer to a sampling period of ) seconds, and can be expressed as a multiple of the basic time unit, where Δf max This can represent the maximum supported subcarrier interval, N fThis may represent the maximum supported Discrete Fourier Transform (DFT) size. The time interval of communication resources may be organized according to radio frames, each having a specified time length (e.g., 10 milliseconds (ms)). Each radio frame may be identified by a system frame number (SFN) (e.g., ranging from 0 to 1023).

[0048] Each frame may contain multiple sequentially numbered subframes or slots, each subframe or slot having the same time length. In some examples, a frame may be divided into subframes (e.g., in the time domain), and each subframe may be further divided into a certain number of slots. Alternatively, each frame may contain a variable number of slots, the number of slots may depend on the subcarrier interval. Each slot may contain a certain number of symbol periods (e.g., depending on the length of the cyclic prefix prepared for each symbol period). In some wireless communication systems 100, a slot may be further divided into a plurality of minislots, each containing one or more symbols. Except for the cyclic prefix, each symbol period may contain one or more (e.g., N) symbols. f It may include a sampling period of (1) units. The duration of the symbol period may depend on the subcarrier interval or the frequency bandwidth of the operation.

[0049] A subframe, slot, minislot, or symbol may be the smallest scheduling unit (e.g., in the time domain) of the wireless communication system 100, and may be called a transmit time interval (TTI). In some examples, the TTI length (e.g., the amount of symbol duration within the TTI) may be variable. Additionally or alternatively, the smallest scheduling unit of the wireless communication system 100 may be dynamically selected (e.g., within a burst of shortened TTIs (sTTIs)).

[0050] Physical channels may be multiplexed on the carrier according to various techniques. Physical control channels and physical data channels may be multiplexed on the downlink carrier using one or more of the following techniques: time-division multiplexing (TDM), frequency-division multiplexing (FDM), or hybrid TDM-FDM. A control region for a physical control channel (e.g., a control resource set (core set)) may be defined by the amount of symbol duration and may extend over the system bandwidth or a subset of the system bandwidth of the carrier. One or more control regions (e.g., core sets) may be configured for a set of UE115s. For example, one or more of the UE115s may monitor or search for control regions for control information according to one or more search space sets, each search space set may include one or more control channel candidates at one or more aggregation levels arranged in a cascaded manner. The aggregation level for control channel candidates may refer to the amount of control channel resources (e.g., control channel elements (CCEs)) associated with encoded information for a control information format having a given payload size. The search space set may include a common search space set configured for sending control information to multiple UE115s, and a UE-specific search space set for sending control information to a specific UE115.

[0051] Each base station 105 may provide communication coverage through one or more cells, such as macrocells, small cells, hotspots, or other types of cells, or any combination thereof. The term “cell” may refer to a logical communication entity used for communication with base station 105 (for example, on a carrier) and may be associated with an identifier for distinguishing neighboring cells (for example, a physical cell identifier (PCID), a virtual cell identifier (VCID), or other). In some examples, a cell may also refer to a geographical coverage area 110 or a portion of geographical coverage area 110 (for example, a sector) on which the logical communication entity operates. Such cells may span from smaller areas (for example, structures, subsets of structures) to larger areas, depending on various factors such as the capabilities of base station 105. For example, a cell may, among other things, be a building, a subset of a building, or external space between or overlapping with geographical coverage area 110.

[0052] Macrocells typically cover relatively large geographical areas (e.g., a radius of several kilometers) and can enable unrestricted access by UE115s subscribed to the services of a network provider that supports macrocells. Small cells may be associated with lower-power base stations 105 compared to macrocells, and small cells may operate in the same or different (e.g., licensed, unlicensed) frequency bands as macrocells. Small cells may provide unrestricted access to UE115s subscribed to the services of a network provider, or they may provide restricted access to UE115s associated with small cells (e.g., UE115s in a limited subscriber group (CSG), UE115s associated with users in a home or office). Base station 105 may support one or more cells and may support communication on one or more cells using one or more component carriers.

[0053] In some examples, a carrier may support multiple cells, and different cells may be configured according to different protocol types (e.g., MTC, Narrowband IoT (NB-IoT), Enhanced Mobile Broadband (eMBB)) that can provide access to different types of devices.

[0054] In some examples, base station 105 may be mobile and therefore provide communication coverage to a moving geographic coverage area 110. In some examples, different geographic coverage areas 110 associated with different technologies may overlap, but different geographic coverage areas 110 may be supported by the same base station 105. In other examples, overlapping geographic coverage areas 110 associated with different technologies may be supported by different base stations 105. The wireless communication system 100 may include, for example, heterogeneous networks in which different types of base stations 105 provide coverage to various geographic coverage areas 110 using the same or different radio access technologies.

[0055] The wireless communication system 100 may support synchronous or asynchronous operation. In synchronous operation, base stations 105 may have similar frame timings, and transmissions from different base stations 105 may be approximately time-coordinated. In asynchronous operation, base stations 105 may have different frame timings, and transmissions from different base stations 105 may, in some cases, not be time-coordinated. The techniques described herein may be used for either synchronous or asynchronous operation.

[0056] Some UE115s, such as MTC devices or IoT devices, may be low-cost or low-complexity devices that can provide automated communication between machines (e.g., via machine-to-machine (M2M) communication). M2M communication or MTC may refer to data communication technology that enables devices to communicate with each other or with base stations 105 without human intervention. In some examples, M2M communication or MTC may include communication from devices that incorporate sensors or meters to measure or capture information and relay that information to a central server or application program that utilizes such information or presents it to a human interacting with the application program. Some UE115s may be designed to collect information or enable automated behavior of machines or other devices. Examples of applications for MTC devices include smart metering, inventory monitoring, water level monitoring, equipment monitoring, healthcare monitoring, wildlife monitoring, weather and geological event monitoring, fleet management and tracking, remote security detection, physical access control, and transaction-based business billing.

[0057] Some UE115s may be configured to use power-saving operating modes, such as half-duplex communication (e.g., modes that support one-way communication via transmit or receive, but not simultaneous transmit and receive). In some examples, half-duplex communication may be performed at a reduced peak rate. Other power-saving techniques for the UE115 include entering a power-saving deep sleep mode when not engaged in active communication, operating over a limited bandwidth (e.g., according to narrowband communication), or a combination of these techniques. For example, some UE115s may be configured for operation using narrowband protocol types related to a defined portion or range within the carrier, within the carrier's guard band, or outside the carrier (e.g., a set of subcarriers or resource blocks (RBs)).

[0058] The wireless communication system 100 may be configured to support ultra-high reliability communication, low latency communication, or various combinations thereof. For example, the wireless communication system 100 may be configured to support ultra-high reliability low latency communication (URLLC) or mission-critical communication. The UE 115 may be designed to support ultra-high reliability, low latency, or critical functions (e.g., mission-critical functions). Ultra-high reliability communication may include private or group communication and may be supported by one or more mission-critical services such as mission-critical push-to-talk (MCPTT), mission-critical video (MCVideo), or mission-critical data (MCData). Support for mission-critical functions may include prioritizing services, and mission-critical services may be used for public safety or general commercial purposes. The terms ultra-high reliability, low latency, mission-critical, and ultra-high reliability low latency may be used interchangeably herein.

[0059] In some examples, UE115 may also be able to communicate directly with other UE115 via a device-to-device (D2D) communication link 135 (for example, using a peer-to-peer (P2P) protocol or a D2D protocol). One or more UE115s utilizing D2D communication may be within the geographical coverage area 110 of base station 105. Other UE115s in such a group may be outside the geographical coverage area 110 of base station 105, or in some cases may not be able to receive transmissions from base station 105. In some examples, a group of UE115s communicating via D2D communication may utilize a one-to-many (1:M) system, such that each UE115 transmits to any other UE115 in the group. In some examples, base station 105 facilitates the scheduling of resources for D2D communication. In other cases, D2D communication is performed between UE115s without the involvement of base station 105.

[0060] In some systems, the D2D communication link 135 may be an example of a communication channel between vehicles (e.g., UE 115), such as a side-link communication channel. In some examples, vehicles may communicate using vehicle-to-everything (V2X) communication, vehicle-to-vehicle (V2V) communication, or any combination thereof. Vehicles may signal information related to traffic conditions, signal scheduling, weather, safety, emergencies, or any other information related to the V2X system. In some examples, vehicles in a V2X system may communicate with roadside infrastructure such as roadside units, or with the network via one or more network nodes (e.g., base station 105) using vehicle-to-network (V2N) communication, or both.

[0061] The core network 130 may provide user authentication, access authorization, tracking, Internet Protocol (IP) connectivity, and other access, routing, or mobility functions. The core network 130 may be an advanced packet core (EPC) or a 5G core (5GC), which may include at least one control plane entity that manages access and mobility (e.g., a mobility management entity (MME), an access and mobility management function (AMF)), and at least one user plane entity that routes packets or interconnects to external networks (e.g., a serving gateway (S-GW), a packet data network (PDN) gateway (P-GW), or a user plane function (UPF)). The control plane entity may manage non-access layer (NAS) functions such as mobility, authentication, and bearer management for UE 115 serviced by base station 105 associated with the core network 130. User IP packets may be forwarded through user plane entities that may provide IP address allocation and other functions. A user plane entity may be connected to one or more network operators' IP services 150. These IP services may include access to the Internet, an intranet, an IP multimedia subsystem (IMS), or a packet-switched streaming service.

[0062] Some of the network devices, such as the base station 105, may include subcomponents such as an access network entity 140, which may be an example of an access node controller (ANC). Each access network entity 140 may communicate with the UE 115 through one or more other access network transmit entities 145, which may be called radio heads, smart radio heads, or transmit / receive points (TRPs). Each access network transmit entity 145 may include one or more antenna panels. In some configurations, the various functions of each access network entity 140 or base station 105 may be distributed across various network devices (e.g., radio heads and ANCs) or integrated into a single network device (e.g., base station 105).

[0063] The wireless communication system 100 may typically operate using one or more frequency bands in the range of 300 megahertz (MHz) to 300 gigahertz (GHz). Generally, the region from 300 MHz to 3 GHz is called the ultra-high frequency (UHF) region or decimeter band, as the wavelengths range from approximately 1 decimeter to 1 meter. While UHF waves may be blocked or redirected by building and environmental characteristics, their waves can penetrate structures well enough for a macrocell to service an indoor UE 115. Transmitting UHF waves may involve smaller antennas and shorter distances (e.g., less than 100 kilometers) compared to transmitting using lower frequencies and longer waves in the shortwave (HF) or very high frequency (VHF) portions of the spectrum below 300 MHz.

[0064] The wireless communication system 100 may also operate in the super high frequency (SHF) region, using a frequency band from 3 GHz to 30 GHz, also known as the centimeter band, or in the extremely high frequency (EHF) region of the spectrum, also known as the millimeter band (for example, from 30 GHz to 300 GHz). In some examples, the wireless communication system 100 may support millimeter-wave (mmW) communication between the UE 115 and the base station 105, and the EHF antennas of each device may be smaller and more densely spaced than UHF antennas. In some examples, this may facilitate the use of antenna arrays within the device. However, the propagation of EHF transmissions may be subject to greater atmospheric attenuation than SHF or UHF transmissions and may have shorter distances. The techniques disclosed herein may be employed across transmissions using one or more different frequency domains, and the specified use of bands across these frequency domains may vary by country or regulatory body.

[0065] The wireless communication system 100 can utilize both licensed and unlicensed radio frequency spectrum bands. For example, the wireless communication system 100 may employ License Assisted Access (LAA), LTE-Unlicensed (LTE-U) radio access technology, or NR technology in unlicensed bands such as the 5 GHz Industrial Scientific Medical (ISM) band. When operating in unlicensed radio frequency spectrum bands, devices such as base station 105 and UE 115 may employ carrier detection for conflict detection and avoidance. In some examples, operation in unlicensed bands may be based on a carrier aggregation configuration in conjunction with component carriers operating in licensed bands (e.g., LAA). Operation in unlicensed spectrums may include, among other things, downlink transmission, uplink transmission, P2P transmission, or D2D transmission.

[0066] Base station 105 or UE115 may be equipped with multiple antennas that can be used to employ techniques such as transmit diversity, receive diversity, multiple-input multiple-output (MIMO) communication, or beamforming. The antennas of base station 105 or UE115 may be located in one or more antenna arrays or antenna panels that can support MIMO operation or transmit beamforming or receive beamforming. For example, one or more base station antennas or antenna arrays may be placed together in an antenna assembly such as an antenna tower. In some examples, the antennas or antenna arrays associated with base station 105 may be located in diverse geographical locations. Base station 105 may have an antenna array having a certain number of rows and columns of antenna ports that base station 105 can use to support beamforming of communication with UE115. Similarly, UE115 may have one or more antenna arrays that can support various MIMO or beamforming operations. As an addition or alternative, an antenna panel may support radio frequency beamforming for signals transmitted through antenna ports.

[0067] A base station 105 or UE115 may use MIMO communication to enhance spectral efficiency by leveraging multipath signal propagation by transmitting or receiving multiple signals through different spatial layers. Such techniques are sometimes called spatial multiplexing. Multiple signals may be transmitted by a transmitting device through different antennas or different combinations of antennas. Similarly, multiple signals may be received by a receiving device through different antennas or different combinations of antennas. Each of the multiple signals may be called a separate spatial stream and may carry bits related to the same data stream (e.g., the same codeword) or different data streams (e.g., different codewords). Different spatial layers may be associated with different antenna ports used for channel measurement and reporting. MIMO techniques include single-user MIMO (SU-MIMO), where multiple spatial layers are transmitted to the same receiving device, and multi-user MIMO (MU-MIMO), where multiple spatial layers are transmitted to multiple devices.

[0068] Beamforming, sometimes called spatial filtering, directional transmission, or directional reception, is a signal processing technique that can be used in a transmitting or receiving device (e.g., base station 105, UE115) to shape or steer an antenna beam (e.g., transmit beam, receive beam) along a spatial path between the transmitting and receiving devices. Beamforming can be achieved by combining signals communicated through the antenna elements of an antenna array such that several signals propagating in a particular orientation relative to the antenna array undergo constructive interference, while other signals undergo destructive interference. Coordination of signals communicated through antenna elements may involve the transmitting or receiving device applying amplitude offset, phase offset, or both to the signals carried through the antenna elements associated with the device. Coordination associated with each antenna element may be defined by a beamforming weight set associated with a particular orientation (e.g., relative to the antenna array of the transmitting or receiving device, or to some other orientation).

[0069] The base station 105 or UE 115 may use beam sweeping techniques as part of its beamforming operation. For example, the base station 105 may use multiple antennas or antenna arrays (e.g., antenna panels) to perform beamforming operations for directional communication with the UE 115. Several signals (e.g., synchronization signals, reference signals, beam selection signals, or other control signals) may be transmitted multiple times by the base station 105 in different directions. For example, the base station 105 may transmit signals according to different beamforming weight sets associated with different transmission directions. Transmissions in different beam directions may be used to identify beam directions for later transmission or reception by the base station 105 (e.g., by a transmitting device such as the base station 105, or by a receiving device such as the UE 115).

[0070] The wireless communication system 100 may be a packet-based network operating according to a layered protocol stack. In the user plane, communication at the bearer or Packet Data Convergence Protocol (PDCP) layer may be IP-based. The Radio Link Control (RLC) layer may perform packet segmentation and reassembly for communication on logical channels. The Medium Access Control (MAC) layer may perform priority processing and multiplexing logical channels to transport channels. The MAC layer may also use error detection techniques, error correction techniques, or both to improve link efficiency by supporting retransmission at the MAC layer. In the control plane, the Radio Resource Control (RRC) protocol layer may establish, configure, and maintain RRC connections between the UE 115 and the base station 105 or core network 130, supporting radio bearers for user plane data. At the physical layer, transport channels may be mapped to physical channels.

[0071] UE115 and base station 105 may support data retransmission to increase the likelihood of successful data reception. Hybrid Automatic Retransmission Request (ARQ) feedback is one technique to increase the likelihood of data being correctly received on communication link 125. HARQ may include a combination of error detection (e.g., using cyclic redundancy check (CRC)), forward error correction (FEC), and retransmission (e.g., Automatic Retransmission Request (ARQ)). HARQ may improve throughput at the MAC layer under poor radio conditions (e.g., low signal-to-noise conditions). In some examples, devices may support same-slot HARQ feedback, where the device may provide HARQ feedback within a slot for data received in a previous symbol within a particular slot. In other cases, the device may provide HARQ feedback in a subsequent slot or according to some other time interval.

[0072] Machine learning, particularly deep machine learning, has become a popular tool in wireless communication systems to facilitate more efficient communication. Among its advantages, a machine learning model deployed in a UE115 can enable the UE115 to make decisions or take action (e.g., using prediction or regression or other objectives) without additional signaling from the base station 105 (e.g., the UE115 can make inferences about what action to take based on input or detected events). In addition, machine learning models can enable the UE115 to prepare transmissions for more efficient communication (e.g., using classification or compression or other objectives). Before a machine learning model can be deployed to a device, it can be prepared and trained (e.g., using a dataset).

[0073] Machine learning models using neural networks can be deployed on a device (e.g., UE115) for different applications (e.g., prediction, classification, compression, regression, or other objectives). For example, a machine learning model, once identified by a device, may include one or more parameters (e.g., a prepared dataset) that enable or support the corresponding application on the device. To build and further train a machine learning model, a network device may collect different datasets to identify the impact or effect of the dataset on different applications. That is, a machine learning model may be trained on one or more datasets using a neural network, and when real-world data is input to the machine learning model, the machine learning model may produce outputs based on the dataset.

[0074] In addition, the UE115 may perform several channel state measurements as part of its communication with the base station 105. For example, these measurements may include per-antenna port channel and interference measurements (e.g., channel state feedback), power measurements from serving cells and neighboring cells, inter-RAT measurements (e.g., from a WiFi network), and sensor measurements. These measurements may generate a large amount of data that will be sent to the base station 105 to assist the base station 105 in network management. In some cases, the UE115 may use machine learning models or neural networks to compress this large amount of data to be sent.

[0075] In some examples, the quantities measured by UE115 (e.g., in a 5G network) may depend on multiple parameters. For example, multiple parameters that may affect the measurement may include the measured results of the antenna design and placement in UE115 and their time-varying shielding, environmental parameters (e.g., the location, shadowing, presence and movement of reflectors in the vicinity of UE115 and / or base station 105, where the placement of reflectors may also create inter-tap correlations such as the resolution of a wide path / beam being divided into many multiple paths), loading to the cell (e.g., resulting in handoff), and the movement of the UE115 in question (e.g., changes in its orientation). To account for these types of changing parameters that may affect the measurement, UE115 may use, or be implemented by, a neural network that learns the dependence of the measured quantities on individual parameters, separates those measured quantities through various layers, and compresses the measurement results while reducing (e.g., minimizing) the compression loss. For example, UE115 may use a neural network to compress the measurement results and reduce the size of the transmission carrying the measurement results. However, further techniques are desired to improve the compression methods of measurement data using neural networks.

[0076] The wireless communication system 100 may support adding an additional layer to the compression operation when the UE 115 is compressing the data in the transmission before sending it to the base station 105. For example, the UE 115 may measure one or more channel conditions and then use a neural network to compress the measured data. In some examples, the UE 115 may quantize the output of the neural network to provide data that is more likely to be communicated. Subsequently, after the neural network compression, the UE 115 may add further layers to further compress the data, and the UE 115 may further compress and encode the measured data before sending the transmission to the base station using differential coding, entropy coding, or both. Differential coding may be an example of encoding the difference between two measured results rather than encoding the absolute value of the measured result. For example, differential coding may use the previous value, initial value, reconstructed value, initial reconstructed value, or a combination thereof to indicate the difference value. When the base station 105 then decodes the encoded, quantized, and compressed transmission from the UE 115, it may use differential decoding, entropy decoding, or both. In some examples, base station 105 may transmit indications of the differential and entropy coding (e.g., and parameters for each coding) that UE 115 should use.

[0077] Figure 2 shows an example of a wireless communication system 200 supporting coding techniques for a neural network architecture according to aspects of the present disclosure. In some examples, the wireless communication system 200 may implement, or be implemented by, aspects of the wireless communication system 100. For example, the wireless communication system 200 may include base stations 105-a and UEs 115-a, which may be examples of the corresponding base stations 105 and UEs 115 as described with reference to Figure 1. In addition, base stations 105-a and UEs 115-a may communicate over the resources of carrier 205 (for example, for downlink communication) and carrier 210 (for example, for uplink communication). Although shown as separate carriers, carriers 205 and 210 may include the same or different resources (e.g., time and frequency resources) for the corresponding transmissions.

[0078] As described herein, using the property that quantities are compressed (e.g., data in a dataset), UE115-a may use a neural network to perform step-by-step extraction and compression of each feature (e.g., dimension) that affects the quantity. For example, UE115-a may identify the features (e.g., dimension) to be compressed. In some examples, UE115-a may perform a certain type of operation in that dimension and a general operation in other dimensions. For example, UE115-a may use fully connected layers in the first dimension and convolutions (e.g., point-by-point convolutions) in the other dimensions. UE115-a may then perform extraction including additional stacked layers. For example, the additional stacked layers may include convolutional layers, fully connected layers, or other layers with or without activation (e.g., residual neural network (ResNet) layers). After extraction, the neural network used by UE115-a may compress the features. For example, a neural network may use convolutional layers, fully connected layers, or other types of layers for this compression. The UE115-a may repeat this process for subsequent features. After each feature has been extracted and compressed, the UE115-a may use one or more additional compression layers (e.g., convolutional layers, fully connected layers, or other types of layers). Using the techniques described herein, the UE115-a may then perform additional coding for final compression, such as differential coding, entropy coding, or both.

[0079] Differential coding may involve encoding the difference between two measurement results, rather than encoding the absolute value of the measurement result. For example, a first signal power measurement result obtained at a first time point may be compared with a second signal power measurement result obtained at a second time point. In differential coding, the difference between the first and second signal power measurement results may be communicated, rather than the absolute value of the second signal power measurement result. Since some measurement results do not change much over time, some differential data may use fewer bits than the absolute value of the data. In addition, entropy coding may involve compressing the amount of encoded bits to reduce the amount of bits used in transmission. In some examples, an entropy coding scheme may use symbols with an amount of bits inversely proportional to the probability of the symbol appearing. For example, symbols that are more likely to appear may be encoded using fewer bits than symbols that are less likely to appear.

[0080] In some examples, UE115-a may perform differential coding, entropy coding, or both to further compress the data after the quantization step associated with neural network processing. Additionally or alternatively, differential coding may be performed at the input stage of the compression operation, where bits representing the difference between the measurement results are first coded and then input to the neural network for processing. When base station 105-a decodes the compressed and coded transmission from UE115-a, it may use differential decoding, entropy decoding, or both.

[0081] Before performing these compression and encoding operations, base station 105-a may indicate that UE115-a should perform different compression and encoding operations. For example, base station 105-a may transmit an indication of encoding parameters 215 to UE115-a (e.g., via carrier 205). In some examples, encoding parameters 215 may include differential encoder parameters 220, entropy encoder parameters 225, or both. One or more encoder parameters may enable or indicate to UE115-a that it should use the corresponding encoder when compressing and encoding the data (e.g., measurement data) before transmitting it to base station 105-a. For example, UE115-a may detect or determine a dataset 230 that should be compressed and encoded based on encoding parameters 215 (e.g., including differential encoder parameters 220, entropy encoder parameters 225, or both). UE115-a may then transmit the compressed and encoded dataset 235 to base station 105-a (for example, via carrier 210) based on having performed a differential encoder operation (for example, according to differential encoder parameter 220), an entropy encoder operation (for example, according to entropy encoder parameter 225), or both. Base station 105-a may then decode the compressed and encoded dataset 235 based on a differential decoder operation, an entropy decoder operation, or both.

[0082] In some examples, the coding parameter 215 may indicate how UE115-a should perform the corresponding encoder operation. For example, according to the differential encoder parameter 220, base station 105-a may indicate which previous values ​​UE115-a should use to perform the differential encoder operation. In some implementations, UE115-a may use an initial value of some data (e.g., in an intracoded frame (I-frame)) to determine the differential value of that data in a later time instance (e.g., in a prediction frame (P-frame)). As an addition or alternative, UE115-a may use a previous reconstructed value of some data to determine the differential value of that data in a later time instance. In some examples, UE115-a may use an initial reconstructed value of some data to determine the differential value of that data in a later time instance. As an addition or alternative, base station 105-a may indicate how UE115-a should perform the entropy coding operation via the entropy encoder parameter 225. For example, the entropy encoder parameter 225 may indicate the type of entropy encoder operation to be performed (e.g., Huffman coding, arithmetic coding, Embedded Zerotree Wavelet (EZW) coding, Lempel-Ziv (LZ) entropy coding, etc.) and any additional parameters to enable the entropy encoder operation.

[0083] As described herein, the UE115-a may perform differential encoder operations, entropy encoder operations, or both on the output of a single-shot encoder. For example, the UE115-a may first encode dataset 230 using a single-shot encoder (e.g., an additional encoder other than the differential encoder and entropy encoder) to compress and encode dataset 230. In some examples, after dataset 230 has been encoded using a single-shot encoder, the UE115-a may quantize the encoded dataset (e.g., the encoded and compressed dataset). Quantizing the encoded dataset may involve storing or converting the values ​​in the encoded dataset with a bit width lower than the floating-point prediction (e.g., floating-point values ​​are converted to bits, so the values ​​in the encoded dataset are represented by integers rather than floating-point values). Once the encoded dataset is quantized (for example, the dataset is encoded, quantized, and compressed here), UE115-a may then perform a differential encoder operation, an entropy encoder operation, or both before transmitting the further encoded, quantized, and compressed dataset to base station 105-a. Additionally or alternatively, UE115-a may perform a differential encoder operation on a single-shot encoder at the input stage and then perform the remaining operations described herein.

[0084] In some examples, UE115-a may determine when to perform a differential encoder operation, an entropy encoder operation, or both, in relation to a single-shot encoder (for example, at the input or output stage), based on the corresponding indicated parameters for each encoder operation. For example, the differential encoder parameter 220 may indicate when UE115-a should perform a differential encoder operation (for example, at the input or output stage of a single-shot encoder operation), and the entropy encoder parameter 225 may indicate when UE115-a should perform an entropy encoder operation (for example, at the output stage of a single-shot encoder operation). Additionally or alternatively, UE115-a may be preconfigured regarding when these encoder operations should be performed.

[0085] In the example in Figure 2, UE115-a is shown performing different compression and encoding operations, but any encoding device (e.g., base station 105, TRP, another type of UE115, etc.) may compress and encode the dataset before transmitting it to additional devices by performing the techniques described herein. In addition, base station 105-a is shown receiving and decoding the compressed and encoded dataset 235 (e.g., encoded, quantized, and compressed dataset), but any decoding device (e.g., UE115, TRP, another type of base station 105, etc.) may decode the compressed and encoded dataset 235 by performing the techniques described herein.

[0086] Figures 3A and 3B show examples of compression procedures 300 and 301 supporting coding techniques for neural network architectures according to embodiments of the present disclosure, respectively. In some examples, compression procedures 300 and 301 may implement embodiments of wireless communication system 100, wireless communication system 200, or both, or may be implemented by one or the other. For example, an coding device (e.g., UE115 or an additional coding device) may be configured to perform one or more operations on samples (e.g., data) received through one or more antennas of the coding device to compress the samples. Subsequently, a decoding device (e.g., base station 105 or an additional decoding device) may be configured to decode the compressed samples to determine information such as channel state feedback.

[0087] In some examples, the encoding device may identify the features to be compressed. For example, the encoding device may perform a first type of operation in a first dimension related to the features to be compressed. The encoding device may perform a second type of operation in other dimensions (e.g., in all other dimensions). For example, the encoding device may perform a fully connected operation in the first dimension and a convolution (e.g., point-by-point convolution) in the other dimensions. The identified features to be compressed are input 305 for the compression procedure 300 (e.g., h(t) or h in It may be part of ). The difference operation shown for compression procedure 300 may include multiple neural network layers and / or operations. The neural networks of the encoding device and decoding device may be formed by a concatenation of one or more of the operations shown.

[0088] The encoding device may perform spatial feature extraction 310 on the data (e.g., input 305). Subsequently, the encoding device may perform tap region feature extraction 315 on the data. In some examples, the encoding device may perform tap region feature extraction before performing spatial feature extraction. In addition, the extraction operation may include multiple operations. For example, the multiple operations may include one or more convolution operations, one or more fully connected operations, etc., which may be activated or deactivated. In some examples, the extraction operation may include one or more ResNet operations.

[0089] After performing spatial feature extraction 310 and tapped region feature extraction 315 (e.g., extraction operations), the encoding device may compress one or more extracted features using feature compression 320. In some examples, feature compression 320 (e.g., compression operations) may include one or more operations, such as one or more convolution operations or one or more fully connected operations. After compression, the number of bits in the output may be less than the number of bits in the input.

[0090] Next, the encoding device may perform quantization 325 (e.g., a quantization operation). In some examples, the encoding device may perform quantization 325 after flattening the output of the compression operation, and / or after flattening the output and performing a fully connected operation. The use of spatial feature extraction 310, tap region feature extraction 315, and feature compression 320 may represent the encoding device performing or using a single-shot encoder. The output of the single-shot encoder may then be passed through quantization 325 before the encoding device sends the encoded data to the decoding device.

[0091] Therefore, the decoding device can then perform feature decompression 330 based on having received the encoded data. In addition, the decoding device can perform tapped region feature reconstruction 335 and spatial feature reconstruction 340 to produce an output 345 (for example, h) corresponding to the input 305. outAlternatively, it may produce h(t). In some examples, the decoding device may perform spatial feature reconstruction 340 before performing tap region feature reconstruction 335. After the reconstruction operation, the decoding device may output the reconstructed version via output 345 based on the input 305 from the encoding device. Using feature decompression 330, tap region feature reconstruction 335, and spatial feature reconstruction 340 may represent performing or using a single-shot encoder in the decoding device.

[0092] In some cases, the decoding device may perform operations in the reverse order of those performed by the encoding device. For example, if the encoding device follows operation (A, B, C, D), the decoding device may follow the reverse operation (D, C, B, A). In addition, the decoding device may perform operations that are perfectly symmetric to those of the encoding device. The use of these symmetric operations can reduce the amount of bits used for neural network construction in the encoding device. As an addition or alternative, the decoding device may perform additional operations (e.g., convolution, fully connected operations, ResNet operations, etc.) in addition to those performed by the encoding device. That is, the decoding device may perform operations that are asymmetric to those of the encoding device.

[0093] Based on the fact that the encoding device encodes the dataset using a neural network for uplink communication, the encoding device (e.g., UE115) may transmit the measurement results (e.g., channel state feedback) along with a reduced payload. This reduced payload may save network resources that might otherwise have been used to transmit the complete dataset, which would otherwise be sampled by the encoding device. As illustrated herein with reference to Figure 2, the encoding device may perform additional encoding operations using additional layers to further compress the data to be transmitted to the decoding device.

[0094] For example, to further compress the feedback 370 sent to the decoding device, as shown in relation to compression procedure 301, the encoding device may apply or use a differential encoder 360, an entropy encoder 365, or both, to the output of the single-shot encoder. The feedback 370 may be an example of information transmitted over an air interface (or some other medium) between the transmitting and receiving devices. In addition, the decoding device may also use an entropy decoder 375, a differential decoder 380, or both, before performing the single-shot decoder operation described in relation to compression procedure 300 in Figure 3A. Additional or alternative, though not shown, the encoding device may apply or use a differential encoder 360 to the single-shot encoder at the input stage, and the decoding device may apply or use a differential decoder 380 to the output of the single-shot decoder. In some examples, compression procedure 301 may be called a time compression operation or procedure.

[0095] As shown in the example in Figure 3B, the encoding device may take in input 305 and pass input 305 through an encoder neural network 350. In some examples, the encoder neural network 350 may correspond to spatial feature extraction 310, tap region feature extraction 315, and feature compression 320, as described in relation to Figure 3A (e.g., a single-shot encoder). After input 305 has passed through the encoder neural network 350, the encoding device may perform a quantization operation (e.g., quantization 325, as described in relation to Figure 3A) using a quantizer 355. Subsequently, the encoding device may use or apply a difference encoder 360, an entropy encoder 365, or both, to the quantized output of the encoder neural network 350. In some examples, the difference encoder 360 and the entropy encoder 365 may be performed as described with reference to Figure 2. The encoding device may then send feedback 370 to the decoding device, which represents the encoded, quantized, and compressed dataset.

[0096] The decoding device may receive feedback 370 and use or apply an entropy decoder 375, a difference decoder 380, or both to partially decode or decode the encoded, quantized, and compressed dataset received with feedback 370. After using or applying the entropy decoder 375, the difference decoder 380, or both, the decoding device may then pass the partially decoded and decoded dataset through a decoder neural network 385 to produce an output 345. In some examples, the decoder neural network 385 may include feature decompression 330, tapped region feature reconstruction 335, and spatial feature reconstruction 340 (for example, and any additional layers), as described in relation to Figure 3A.

[0097] Figure 4 shows an example of a compression configuration 400 supporting an encoding technique for a neural network architecture according to an aspect of the present disclosure. In some examples, the compression configuration 400 may implement an aspect of wireless communication system 100, wireless communication system 200, or both, or may be implemented by one or the other. For example, an encoding device (e.g., UE115 or an additional encoding device) may be configured to perform one or more operations on samples (e.g., data) received through one or more antennas of the encoding device to compress the samples. Subsequently, a decoding device (e.g., base station 105 or an additional decoding device) may be configured to decode the compressed samples to determine information such as channel state feedback. In some examples, the compression configuration 400 may represent an example of a channel state feedback compression operation or procedure using entropy coding.

[0098] The encoding device may identify an input 405 for performing a compression configuration 400. For example, the input 405 may contain a dataset to be compressed and encoded (e.g., multiple measurements such as channel state feedback measurements). In some examples, the encoding device may receive samples from an antenna, and the input 405 represents a 64x64 dimensional dataset received from that antenna, based on the antenna size, the number of samples per antenna, and the tap features. A 64x64 dimensional dataset may represent one non-restrictive example for the input 405. The encoding device may pass or let the input 405 through a first convolutional layer 410. For example, the first convolutional layer 410 may represent an initial layer for spatial and short-time (tap) feature extraction using a one-dimensional convolution (e.g., Conv1D). The first convolutional layer 410 (e.g., an additional convolutional layer) could be a fully connected layer (e.g., in an antenna) and a simple convolution with a small kernel size (e.g., 3) in the tap dimension (for extracting short-tap features). The output from such a 64 × W one-dimensional convolution operation could be a W × 64 matrix.

[0099] After the first convolutional layer 410, the encoding device may perform one or more ResNet operations using one or more corresponding ResNet blocks, such as ResNet block 415 and ResNet block 420. One or more ResNet operations may further refine the spatial and / or temporal features. In some examples, a ResNet operation may include multiple operations related to the features. For example, a ResNet operation may include multiple (e.g., three) one-dimensional convolutions, skip connections (e.g., between the input and output of a ResNet to avoid applying one-dimensional convolutions), an addition operation of a path through multiple one-dimensional convolutions and a path through skip connections, or additional operations. In some examples, multiple one-dimensional convolution operations may include a W×256 convolution with kernel size 3, the output of which is fed into a batch normalization (BN) layer, followed by normalized linear unit (ReLU) activation (e.g., LeakyReLU activation), which produces a 256×64 output dataset, a 256×512 convolution with kernel size 3, the output of which is fed into a BN layer, followed by ReLU activation, which produces a 512×64 output dataset, a 512×W convolution with kernel size 3, which outputs a W×64 BN dataset. The output from one or more ResNet operations may be a W×64 matrix.

[0100] The encoding device may use an additional convolutional layer 425 to perform a W×V convolution on the outputs from one or more ResNet operations. The W×V convolution may include point-by-point (e.g., tap-by-tap) convolutions. The W×V convolution may compress the spatial features to a reduced dimension for each tap. The W×V convolution may have W feature inputs and V feature outputs. The output from the W×V convolution may be a V×64 matrix.

[0101] The encoding device may perform a flattening operation 430 to flatten a V×64 matrix into a 64V element vector. The encoding device may use a fully connected layer 435 to perform a 64V×M fully connected operation to further compress the spacetime feature dataset into a low-dimensional vector of size M for wireless transmission to the decoding device. The encoding device may perform quantization 440 before wireless transmission of the low-dimensional vector of size M to map the transmission samples to discrete values ​​of the low-dimensional vector of size M. After quantization 440, the encoding device may use or apply an entropy encoder 445 to further compress the output of quantization 440.

[0102] The decoding device may receive entropy-encoded data and decode the encoded transmission using or applying the entropy decoder 450. In addition, the decoding device may use a fully connected layer 455 to perform an M×64V fully connected operation to decompose a low-dimensional vector of size M into a spatial-time feature dataset. The decoding device may perform a reshaping operation 460 to reshape the 64V element vector into a 2-dimensional V×64 matrix. The decoding device may use a convolutional layer 465 to perform a V×W (kernel size 1) convolution on the output from the reshaping operation 460. The V×W convolution may include point-by-point (e.g., tap-by-tap) convolutions. In some examples, the V×W convolution may decompose spatial features from reduced dimensions for each tap. The V×W convolution may have V feature inputs and W feature outputs. The output from the V×W convolution may be a W×64 matrix.

[0103] The decoding device may then perform one or more ResNet operations using one or more corresponding ResNet blocks, such as ResNet block 470 and ResNet block 475. These one or more ResNet operations may further decode spatial features, temporal features, or both. In some examples, a ResNet operation may include multiple (e.g., three) one-dimensional convolutions, skip connections (e.g., to avoid applying one-dimensional convolutions), summation of paths through multiple convolutions and paths through skip connections, or additional operations. The output from one or more ResNet operations may be a W×64 matrix.

[0104] The decoding device may then use the convolutional layer 480 to perform spatial and temporal feature reconstruction. In some examples, spatial and temporal feature reconstruction can be achieved through the use of a one-dimensional convolution operation, which is a simple convolution that is fully connected in the spatial dimension (to reconstruct spatial features) and has a small kernel size (e.g., 3) in the tap dimension (to reconstruct short tap features). The output from the 64×W convolution operation may be a 64×64 matrix and may be represented by output 485. In some examples, the values ​​of M, W, and / or V may be configurable to adjust feature weights, payload size, etc. In addition, while a 64×64 dimensional dataset (e.g., a 64×64 matrix) is illustrated in the example of Figure 4 for the corresponding operation equivalent to a 64×64 dimensional dataset, datasets of any size can be used to perform the operations described herein.

[0105] Figure 5 shows an example of a compression and coding configuration 500 supporting coding techniques for a neural network architecture according to aspects of the present disclosure. In some examples, the compression and coding configuration 500 may implement an embodiment of a wireless communication system 100, a wireless communication system 200, or both, or may be implemented by one embodiment or the other. For example, a coding device (e.g., UE 115 or an additional coding device) may be configured to perform one or more operations on an input 505 of samples (e.g., data) received through one or more antennas of the coding device to compress the samples. Subsequently, a decoding device (e.g., base station 105 or an additional decoding device) may be configured to decode the compressed samples based on the input 505 to determine an output 530 of information such as channel state feedback. In some examples, the compression and coding configuration 500 may represent a compression and coding procedure that includes a differential encoder 510, an entropy encoder 515, or both at the output of an encoder (e.g., a single-shot encoder, an encoder neural network, etc.). By using or applying a differential encoder 510 at the output of the encoder, the coding device can reduce or limit error propagation.

[0106] The symbolization device may take in the input 505 and first encode, compress, and quantize the input 505 using one or more operations of an encoder, such as one or more convolutional layers, one or more ResNet blocks, one or more fully connected layers, and a quantizer, as described in relation to FIG. 4. After being quantized (e.g., at the output of the encoder), the symbolization device may use or apply a differential encoder 510. As described in relation to FIG. 2, the differential encoder 510 may encode data using previous values of the data. In some examples, a decoding device or network device may indicate to the symbolization device which previous values to use for the differential encoder 510 (e.g., via one or more parameters for the differential encoding operation). For example, the symbolization device may determine a difference value for that data at a later time instance (e.g., in a P frame) using an initial value of the data (e.g., in an I frame), may determine a difference value for that data at a later time instance using a previously reconstructed value of the data, may determine a difference value for that data at a later time instance using an initially reconstructed value of the data, or a combination thereof.

[0107] For example, when determining a difference value for that data at a later time instance using an initial value of the data, the symbolization device may calculate a P subframe (e.g., a P frame) with respect to an I frame. A given difference value for a time instance (N - 1) may be given by Equation 1. (x N-1 - x0) (1) x0 may represent an initial value of the data (e.g., determined in an I frame), and x N-1 may represent a given value of that data at time N - 1 (e.g., determined in a P frame). The symbolization device calculates the difference value for the data given by Equation 1 [Number] You can divide by x, where E represents the error propagation value and Δ is a given value of the data (for example, x N-1 This represents the difference between the initial value of the data (for example, x0).

number

number

[0108] On the decoding device side, the decoding device is E P / D P The device can receive a representation of the data provided by the device. Subsequently, the decoding device receives the received representation of the data (for example, an encoded version of the data).

number

number

[0109] The decoding device then reconstructs the value relative to the initial value of the data determined by the decoding device (for example, in an I-frame).

number

number

number

[0110] As an addition or alternative, the encoding device may use P-subframe construction and reconstruction based on the channel being reconstructed when using or applying the differential encoder 510 (for example, based on one or more parameters signaled to the encoding device for the differential encoder 510). Using P-subframe construction and reconstruction based on the channel being reconstructed can reduce or prevent error propagation. In some examples, the encoding device may determine the reconstructed value for a given data value. For example, the encoding device may

number

number

number

number

[0111] Next, for a given P-frame following an I-frame (e.g., P-frame 1, P-frame 2, etc.), the encoding device may encode the difference value for the value in the given P-frame using the reconstructed value for the previous value. For example, for P-frame N-1 (e.g., at time N-1), the encoding device may determine the difference value for the data at time N-1 based on Equation 3.

number

[0112] The encoding device can then perform operations similar to those previously described (for example, the difference value for the data given by Equation 3).

number

number

number

number

number

number

[0113] Therefore, the decoding device may receive an encoded, quantized, and compressed dataset from the encoding device and partially decode the encoded, quantized, and compressed dataset using the entropy decoder 520 and the difference decoder 525. In addition, the decoding device may determine and obtain the output 530 corresponding to the input 505 using one or more fully connected layers, one or more convolutional layers, and one or more ResNet blocks, as described in relation to Figure 3A.

[0114] Figure 6 shows an example of a compression and coding configuration 600 supporting coding techniques for a neural network architecture according to aspects of the present disclosure. In some examples, the compression and coding configuration 600 may implement an embodiment of a wireless communication system 100, a wireless communication system 200, or both, or may be implemented by any embodiment. For example, a coding device (e.g., UE 115 or an additional coding device) may be configured to perform one or more operations on an input 605 of samples (e.g., data) received through one or more antennas of the coding device to compress the samples. Subsequently, a decoding device (e.g., base station 105 or an additional decoding device) may be configured to decode the compressed samples based on the input 605 to determine an output 640 of information such as channel state feedback. In some examples, the compression and coding configuration 600 may represent a compression and coding procedure that includes a differential encoder 610 at the input of an encoder (e.g., a single-shot encoder, an encoder neural network, etc.) and an entropy encoder 620 at the output of the encoder.

[0115] Along with the differential encoder 610, the encoding device may use I-frames and one or more P-frames. I-frames may appear once every N subframes, and P-frames may appear for the remaining N-1 subframes. In some examples, the encoding device may use two neural networks for the differential encoder 610, such as a first neural network for I-frames and a second neural network for P-frames. The encoding device may determine the initial values ​​of the data in the I-frames based on the techniques described in relation to Figure 5 (for example,

number

[0116] The decoding device then determines the reconstructed value for each time instance of the data based on this difference, as illustrated in relation to Figure 5 (for example,

number

number

number

number

[0117] In addition to using or applying the differential encoder 610 at the encoder input, the encoding device may normalize the differentially encoded data by performing a normalization operation 615 before further encoding. Furthermore, the decoding device may perform a normalization cancellation operation 630 before using or applying the differential decoder 635. In some examples, the decoding device may also include an entropy decoder 625 that decodes the output of the entropy encoder 620.

[0118] Figure 7 shows an example of a process flow 700 supporting an encoding technique for a neural network architecture according to an aspect of the present disclosure. In some examples, the process flow 700 may implement an aspect of wireless communication system 100, wireless communication system 200, or both, or may be implemented in any one of these aspects. For example, the process flow 700 may include base stations 105-b and UE115-b, which may represent examples of corresponding base stations 105 and UE115, respectively, as described in relation to Figures 1 to 6.

[0119] In the following description of process flow 700, the operations between UE115-b and base station 105-b may be performed in a different order or at different times. Some operations may be omitted from process flow 700, or other operations may be added to process flow 700. Although UE115-b and base station 105-b are shown to perform a certain amount of operations in process flow 700, any wireless device may perform the indicated operations.

[0120] In 705, UE115-b may receive an indication of one or more coding operations to be used to encode the compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. For example, UE115-b may receive one or more parameters corresponding to one or more coding operations, each of which corresponds to each of the one or more coding operations, and the quantized and compressed dataset is encoded based on the one or more parameters.

[0121] In 710, the UE115-b may encode the dataset using the neural network used in or utilized by the UE115-b to generate a compressed dataset. In 715, the UE115-b may quantize the compressed dataset encoded by the neural network.

[0122] In 720, UE115-b may encode a quantized and compressed dataset based on the reception of an indication of one or more coding operations. In some examples, UE115-b may encode a quantized and compressed dataset using an entropy coding operation. For example, an entropy coding operation may involve encoding a compressed dataset using one or more symbols having lengths that vary based on the probability of the symbols appearing.

[0123] As an addition or alternative, UE115-b may encode a quantized and compressed dataset using differential coding operations. For example, differential coding operations may involve encoding the amount of data in a compressed dataset based on a previous value for that amount of data. In some examples, UE115-b may quantize a compressed dataset encoded by a neural network, and then determine the difference between a first value of data in the quantized and compressed dataset in a first time instance and a second value of data in a second time instance after the first time instance, where the difference value is determined based on one or more coding operation markings (for example, these markings indicate initial values ​​that UE115-b should use when determining the difference value), and the quantized and compressed dataset is encoded based on that difference value. As an addition or alternative, UE115-b may, after quantizing a compressed dataset encoded by a neural network, determine the difference between a first reconstructed value of the data in the quantized and compressed dataset in a first time instance and a second reconstructed value of the data in a second time instance following the first time instance, the difference value being determined based on one or more coding operation markings (for example, these markings indicate the previous reconstructed value that UE115-b should use when determining the difference value), and the quantized and compressed dataset being encoded based on that difference value.

[0124] In some examples, when performing a differential coding operation, UE115-b may, after quantizing the compressed dataset coded by the neural network, determine an initial reconstruction value for the data in the quantized and compressed dataset in the initial time instance related to coding the dataset. Subsequently, UE115-b may determine the difference between the initial reconstruction value for the data and additional reconstruction values ​​in additional time instances after the initial time instance, after quantization, where the difference value is determined based on one or more coding operation markings (for example, this marking indicates the initial reconstructed value that UE115-b should use when determining the difference value), and the quantized and compressed dataset is coded based on that difference value.

[0125] In 725, UE115-b may, after encoding a compressed dataset based on one or more encoding operations, transmit the encoded, quantized, and compressed dataset (e.g., the compressed and encoded dataset after a single-shot encoder) to a second device (e.g., base station 105-b). In some examples, the encoded, quantized, and compressed dataset may include differential values ​​to the data in the dataset, based on the initial values ​​of the data, previous reconstruction values ​​of the data, initial reconstruction values ​​of the data, or a combination thereof.

[0126] In 730, base station 105-b may decode the encoded, quantized, and compressed dataset received from UE 115-b. For example, base station 105-b may first decode the encoded, quantized, and compressed dataset and generate a compressed dataset based on one or more encoding operations. In addition, base station 105-b may then decode the compressed dataset using a neural network (for example, in or utilized by base station 105-b) and generate a dataset based on the decoding of the encoded, quantized, and compressed dataset.

[0127] Figure 8 shows a block diagram 800 of a device 805 supporting coding techniques for a neural network architecture according to an aspect of this disclosure. Device 805 may be an example of an aspect of UE115 as described herein. Device 805 may include a receiver 810, a transmitter 815, and a communications manager 820. Device 805 may also include a processor. Each of these components may communicate with one another (for example, via one or more buses).

[0128] Receiver 810 may provide means for receiving information such as packets, user data, control information, or any combination thereof related to various information channels (e.g., control channels, data channels, and information channels relating to coding techniques for neural network architectures). The information may be passed to other components of device 805. Receiver 810 may utilize a single antenna or a set of multiple antennas.

[0129] The transmitter 815 may provide a means for transmitting signals generated by other components of device 805. For example, the transmitter 815 may transmit information such as packets, user data, control information, or any combination thereof related to various information channels (e.g., control channels, data channels, and information channels relating to coding techniques for neural network architectures). In some examples, the transmitter 815 may be placed alongside the receiver 810 in the transceiver module. The transmitter 815 may utilize a single antenna or a set of multiple antennas.

[0130] The communication manager 820, receiver 810, transmitter 815, or various combinations thereof or various components thereof may be examples of means for performing various aspects of coding techniques for neural network architectures such as those described herein. For example, the communication manager 820, receiver 810, transmitter 815, or various combinations thereof or components thereof may support methods for performing one or more of the functions described herein.

[0131] In some examples, the communications manager 820, the receiver 810, the transmitter 815, or various combinations or components thereof may be implemented in hardware (for example, in a communications management circuit). The hardware may include a processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof configured as a means for performing or otherwise supporting the functions described herein. In some examples, a processor and memory coupled to the processor may be configured to perform one or more of the functions described herein (for example, by the processor executing instructions stored in the memory).

[0132] As an addition or alternative, in some examples, the communications manager 820, receiver 810, transmitter 815, or various combinations or components thereof may be implemented in code executed by a processor (for example, as communications management software or firmware). When implemented in code executed by a processor, the functions of the communications manager 820, receiver 810, transmitter 815, or various combinations or components thereof may be performed by a general-purpose processor, DSP, central processing unit (CPU), ASIC, FPGA, or any combination thereof or other programmable logic device (for example, configured as a means for performing or otherwise supporting the functions described herein).

[0133] In some examples, the communications manager 820 may be configured to perform various operations (e.g., receiving, monitoring, transmitting) using or otherwise cooperating with the receiver 810, the transmitter 815, or both. For example, the communications manager 820 may receive information from the receiver 810 and transmit information to the transmitter 815, or be integrated with the receiver 810, the transmitter 815, or both to receive information, transmit information, or perform various other operations as described herein.

[0134] The communication manager 820 may support wireless communication in the UE in accordance with the examples disclosed herein. For example, the communication manager 820 may be configured as a means for receiving, or otherwise supporting, an indication of one or more coding operations to be used to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The communication manager 820 may be configured as a means for encoding a dataset to generate a compressed dataset by a neural network, or otherwise supporting, this. The communication manager 820 may be configured as a means for quantizing a compressed dataset encoded by a neural network, or otherwise supporting, this. The communication manager 820 may be configured as a means for encoding a quantized and compressed dataset based on having received an indication of one or more coding operations, or otherwise supporting, this. The communication manager 820 may be configured as a means for transmitting an encoded, quantized, and compressed dataset to a second device after encoding the compressed dataset based on one or more coding operations, or otherwise supporting, this.

[0135] By including or configuring the communications manager 820 in accordance with the examples described herein, the device 805 (e.g., a processor controlling the receiver 810, transmitter 815, communications manager 820, or a combination thereof, or otherwise coupled thereto) may support techniques for more efficient use of communications resources. For example, by using differential encoder operations, entropy encoder operations, or both, the communications manager 820 may further compress the dataset for transmission to a second device, thereby reducing signaling overhead and using fewer communications resources to carry the compressed dataset.

[0136] Figure 9 shows a block diagram 900 of a device 905 supporting coding techniques for a neural network architecture according to an aspect of this disclosure. Device 905 may be an example of an aspect of device 805 or UE115 as described herein. Device 905 may include a receiver 910, a transmitter 915, and a communications manager 920. Device 905 may also include a processor. Each of these components may communicate with one another (for example, via one or more buses).

[0137] The receiver 910 may provide means for receiving information such as packets, user data, control information, or any combination thereof related to various information channels (e.g., control channels, data channels, and information channels relating to coding techniques for neural network architectures). The information may be passed to other components of device 905. The receiver 910 may utilize a single antenna or a set of multiple antennas.

[0138] The transmitter 915 may provide means for transmitting signals generated by other components of device 905. For example, the transmitter 915 may transmit information such as packets, user data, control information, or any combination thereof related to various information channels (e.g., control channels, data channels, and information channels relating to coding techniques for neural network architectures). In some examples, the transmitter 915 may be placed alongside the receiver 910 in the transceiver module. The transmitter 915 may utilize a single antenna or a set of multiple antennas.

[0139] Device 905 or its various components may be examples of means for performing various embodiments of coding techniques for neural network architectures as described herein. For example, communication manager 920 may include an encoder indicator component 925, a neural network encoder component 930, a quantization component 935, an encoder component 940, a coded dataset transmission component 945, or any combination thereof. Communication manager 920 may be an example of an embodiment of communication manager 820 as described herein. In some examples, communication manager 920 or its various components may be configured to perform various operations (e.g., receiving, monitoring, transmitting) using or otherwise cooperating with receiver 910, transmitter 915, or both. For example, communication manager 920 may receive information from receiver 910 and transmit information to transmitter 915, or be integrated with receiver 910, transmitter 915, or both to receive information, transmit information, or perform various other operations as described herein.

[0140] The communication manager 920 may support wireless communication in the UE according to the examples disclosed herein. The encoder indicator component 925 may be configured as a means for receiving, or otherwise supporting, an indicator of one or more coding operations to be used to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The neural network encoder component 930 may be configured as a means for encoding a dataset by a neural network to generate a compressed dataset, or otherwise supporting, it. The quantization component 935 may be configured as a means for quantizing a compressed dataset encoded by a neural network, or otherwise supporting, it. The encoder component 940 may be configured as a means for encoding a quantized and compressed dataset based on having received an indicator of one or more coding operations, or otherwise supporting, it. The encoded data transmission component 945 may be configured as a means for transmitting an encoded, quantized, and compressed dataset to a second device after encoding the compressed dataset based on one or more coding operations, or otherwise supporting, it.

[0141] Figure 10 shows a block diagram 1000 of a communications manager 1020 supporting coding techniques for a neural network architecture according to an aspect of this disclosure. Communications manager 1020 may be an example of communications manager 820, communications manager 920, or both, as described herein. Communications manager 1020 or its various components may be examples of means for performing various aspects of coding techniques for a neural network architecture as described herein. For example, communications manager 1020 may include an encoder marking component 1025, a neural network encoder component 1030, a quantization component 1035, an encoder component 1040, a coded dataset transmission component 1045, a differential encoder component 1050, an entropy encoder component 1055, or any combination thereof. Each of these components may communicate with one another directly or indirectly (for example, via one or more buses).

[0142] The communication manager 1020 may support wireless communication in the UE according to the examples disclosed herein. The encoder marking component 1025 is configured as a means for receiving, or otherwise supporting, a marking of one or more coding operations to be used to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The neural network encoder component 1030 is configured as a means for encoding a dataset by a neural network to generate a compressed dataset, or otherwise supporting, a means for quantization of a compressed dataset encoded by a neural network. The quantization component 1035 is configured as a means for quantizing a compressed dataset encoded by a neural network, or otherwise supporting, a means for encoding a quantized and compressed dataset based on having received a marking of one or more coding operations. The encoded data transmission component 1045 may be configured, or otherwise support, means for transmitting the encoded, quantized, and compressed dataset to a second device after encoding the compressed dataset based on one or more encoding operations.

[0143] In some examples, to support receiving indications for one or more coding operations, the encoder indication component 1025 may be configured as a means for receiving one or more parameters corresponding to one or more coding operations, or may support it otherwise, each of the one or more parameters corresponding to each coding operation of the one or more coding operations, and the quantized and compressed dataset is coded based on the one or more parameters.

[0144] In some examples, to support the encoding of quantized and compressed datasets, the differential encoder component 1050 may be configured as a means for encoding the quantized and compressed dataset using differential encoding operations after encoding the dataset using a neural network, or it may support this in other ways.

[0145] In some examples, to support the coding of quantized and compressed datasets, the entropy encoder component 1055 may be configured, or otherwise support, to code the quantized and compressed dataset using an entropy coding operation after coding the dataset using a neural network.

[0146] In some examples, to support the encoding of a quantized and compressed dataset, the difference encoder component 1050 may be configured, or otherwise supported, as a means for determining, after quantizing the compressed dataset to be encoded by a neural network, the difference value between a first value of data in the quantized and compressed dataset in an initial time instance and a second value of data in a second time instance after the initial time instance, wherein the difference value is determined based on the indication of one or more encoding operations, and the quantized and compressed dataset is encoded based on the difference value.

[0147] In some examples, to support the encoding of a quantized and compressed dataset, the difference encoder component 1050 may be configured, or otherwise support, a means for determining, after quantizing the compressed dataset to be encoded by a neural network, the difference between a first reconstructed value of the data in the quantized and compressed dataset in a first time instance and a second reconstructed value of the data in a second time instance following the first time instance, wherein the difference value is determined based on one or more coding operation markings, and the quantized and compressed dataset is encoded based on the difference value.

[0148] In some examples, to support encoding a quantized and compressed dataset, the differential encoder component 1050 may be configured, or otherwise supported, as a means for determining an initial reconstruction value for the data in the quantized and compressed dataset in an initial time instance related to encoding the dataset, after the compressed dataset encoded by the neural network has been quantized. In some examples, to support encoding a quantized and compressed dataset, the differential encoder component 1050 may be configured, or otherwise supported, as a means for determining a difference between an additional reconstruction value for the data and an initial reconstruction value for the data in an additional time instance after the initial time instance, after quantization, the difference value is determined based on one or more coding operation markings, and the quantized and compressed dataset is encoded based on the difference value.

[0149] In some examples, differential coding operations involve coding the amount of data in a compressed dataset based on its previous value relative to the amount of data in that dataset.

[0150] In some examples, entropy coding operations involve coding a compressed dataset using one or more symbols with lengths that vary based on the probability of the symbols appearing.

[0151] Figure 11 shows a diagram of a system 1100 including a device 1105 that supports coding techniques for a neural network architecture according to an aspect of this disclosure. Device 1105 may be, or include, an example of a component of device 805, device 905, or UE 115 as described herein. Device 1105 may communicate wirelessly with one or more base stations 105, UE 115, or any combination thereof. Device 1105 may include components for bidirectional voice and data communication, including components for transmitting and receiving communications, such as a communications manager 1120, an input / output (I / O) controller 1110, a transceiver 1115, an antenna 1125, a memory 1130, a code 1135, and a processor 1140. These components may communicate electronically via one or more buses (e.g., bus 1145) or may be coupled in other ways (e.g., operationally, communicatively, functionally, electronically, or electrically).

[0152] The I / O controller 1110 may manage input and output signals for device 1105. The I / O controller 1110 may also manage peripherals not integrated into device 1105. In some cases, the I / O controller 1110 may represent physical connections or ports to external peripherals. In some cases, the I / O controller 1110 may utilize an operating system such as iOS®, ANDROID®, MS-DOS®, MS-WINDOWS®, OS / 2®, UNIX®, LINUX®, or another known operating system. Additionally or alternatively, the I / O controller 1110 may represent, or interact with, a modem, keyboard, mouse, touchscreen, or similar device. In some cases, the I / O controller 1110 may be implemented as part of a processor, such as processor 1140. In some cases, the user may interact with device 1105 via I / O controller 1110 or via hardware components controlled by I / O controller 1110.

[0153] In some cases, device 1105 may include a single antenna 1125. However, in some other cases, device 1105 may have two or more antennas 1125, and these antennas may be capable of transmitting or receiving multiple wireless transmissions simultaneously. Transceiver 1115 may communicate bidirectionally via one or more antennas 1125, a wired link, or a wireless link, as described herein. For example, transceiver 1115 may represent a wireless transceiver and communicate bidirectionally with another wireless transceiver. Transceiver 1115 may also include a modem for modulating packets and providing the modulated packets to one or more antennas 1125 for transmission, and for demodulating packets received from one or more antennas 1125. Transceiver 1115, or transceiver 1115 and one or more antennas 1125, may be examples of transmitter 815, transmitter 915, receiver 810, receiver 910, or any combination thereof or components thereof, as described herein.

[0154] Memory 1130 may include random access memory (RAM) and read-only memory (ROM). Memory 1130 may store computer-readable, computer-executable code 1135, which, when executed by processor 1140, includes instructions causing device 1105 to perform various functions described herein. Code 1135 may be stored in a non-temporary computer-readable medium such as system memory or another type of memory. In some cases, code 1135 may not be directly executable by processor 1140, but may cause the computer to perform the functions described herein (for example, when compiled and executed). In some cases, memory 1130 may include a basic I / O system (BIOS) that can control basic hardware or software operations, such as, among other things, interaction with peripheral components or peripheral devices.

[0155] The processor 1140 may include intelligent hardware devices (e.g., general-purpose processors, DSPs, CPUs, microcontrollers, ASICs, FPGAs, programmable logic devices, discrete gate or transistor logic components, discrete hardware components, or any combination thereof). In some cases, the processor 1140 may be configured to operate a memory array using a memory controller. In some other cases, the memory controller may be integrated into the processor 1140. The processor 1140 may be configured to execute computer-readable instructions stored in memory (e.g., memory 1130) in order to cause device 1105 to perform various functions (e.g., functions or tasks supporting coding techniques for neural network architectures). For example, device 1105 or a component of device 1105 may include the processor 1140 and memory 1130 coupled to the processor 1140, and the processor 1140 and memory 1130 may be configured to perform various functions described herein.

[0156] The communication manager 1120 may support wireless communication in the UE in accordance with the examples disclosed herein. For example, the communication manager 1120 may be configured as a means for receiving, or otherwise supporting, an indication of one or more coding operations to be used to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The communication manager 1120 may be configured as a means for encoding a dataset to generate a compressed dataset by a neural network, or otherwise supporting, this. The communication manager 1120 may be configured as a means for quantizing a compressed dataset encoded by a neural network, or otherwise supporting, this. The communication manager 1120 may be configured as a means for encoding a quantized and compressed dataset based on having received an indication of one or more coding operations, or otherwise supporting, this. The communication manager 1120 may be configured as a means for transmitting an encoded, quantized, and compressed dataset to a second device after encoding the compressed dataset based on one or more coding operations, or otherwise supporting, this.

[0157] By including or configuring the communication manager 1120 in accordance with the examples described herein, device 1105 may support techniques for more efficient use of communication resources, improved inter-device coordination, and improved utilization of processing power. For example, by using differential encoder operations, entropy encoder operations, or both, the communication manager 1120 may further compress the dataset for transmission to a second device, thereby reducing signaling overhead to carry the compressed dataset and using fewer communication resources. In addition, by receiving parameter indications for different encoder operations from network devices, the communication manager 1120 may support coordination between device 1105 and network devices. By using differential encoder operations, entropy encoder operations, or both, it is also possible to use more of device 1105's processing power than fewer encoding operations that result in less compressed data than would have been obtained without differential and entropy encoder operations.

[0158] In some examples, the communications manager 1120 may be configured to perform various operations (e.g., receiving, monitoring, transmitting) using or otherwise cooperating with the transceiver 1115, one or more antennas 1125, or any combination thereof. Although the communications manager 1120 is shown as a separate component, in some examples, one or more functions described in relation to the communications manager 1120 may be supported or performed by the processor 1140, memory 1130, code 1135, or any combination thereof. For example, code 1135 may include instructions executable by the processor 1140 to cause the device 1105 to perform various aspects of coding techniques for neural network architectures such as those described herein, or the processor 1140 and memory 1130 may be otherwise configured to perform or support such operations.

[0159] Figure 12 shows a block diagram 1200 of a device 1205 supporting coding techniques for a neural network architecture according to an aspect of this disclosure. Device 1205 may be an example of an embodiment of a base station 105 as described herein. Device 1205 may include a receiver 1210, a transmitter 1215, and a communications manager 1220. Device 1205 may also include a processor. Each of these components may communicate with one another (for example, via one or more buses).

[0160] Receiver 1210 may provide means for receiving information such as packets, user data, control information, or any combination thereof related to various information channels (e.g., control channels, data channels, and information channels relating to coding techniques for neural network architectures). The information may be passed to other components of device 1205. Receiver 1210 may utilize a single antenna or a set of multiple antennas.

[0161] The transmitter 1215 may provide a means for transmitting signals generated by other components of device 1205. For example, the transmitter 1215 may transmit information such as packets, user data, control information, or any combination thereof related to various information channels (e.g., control channels, data channels, and information channels relating to coding techniques for neural network architectures). In some examples, the transmitter 1215 may be placed alongside the receiver 1210 in the transceiver module. The transmitter 1215 may utilize a single antenna or a set of multiple antennas.

[0162] The communication manager 1220, receiver 1210, transmitter 1215, or various combinations thereof or various components thereof may be examples of means for performing various embodiments of coding techniques for neural network architectures such as those described herein. For example, the communication manager 1220, receiver 1210, transmitter 1215, or various combinations thereof or components thereof may support methods for performing one or more of the functions described herein.

[0163] In some examples, the communications manager 1220, the receiver 1210, the transmitter 1215, or various combinations or components thereof may be implemented in hardware (for example, in a communications management circuit). The hardware may include a processor, a DSP, an ASIC, an FPGA or other programmable logic device, discrete gates or transistor logic, discrete hardware components, or any combination thereof configured as a means for performing, or otherwise supporting, the functions described herein. In some examples, a processor and memory coupled to the processor may be configured to perform one or more of the functions described herein (for example, by the processor executing instructions stored in memory).

[0164] As an addition or alternative, in some examples, the communications manager 1220, receiver 1210, transmitter 1215, or various combinations or components thereof may be implemented in code executed by a processor (for example, as communications management software or firmware). When implemented in code executed by a processor, the functions of the communications manager 1220, receiver 1210, transmitter 1215, or various combinations or components thereof may be performed by a general-purpose processor, DSP, CPU, ASIC, FPGA, or any combination thereof or other programmable logic device (for example, configured as a means for performing or otherwise supporting the functions described herein).

[0165] In some examples, the communications manager 1220 may be configured to perform various operations (e.g., receiving, monitoring, transmitting) using or otherwise cooperating with the receiver 1210, the transmitter 1215, or both. For example, the communications manager 1220 may receive information from the receiver 1210 and transmit information to the transmitter 1215, or be integrated with the receiver 1210, the transmitter 1215, or both to receive information, transmit information, or perform various other operations as described herein.

[0166] The communication manager 1220 may support wireless communication in the device in accordance with the examples disclosed herein. For example, the communication manager 1220 may be configured as a means for transmitting to the UE an indication of one or more coding operations to be used by the UE to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The communication manager 1220 may be configured as a means for receiving an encoded, quantized, and compressed dataset from the UE after the compressed dataset has been encoded following a quantization operation based on one or more coding operations, or it may be configured as a means for decoding an encoded, quantized, and compressed dataset to generate a compressed dataset based on at least part of one or more coding operations, or it may be configured as a means for decoding a compressed dataset to generate data based on the decoding of the encoded, quantized, and compressed dataset by a neural network, or it may be configured as a means for decoding a compressed dataset to generate data.

[0167] Figure 13 shows a block diagram 1300 of a device 1305 supporting coding techniques for a neural network architecture according to an aspect of this disclosure. Device 1305 may be an example of an aspect of device 1205 or base station 105 as described herein. Device 1305 may include a receiver 1310, a transmitter 1315, and a communications manager 1320. Device 1305 may also include a processor. Each of these components may communicate with one another (for example, via one or more buses).

[0168] Receiver 1310 may provide means for receiving information such as packets, user data, control information, or any combination thereof related to various information channels (e.g., control channels, data channels, and information channels relating to coding techniques for neural network architectures). The information may be passed to other components of device 1305. Receiver 1310 may utilize a single antenna or a set of multiple antennas.

[0169] Transmitter 1315 may provide means for transmitting signals generated by other components of device 1305. For example, transmitter 1315 may transmit information such as packets, user data, control information, or any combination thereof related to various information channels (e.g., control channels, data channels, and information channels relating to coding techniques for neural network architectures). In some examples, transmitter 1315 may be placed alongside receiver 1310 in a transceiver module. Transmitter 1315 may utilize a single antenna or a set of multiple antennas.

[0170] Device 1305 or its various components may be examples of means for performing various embodiments of coding techniques for neural network architectures described herein. For example, communication manager 1320 may include coding operation indicator component 1325, coding dataset receiving component 1330, decoder component 1335, neural network decoder component 1340, or any combination thereof. Communication manager 1320 may be an example of an embodiment of communication manager 1220 described herein. In some examples, communication manager 1320 or its various components may be configured to perform various operations (e.g., receiving, monitoring, transmitting) using or otherwise cooperating with receiver 1310, transmitter 1315, or both. For example, communication manager 1320 may receive information from receiver 1310 and transmit information to transmitter 1315, or be integrated with receiver 1310, transmitter 1315, or both to receive information, transmit information, or perform various other operations described herein.

[0171] The communication manager 1320 may support wireless communication in the device according to the examples disclosed herein. The coding operation indicator component 1325 may be configured as a means for transmitting an indicator of one or more coding operations that the UE should use to code a compressed dataset, or otherwise support it, wherein the one or more coding operations include differential coding operations, entropy coding operations, or both. The coded dataset receiving component 1330 may be configured as a means for receiving an coded, quantized, and compressed dataset from the UE after the compressed dataset has been coded following a quantization operation based on one or more coding operations, or otherwise support it. The decoder component 1335 may be configured as a means for decoding an coded, quantized, and compressed dataset to generate a compressed dataset based on one or more coding operations, or otherwise support it. The neural network decoder component 1340 may be configured as a means for decoding a compressed dataset to generate a dataset based on the decoding of the coded, quantized, and compressed dataset by a neural network, or otherwise support it.

[0172] Figure 14 shows a block diagram 1400 of a communications manager 1420 supporting coding techniques for a neural network architecture according to an aspect of this disclosure. The communications manager 1420 may be an example of communications manager 1220, communications manager 1320, or both, as described herein. The communications manager 1420 or various components thereof may be examples of means for performing various aspects of coding techniques for a neural network architecture as described herein. For example, the communications manager 1420 may include a coding operation indicator component 1425, a coding dataset receiver component 1430, a decoder component 1435, a neural network decoder component 1440, a difference decoder component 1445, an entropy decoder component 1450, or any combination thereof. Each of these components may communicate with one another directly or indirectly (for example, via one or more buses).

[0173] The communication manager 1420 may support wireless communication in the device according to the examples disclosed herein. The coding operation indicator component 1425 may be configured as a means for transmitting an indicator of one or more coding operations to be used by the UE to code a compressed dataset, or otherwise support it, wherein the one or more coding operations include differential coding operations, entropy coding operations, or both. The coded dataset receiving component 1430 may be configured as a means for receiving an coded, quantized, and compressed dataset from the UE after the compressed dataset has been coded following a quantization operation based on one or more coding operations, or otherwise support it. The decoder component 1435 may be configured as a means for decoding an coded, quantized, and compressed dataset based on one or more coding operations to produce a compressed dataset, or otherwise support it. The neural network decoder component 1440 may be configured as a means for decoding a compressed dataset to produce a dataset based on the decoding of the coded, quantized, and compressed dataset by a neural network, or otherwise support it.

[0174] In some examples, to support the transmission of one or more coding operation indicators, the coding operation indicator component 1425 may be configured as a means for transmitting one or more parameters corresponding to one or more coding operations, each of which one or more parameters corresponds to each coding operation of the one or more coding operations, and the quantized and compressed dataset is coded based on one or more parameters.

[0175] In some examples, to support decoding of encoded, quantized, and compressed datasets, the differential decoder component 1445 may be configured, or otherwise support, a means for decoding encoded, quantized, and compressed datasets using differential decoding operations.

[0176] In some examples, to support decoding of encoded, quantized, and compressed datasets, the entropy decoder component 1450 may be configured, or otherwise support, a means for decoding encoded, quantized, and compressed datasets using entropy decoding operations.

[0177] In some examples, to support the reception of encoded, quantized, and compressed datasets, the difference decoder component 1445 may be configured, or otherwise support, a means for receiving encoded, quantized, and compressed datasets that include difference values ​​to the data in a dataset based on the initial values ​​of the data.

[0178] In some examples, to support the reception of encoded, quantized, and compressed datasets, the difference decoder component 1445 may be configured, or otherwise support, for receiving encoded, quantized, and compressed datasets that include the difference values ​​to the data in the dataset based on previous reconstruction values ​​for the data.

[0179] In some examples, to support the reception of encoded, quantized, and compressed datasets, the difference decoder component 1445 may be configured, or otherwise support, for receiving encoded, quantized, and compressed datasets that include difference values ​​to the data in the dataset based on the initial reconstructed values ​​of the data.

[0180] In some examples, differential decoding operations involve encoding the amount of data in a compressed dataset based on its previous value relative to the amount of data in that dataset.

[0181] In some examples, entropy decoding operations involve encoding compressed datasets using one or more symbols with lengths that vary based on the probability of the symbols appearing.

[0182] Figure 15 shows a diagram of a system 1500 including a device 1505 that supports coding techniques for a neural network architecture according to an aspect of the present disclosure. Device 1505 may be, or include, an example of a component of device 1205, device 1305, or base station 105 as described herein. Device 1505 may communicate wirelessly with one or more base stations 105, UE 115, or any combination thereof. Device 1505 may include components for bidirectional voice and data communication, including components for transmitting and receiving communications, such as a communications manager 1520, a network communications manager 1510, a transceiver 1515, an antenna 1525, a memory 1530, a code 1535, a processor 1540, and an inter-station communications manager 1545. These components may communicate electronically over one or more buses (e.g., bus 1550) or may be coupled in other ways (e.g., operationally, communicatively, functionally, electronically, or electrically).

[0183] The network communication manager 1510 may manage communication with the core network 130 (for example, via one or more wired backhaul links). For example, the network communication manager 1510 may manage the transfer of data communications for one or more client devices such as UE 115.

[0184] In some cases, device 1505 may include a single antenna 1525. However, in some other cases, device 1505 may have two or more antennas 1525, and these antennas may be capable of transmitting or receiving multiple wireless transmissions simultaneously. Transceiver 1515 may communicate bidirectionally via one or more antennas 1525, a wired link, or a wireless link, as described herein. For example, transceiver 1515 may represent a wireless transceiver and communicate bidirectionally with another wireless transceiver. Transceiver 1515 may also include a modem for modulating packets and providing the modulated packets to one or more antennas 1525 for transmission, and for demodulating packets received from one or more antennas 1525. Transceiver 1515, or transceiver 1515 and one or more antennas 1525, may be examples of transmitters 1215, transmitters 1315, receivers 1210, receivers 1310, or any combination thereof or components thereof, as described herein.

[0185] Memory 1530 may include RAM and ROM. Memory 1530 may store computer-readable, computer-executable code 1535, which, when executed by processor 1540, includes instructions that cause device 1505 to perform various functions described herein. Code 1535 may be stored in a non-temporary computer-readable medium such as system memory or another type of memory. In some cases, code 1535 may not be directly executable by processor 1540, but (for example, when compiled and executed) may cause the computer to perform the functions described herein. In some cases, memory 1530 may include a BIOS that can control basic hardware or software operations, such as interaction with peripheral components or devices, among other things.

[0186] The processor 1540 may include intelligent hardware devices (e.g., general-purpose processors, DSPs, CPUs, microcontrollers, ASICs, FPGAs, programmable logic devices, discrete gate or transistor logic components, discrete hardware components, or any combination thereof). In some cases, the processor 1540 may be configured to operate a memory array using a memory controller. In some other cases, the memory controller may be integrated into the processor 1540. The processor 1540 may be configured to execute computer-readable instructions stored in memory (e.g., memory 1530) in order to cause device 1505 to perform various functions (e.g., functions or tasks supporting coding techniques for neural network architectures). For example, device 1505 or a component of device 1505 may include the processor 1540 and memory 1530 coupled to the processor 1540, and the processor 1540 and memory 1530 may be configured to perform various functions described herein.

[0187] The inter-station communication manager 1545 may manage communication with other base stations 105 and may include a controller or scheduler for coordinating with other base stations 105 to control communication with the UE 115. For example, the inter-station communication manager 1545 may coordinate scheduling for transmissions to the UE 115 for various interference mitigation techniques such as beamforming or joint transmission. In some examples, the inter-station communication manager 1545 may provide an X2 interface within the LTE / LTE-A wireless communication network technology for communication between base stations 105.

[0188] The communication manager 1520 may support wireless communication in a device in accordance with the examples disclosed herein. For example, the communication manager 1520 may be configured as a means for transmitting to a UE an indication of one or more coding operations to be used by the UE to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The communication manager 1520 may be configured as a means for receiving an encoded, quantized, and compressed dataset from the UE after the compressed dataset has been encoded following a quantization operation based on one or more coding operations, or it may be configured as a means for decoding an encoded, quantized, and compressed dataset to generate a compressed dataset based on at least part of one or more coding operations, or it may be configured as a means for decoding a compressed dataset to generate a dataset based on the decoding of the encoded, quantized, and compressed dataset by a neural network, or it may be configured as a means for decoding a compressed dataset to generate a dataset.

[0189] In some examples, the communications manager 1520 may be configured to perform various operations (e.g., receiving, monitoring, transmitting) using or otherwise cooperating with the transceiver 1515, one or more antennas 1525, or any combination thereof. Although the communications manager 1520 is shown as a separate component, in some examples, one or more functions described in relation to the communications manager 1520 may be supported or performed by the processor 1540, memory 1530, code 1535, or any combination thereof. For example, code 1535 may include instructions executable by the processor 1540 to cause the device 1505 to perform various aspects of coding techniques for neural network architectures such as those described herein, or the processor 1540 and memory 1530 may be otherwise configured to perform or support such operations.

[0190] Figure 16 shows a flowchart illustrating a method 1600 supporting coding techniques for a neural network architecture according to aspects of this disclosure. The operation of method 1600 may be performed by a UE or a component thereof, as described herein. For example, the operation of method 1600 may be performed by a UE 115, as described with reference to Figures 1 to 11. In some examples, the UE may execute a set of instructions to control a functional element of the UE to perform the described functions. Additionally or alternatively, the UE may perform aspects of the described functions using dedicated hardware.

[0191] In 1605, the method may include the step of receiving an indication of one or more coding operations to be used to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The operation of 1605 may be performed according to examples such as those disclosed herein. In some examples, the operation of 1605 may be performed by an encoder indication component 1025, as described with reference to Figure 10.

[0192] In 1610, the method may include the step of encoding a dataset by a neural network to generate a compressed dataset. The operation of 1610 may be performed according to examples such as those disclosed herein. In some examples, the operation of 1610 may be performed by a neural network encoder component 1030, as described with reference to Figure 10.

[0193] In 1615, the method may include the step of quantizing a compressed dataset encoded by a neural network. The operation of 1615 may be performed according to examples such as those disclosed herein. In some examples, the operation of 1615 may be performed by a quantization component 1035 as described with reference to Figure 10.

[0194] In 1620, the method may include the step of encoding a quantized and compressed dataset based on the reception of an indication of one or more encoding operations. The operation of 1620 may be performed according to examples such as those disclosed herein. In some examples, the operation of 1620 may be performed by an encoder component 1040, as described with reference to Figure 10.

[0195] In 1625, the method may include the step of encoding a compressed dataset based on one or more encoding operations, and then transmitting the encoded, quantized, and compressed dataset to a second device. The operation of 1625 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1625 may be performed by an encoded dataset transmission component 1045, as described with reference to Figure 10.

[0196] Figure 17 shows a flowchart illustrating a method 1700 supporting coding techniques for a neural network architecture according to aspects of this disclosure. The operation of method 1700 may be performed by a UE or a component thereof, as described herein. For example, the operation of method 1700 may be performed by a UE 115, as described with reference to Figures 1 to 11. In some examples, the UE may execute a set of instructions to control a functional element of the UE to perform the described functions. Additional or alternative, the UE may perform aspects of the described functions using dedicated hardware.

[0197] In 1705, the method may include the step of receiving an indication of one or more coding operations to be used to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The operation of 1705 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1705 may be performed by an encoder indication component 1025, as described with reference to Figure 10.

[0198] In 1710, the method may include the step of encoding a dataset by a neural network to generate a compressed dataset. The operation of 1710 may be performed according to examples such as those disclosed herein. In some examples, the operation of 1710 may be performed by a neural network encoder component 1030, as described with reference to Figure 10.

[0199] In 1715, the method may include the step of quantizing a compressed dataset encoded by a neural network. The operation of 1715 may be performed according to examples such as those disclosed herein. In some examples, the operation of 1715 may be performed by a quantization component 1035 as described with reference to Figure 10.

[0200] In 1720, the method may include the step of encoding a quantized and compressed dataset based on the reception of an indication of one or more encoding operations. The operation of 1720 may be performed according to examples such as those disclosed herein. In some examples, the operation of 1720 may be performed by an encoder component 1040, as described with reference to Figure 10.

[0201] In 1725, the method may include the step of encoding a quantized and compressed dataset using a differential coding operation. The operation of 1725 may be performed according to examples such as those disclosed herein. In some examples, the operation of 1725 may be performed by a differential encoder component 1050, as described with reference to Figure 10.

[0202] In 1730, the method may include the step of encoding a compressed dataset based on one or more encoding operations, and then transmitting the encoded, quantized, and compressed dataset to a second device. The operation of 1730 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1730 may be performed by an encoded dataset transmission component 1045, as described with reference to Figure 10.

[0203] Figure 18 shows a flowchart illustrating a method 1800 supporting coding techniques for a neural network architecture according to aspects of this disclosure. The operation of method 1800 may be performed by a UE or a component thereof, as described herein. For example, the operation of method 1800 may be performed by a UE 115, as described with reference to Figures 1 to 11. In some examples, the UE may execute a set of instructions to control a functional element of the UE to perform the described functions. In addition or alternatively, the UE may perform aspects of the described functions using dedicated hardware.

[0204] In 1805, the method may include the step of receiving an indication of one or more coding operations to be used to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The operation of 1805 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1805 may be performed by an encoder indication component 1025, as described with reference to Figure 10.

[0205] In 1810, the method may include the step of encoding a dataset by a neural network to generate a compressed dataset. The operation of 1810 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1810 may be performed by a neural network encoder component 1030, as described with reference to Figure 10.

[0206] In 1815, the method may include the step of quantizing a compressed dataset encoded by a neural network. The operation of 1815 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1815 may be performed by a quantization component 1035 as described with reference to Figure 10.

[0207] In 1820, the method may include the step of encoding a quantized and compressed dataset based on the reception of an indication of one or more encoding operations. The operation of 1820 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1820 may be performed by an encoder component 1040, as described with reference to Figure 10.

[0208] In 1825, the method may include the step of encoding a quantized and compressed dataset using an entropy coding operation. The operation of 1825 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1825 may be performed by an entropy encoder component 1055, as described with reference to Figure 10.

[0209] In 1830, the method may include the step of encoding a compressed dataset based on one or more encoding operations, and then transmitting the encoded, quantized, and compressed dataset to a second device. The operation of 1830 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1830 may be performed by an encoded dataset transmission component 1045, as described with reference to Figure 10.

[0210] Figure 19 shows a flowchart illustrating a method 1900 supporting coding techniques for a neural network architecture according to aspects of this disclosure. The operation of method 1900 may be performed by a base station or a component thereof, as described herein. For example, the operation of method 1900 may be performed by a base station 105, as described with reference to Figures 1 to 7 and Figures 12 to 15. In some examples, the base station may execute a set of instructions to control the functional elements of the base station to perform the functions described. In addition or alternatively, the base station may perform aspects of the functions described using dedicated hardware.

[0211] In 1905, the method may include sending to the UE an indication of one or more coding operations that the UE should use to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The operation of 1905 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1905 may be performed by a coding operation indication component 1425, as described with reference to Figure 14.

[0212] In 1910, the method may include the step of receiving an encoded, quantized, and compressed dataset from the UE after the compressed dataset has been encoded based on one or more encoding operations, following a quantization operation. The operation of 1910 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1910 may be performed by an encoded dataset receiving component 1430, as described with reference to Figure 14.

[0213] In 1915, the method may include the step of decoding an encoded, quantized, and compressed dataset to generate a compressed dataset based on one or more encoding operations. The operation of 1915 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1915 may be performed by a decoder component 1435, as described with reference to Figure 14.

[0214] In 1920, the method may include the step of decoding a compressed dataset in order to generate a dataset based on the decoding of an encoded, quantized, and compressed dataset by a neural network. The operation of 1920 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 1920 may be performed by a neural network decoder component 1440, as described with reference to Figure 14.

[0215] Figure 20 shows a flowchart illustrating a method 2000 that supports coding techniques for a neural network architecture according to aspects of this disclosure. The operation of method 2000 may be performed by a base station or a component thereof, as described herein. For example, the operation of method 2000 may be performed by base station 105, as described with reference to Figures 1 to 7 and Figures 12 to 15. In some examples, the base station may execute a set of instructions to control the functional elements of the base station to perform the functions described. In addition or alternatively, the base station may perform aspects of the functions described using dedicated hardware.

[0216] In 2005, the method may include the step of sending to the UE an indication of one or more coding operations that the UE should use to encode a compressed dataset, the one or more coding operations including differential coding operations, entropy coding operations, or both. The operation of 2005 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 2005 may be performed by a coding operation indication component 1425, as described with reference to Figure 14.

[0217] In 2010, the method may include the step of transmitting one or more parameters corresponding to one or more coding operations, each of which corresponds to each coding operation of the one or more coding operations, and the quantized and compressed dataset is coded based on the one or more parameters. The operation of 2010 can be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 2010 can be performed by a coding operation indicator component 1425, as described with reference to Figure 14.

[0218] In 2015, the method may include the step of receiving an encoded, quantized, and compressed dataset from the UE after the compressed dataset has been encoded based on one or more encoding operations, following a quantization operation. The operation of 2015 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 2015 may be performed by an encoded dataset receiving component 1430, as described with reference to Figure 14.

[0219] In 2020, the method may include the step of decoding an encoded, quantized, and compressed dataset to generate a compressed dataset based on one or more encoding operations. The operation of 2020 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 2020 may be performed by a decoder component 1435, as described with reference to Figure 14.

[0220] In 2025, the method may include the step of decoding a compressed dataset in order to generate a dataset based on the decoding of an encoded, quantized, and compressed dataset by a neural network. The operation of 2025 may be performed according to examples such as those disclosed herein. In some examples, aspects of the operation of 2025 may be performed by a neural network decoder component 1440, as described with reference to Figure 14.

[0221] The following provides an overview of the aspects of this disclosure.

[0222] Embodiment 1: A method for wireless communication in a UE, comprising: receiving an indication of one or more coding operations to be used to encode a compressed dataset, wherein the one or more coding operations comprise a differential coding operation or an entropy coding operation or both; encoding a dataset by a neural network to generate a compressed dataset; quantizing the compressed dataset encoded by the neural network; encoding the quantized and compressed dataset at least in part on having received an indication of one or more coding operations; and transmitting the encoded, quantized and compressed dataset to a second device after encoding the compressed dataset at least in part on one or more coding operations.

[0223] Embodiment 2: The method of Embodiment 1, wherein the step of receiving an indication of one or more coding operations comprises the step of receiving one or more parameters corresponding to one or more coding operations, each of the one or more parameters corresponding to each of the one or more coding operations, and the quantized and compressed dataset is coded based at least in part on the one or more parameters.

[0224] Embodiment 3: Any method of Embodiments 1 to 2, wherein the step of encoding a quantized and compressed dataset comprises encoding the dataset using a neural network and then encoding the quantized and compressed dataset using a differential encoding operation.

[0225] Embodiment 4: Any method of Embodiments 1 to 3, wherein the step of encoding a quantized and compressed dataset comprises encoding the dataset using a neural network and then encoding the quantized and compressed dataset using an entropy coding operation.

[0226] Embodiment 5: Any method of Embodiments 1 to 4, wherein the step of encoding a quantized and compressed dataset comprises determining, after quantizing the compressed dataset to be encoded by a neural network, a difference between a first value of data in the quantized and compressed dataset in an initial time instance and a second value of data in a second time instance after the initial time instance, wherein the difference value is determined at least in part on the marking of one or more encoding operations, and the quantized and compressed dataset is encoded at least in part on the difference value.

[0227] Embodiment 6: Any method of Embodiments 1 to 5, wherein the step of encoding a quantized and compressed dataset comprises determining, after quantizing the compressed dataset to be encoded by a neural network, a difference between a first reconstructed value of the data in the quantized and compressed dataset in a first time instance and a second reconstructed value of the data in a second time instance after the first time instance, wherein the difference value is determined based at least in part on the marking of one or more encoding operations, and the quantized and compressed dataset is encoded at least in part on the difference value.

[0228] Embodiment 7: Any method of Embodiments 1 to 6, wherein the step of encoding a quantized and compressed dataset comprises: determining an initial reconstruction value of the data in the quantized and compressed dataset in an initial time instance related to encoding the dataset after quantizing the compressed dataset to be encoded by a neural network; and determining, after quantization, the difference between an additional reconstruction value of the data in an additional time instance after the initial time instance and the initial reconstruction value of the data, wherein the difference value is determined at least in part on the marking of one or more encoding operations, and the quantized and compressed dataset is encoded at least in part on the difference value.

[0229] Aspect 8: A method according to any of aspects 1 to 7, wherein the differential encoding operation comprises encoding the amount of data based at least in part on previous values with respect to the amount of data in the compressed data set.

[0230] Aspect 9: A method according to any of aspects 1 to 8, wherein the entropy encoding operation comprises encoding the compressed data set using one or more symbols having lengths that vary at least in part based on the probability that the symbols appear.

[0231] Aspect 10: A method for wireless communication in a device, comprising: transmitting, to a UE, an indication of one or more encoding operations for the UE to use to encode a compressed data set, wherein the one or more encoding operations comprise a differential encoding operation or an entropy encoding operation or both; receiving, from the UE, the compressed data set after it has been encoded, quantized, and compressed, following a quantization operation based at least in part on the one or more encoding operations; decoding the encoded, quantized, and compressed data set to generate the compressed data set based at least in part on the one or more encoding operations; and decoding the compressed data set to generate a data set based at least in part on the neural network having decoded the encoded, quantized, and compressed data set.

[0232] Aspect 11: The method of aspect 10, wherein transmitting the indication of the one or more encoding operations comprises transmitting one or more parameters corresponding to the one or more encoding operations, each of the one or more parameters corresponding to a respective encoding operation of the one or more encoding operations, and the quantized and compressed data set is encoded based at least in part on the one or more parameters.

[0233] Aspect 12: The method according to any one of Aspects 10 to 11, wherein the step of decoding the encoded, quantized, and compressed data set comprises decoding the encoded, quantized, and compressed data set using a differential decoding operation before decoding the data set using a neural network.

[0234] Aspect 13: The method according to any one of Aspects 10 to 12, wherein the step of decoding the encoded, quantized, and compressed data set comprises decoding the encoded, quantized, and compressed data set using an entropy decoding operation before decoding the data set using a neural network.

[0235] Aspect 14: The method according to any one of Aspects 10 to 13, wherein the step of receiving the encoded, quantized, and compressed data set comprises receiving the encoded, quantized, and compressed data set with a differential value for the data based at least in part on an initial value of the data in the data set.

[0236] Aspect 15: The method according to any one of Aspects 10 to 14, wherein the step of receiving the encoded, quantized, and compressed data set comprises receiving the encoded, quantized, and compressed data set with a differential value for the data based at least in part on a previous reconstruction value of the data in the data set.

[0237] Aspect 16: The method according to any one of Aspects 10 to 15, wherein the step of receiving the encoded, quantized, and compressed data set comprises receiving the encoded, quantized, and compressed data set with a differential value for the data based at least in part on an initial reconstruction value of the data in the data set.

[0238] Aspect 17: The method according to any one of Aspects 10 to 16, wherein the differential decoding operation comprises encoding the amount of the data based at least in part on a previous value for the amount of the compressed data set.

[0239] Embodiment 18: Any method of Embodiments 10 to 17, comprising encoding a compressed dataset using one or more symbols having lengths that vary at least in part on the probability of the symbols appearing.

[0240] Embodiment 19: A device for wireless communication in a UE, comprising a processor, a memory coupled to the processor, and instructions stored in the memory that can be executed by the processor to cause the device to perform any of the methods of Embodiments 1 to 9.

[0241] Embodiment 20: Apparatus for wireless communication in a UE, comprising at least one means for performing any of the methods of Embodiments 1 to 9.

[0242] Embodiment 21: A non-temporary computer-readable medium for storing code for wireless communication in a UE, wherein the code comprises instructions that can be executed by a processor to perform any of Embodiments 1 to 9.

[0243] Embodiment 22: A device for wireless communication in a device, comprising a processor, a memory coupled to the processor, and instructions stored in the memory that can be executed by the processor to cause the device to perform any of the methods in Embodiments 10 to 18.

[0244] Embodiment 23: An apparatus for wireless communication in a device, comprising at least one means for performing any of the methods of Embodiments 10 to 18.

[0245] Embodiment 24: A non-temporary computer-readable medium for storing code for wireless communication in a device, wherein the code comprises instructions that can be executed by a processor to perform any of the methods of Embodiments 10 to 18.

[0246] It should be noted that the methods described herein describe possible implementations, and that the operation and steps may be reconfigured or otherwise modified, and that other implementations are possible. Furthermore, two or more embodiments of these methods may be combined.

[0247] While embodiments of LTE, LTE-A, LTE-A Pro, or NR systems may be described as examples, and the terms LTE, LTE-A, LTE-A Pro, or NR may be used extensively in the description, the techniques described herein are applicable to networks other than LTE, LTE-A, LTE-A Pro, or NR. For example, the techniques described may be applicable to various other wireless communication systems such as Ultra-Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20, Flash-OFDM, and other systems and wireless technologies not expressly mentioned herein.

[0248] The information and signals described herein may be represented using any of a wide variety of techniques and methods. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be mentioned throughout this description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, light fields or optical particles, or any combination thereof.

[0249] The various exemplary blocks and components described in this disclosure may be implemented or run using general-purpose processors, DSPs, ASICs, CPUs, FPGAs or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general-purpose processor may be a microprocessor, but alternatively, a processor may be any processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices (e.g., a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors working with a DSP core, or any other such configuration).

[0250] The functions described herein may be implemented in hardware, software executed by a processor, firmware, or any combination thereof. When implemented in software executed by a processor, the functions may be stored in or transmitted via a computer-readable medium as one or more instructions or codes. Other examples and implementations are within the scope of this disclosure and the accompanying claims. For example, due to the nature of the software, the functions described herein may be implemented using software executed by a processor, hardware, firmware, hardwiring, or any combination thereof. Features implementing the functions may also be physically arranged in various locations, including being distributed so that parts of the functions are implemented in different physical locations.

[0251] Computer-readable media include both non-temporary computer storage media and communication media, including any media that enables the transfer of computer programs from one location to another. Non-temporary storage media can be any available media that can be accessed by a general-purpose or dedicated computer. Examples, rather than limitations, of non-temporary computer-readable media may include RAM, ROM, electrically erasable programmable ROM (EEPROM), flash memory, compact disk (CD) ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other non-temporary media used to carry or store desired program code means in the form of instructions or data structures, and that can be accessed by a general-purpose or dedicated computer or general-purpose or dedicated processor. Any connection is also appropriately referred to as computer-readable media. For example, if software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of computer-readable media. As used herein, the terms "disk" and "disc" include CDs, laserdiscs, optical discs, digital multipurpose discs (DVDs), floppy disks, and Blu-ray® discs, where a disk typically reproduces data magnetically and a disc reproduces data optically using a laser. Combinations of these terms are also included within the scope of computer-readable media.

[0252] When used herein, including within the claims, “or” as used in an enumeration of items (for example, an enumeration of items beginning with a phrase such as “at least one of” or “one or more of”) indicates an inclusive enumeration, such as the enumeration “at least one of A, B, or C” meaning A or B or C or AB or AC or BC or ABC (i.e., A and B and C). Furthermore, the phrase “based on” as used herein should not be construed as a reference to a closed set of conditions. For example, an exemplary step described as “based on condition A” may be based on both condition A and condition B without departing from the scope of this disclosure. In other words, the phrase “based on” as used herein should be construed in the same way as the phrase “at least partially based on.”

[0253] The term “decide” or “to decide” encompasses a wide range of actions, and therefore “deciding” can include calculating, calculating, processing, deriving, investigating, searching (for example, searching in a table, database, or other data structure), confirming, etc. It can also include receiving (such as receiving information), accessing (such as accessing data in memory), etc. Furthermore, it can include resolving, selecting, choosing, establishing, or other similar actions.

[0254] In the attached diagrams, similar components or features may have the same reference label. Furthermore, various components of the same type may be distinguished by following the reference label with a dash and a second label that distinguishes similar components. When only the first reference label is used herein, the description is applicable to any similar component having the same first reference label, regardless of the second reference label or any other subsequent reference labels.

[0255] The description set forth in this specification with respect to the accompanying drawings describes exemplary configurations and does not necessarily represent all examples that can be implemented or fall within the scope of the claims. The term "exemplary" as used herein means "serving as an example, instance, or illustration" and does not mean "preferred" or "advantageous over other examples." The detailed description includes specific details for the purpose of providing an understanding of the described techniques. However, these techniques can be practiced without these specific details. In some instances, well-known structures and devices are shown in block diagram form to avoid obscuring the concepts of the examples being described.

[0256] The description in this specification is provided to enable a person skilled in the art to make or use the present disclosure. Various modifications to the present disclosure will be apparent to those skilled in the art, and the general principles defined herein can be applied to other variations without departing from the scope of the present disclosure. Therefore, the present disclosure is not limited to the examples and designs described in this specification, and the broadest scope consistent with the principles and novel features disclosed herein should be given.

Description of Reference Numerals

[0257] 105 Base Station 110 Coverage Area 115 UE 120 Backhaul Link 125 Communication Link 130 Core Network 135 D2D Communication Link 140 Access Network Entity 145 Access Network Transmission Entity 150 IP Service 200 Wireless Communication System 205 Carrier 210 Carrier 215 Encoding Parameter 220 Differential Encoder Parameter 225 Entropy Encoder Parameter 230 datasets to be compressed and encoded 235 Compressed and Encoded Datasets 305 Input 310 Spatial Feature Extraction 315 Tap Region Feature Extraction 320 Feature Compression 325 Quantization 330 Feature Extraction 335 Tap Region Feature Reconstruction 340 Spatial Feature Reconstruction 345 Output 350 Encoder Neural Network 355 Quantizer 360 Differential Encoder 365 Entropy Encoder 370 Feedback 375 Entropy Decoder 380 Differential Decoder 385 Decoder Neural Network 405 Input 410 Convolutional Layers 415 ResNet blocks 420 ResNet blocks 425 Convolutional Layer 430 Flattening operation 435 Fully connected layer 440 Quantization 445 Entropy Encoder 450 Entropy Decoder 455 Fully connected layer 460 Reshaping operation 465 Convolutional Layers 470 ResNet blocks 475 ResNet blocks 480 Convolutional Layers 485 output 505 Input 510 Differential Encoder 515 Entropy Encoder 520 Entropy Decoder 525 Differential Decoder 530 output 605 Input 610 Differential Encoder 615 Normalization operation 620 Entropy Encoder 625 Entropy Decoder 630 Normalization Cancellation Operation 635 Differential Decoder 640 output 805 devices 810 Receiver 815 Transmitter 820 Communications Manager 905 Device 910 Receiver 915 Transmitter 920 Communications Manager 925 Encoder Indication Component 930 Neural Network Encoder Component 935 Quantization Components 940 Encoder Component 945 Encoded Dataset Transmission Component 1020 Communications Manager 1025 Encoder Indication Component 1030 Neural Network Encoder Component 1035 Quantization Component 1040 Encoder Component 1045 Encoded Dataset Transmission Component 1050 Differential Encoder Component 1055 Entropy Encoder Component 1105 devices 1110 I / O Controller 1115 Transceiver 1120 Communications Manager 1125 Antenna 1130 memory 1135 Code 1140 Processor 1145 Bus 1205 devices 1210 Receiver 1215 Transmitter 1220 Communications Manager 1305 devices 1310 Receiver 1315 Transmitter 1320 Communications Manager 1325 Encoding operation indicator component 1330 Encoded Dataset Receiver Component 1335 Decoder Component 1340 Neural Network Decoder Component 1420 Communications Manager 1425 Encoding operation indicator component 1430 Encoded Dataset Receiver Component 1435 Decoder Component 1440 Neural Network Decoder Component 1445 Differential Decoder Component 1450 Entropy Decoder Component 1505 Device 1510 Network Communications Manager 1515 Transceiver 1520 Communications Manager 1525 Antenna 1530 memory 1535 Code 1540 processor 1545 Inter-station communications manager 1550 Bus

Claims

1. A method for wireless communication in user equipment (UE), A step of receiving an indication from a second device via a communication link through a carrier of one or more encoding operations to be used to encode a compressed dataset, wherein the one or more encoding operations include a differential encoding operation. The steps include: encoding the dataset using a neural network to generate the compressed dataset; The steps include: quantizing the compressed dataset encoded by the neural network; The steps of encoding the compressed and quantized dataset using the differential encoding operation, at least in part, based on having received the indication of one or more encoding operations, The steps include: encoding the compressed dataset based at least in part on one or more encoding operations, and then transmitting the compressed, quantized, and encoded dataset to the second device; A method that includes [a certain feature].

2. The step of receiving the indication of one or more coding operations is: The method according to claim 1, comprising the step of receiving one or more parameters corresponding to one or more encoding operations, each of the one or more parameters corresponding to each of the one or more encoding operations, and the compressed and quantized dataset is encoded based at least in part on the one or more parameters.

3. The step of encoding the compressed and quantized dataset is: The method according to claim 1, comprising the step of determining a difference between a first value of data in the compressed and quantized dataset in an initial time instance and a second value of the data in a second time instance after the initial time instance, after quantizing the compressed dataset encoded by the neural network, wherein the difference value is determined based at least in part on the markings of one or more encoding operations, and the compressed and quantized dataset is encoded based at least in part on the difference value.

4. The step of encoding the compressed and quantized dataset is: The method according to claim 1, comprising the step of determining a difference between a first reconstruction value of the data in the compressed and quantized dataset in a first time instance and a second reconstruction value of the data in a second time instance after the first time instance, after quantizing the compressed dataset encoded by the neural network, wherein the difference value is determined based at least in part on the markings of the one or more encoding operations, and the compressed and quantized dataset is encoded based at least in part on the difference value.

5. The step of encoding the compressed and quantized dataset is: The steps include: quantizing the compressed dataset encoded by the neural network, and then determining initial reconstruction values ​​for the data in the compressed and quantized dataset in the initial time instance related to encoding the dataset; The method according to claim 1, comprising the step of determining, after the quantization, a difference between an additional reconstructed value of the data in an additional time instance after the initial time instance and the initial reconstructed value of the data, wherein the difference value is determined based at least in part on the marking of the one or more encoding operations, and the compressed and quantized dataset is encoded based at least in part on the difference value.

6. The method according to claim 1, wherein the differential coding operation comprises coding the amount of data based at least in part on a previous value for the amount of data in the compressed dataset.

7. A method for wireless communication in a device, A step of transmitting to a user device (UE) via a carrier communication link an indication of one or more encoding operations to be used by the UE to encode a compressed dataset, wherein the one or more encoding operations include a differential encoding operation. The steps include receiving the compressed, quantized, and encoded dataset from the UE after the compressed dataset has been encoded following a quantization operation, based at least in part on the one or more encoding operations, The steps include: decoding the compressed, quantized, and encoded dataset using a differential decoding operation in order to generate the compressed dataset based on at least part of the one or more encoding operations; A method comprising the steps of: decoding a compressed dataset in order to generate a dataset based at least in part on decoding the compressed, quantized, and encoded dataset by a neural network.

8. The step of transmitting the indication of one or more coding operations is: The method according to claim 7, comprising the step of transmitting one or more parameters corresponding to one or more encoding operations, each of the one or more parameters corresponding to each of the one or more encoding operations, and the compressed and quantized dataset is encoded based at least in part on the one or more parameters.

9. The step of decoding the compressed, quantized, and encoded dataset is: The method according to claim 7, further comprising the step of decoding the compressed, quantized, and encoded dataset using an entropy decoding operation before decoding the dataset using the neural network.

10. The step of receiving the compressed, quantized, and encoded dataset is: The method according to claim 7, comprising the step of receiving the compressed, quantized, and encoded dataset, which comprises difference values ​​to the data based at least in part on initial values ​​of the data in the dataset.

11. The step of receiving the compressed, quantized, and encoded dataset is: The method according to claim 7, comprising the step of receiving the compressed, quantized, and encoded dataset, which comprises differential values ​​to the data based at least in part on previous reconstruction values ​​of the data in the dataset.

12. The step of receiving the compressed, quantized, and encoded dataset is: The method according to claim 7, comprising the step of receiving the compressed, quantized, and encoded dataset, which comprises differential values ​​to the data based at least in part on initial reconstruction values ​​of the data in the dataset.

13. A device for wireless communication in user equipment (UE), One or more processors, One or more memories coupled to the one or more processors, The device comprises instructions stored in one or more of the aforementioned memories, and the instructions are transmitted to the device. Receiving, via a communication link through a carrier, an indication from a second device of one or more coding operations to be used to encode a compressed dataset, wherein the one or more coding operations include a differential coding operation, The neural network encodes the dataset in order to generate the compressed dataset, Quantizing the compressed dataset encoded by the neural network, Encoding the compressed and quantized dataset using the differential coding operation, based at least in part on the reception of the indication of one or more coding operations, After the compressed dataset is encoded based on at least part of the one or more encoding operations, the compressed, quantized, and encoded dataset is transmitted to the second device. A device that can be executed by one or more processors to perform the following actions.

14. A device for wireless communication in a device, One or more processors, One or more memories coupled to the one or more processors, The device comprises instructions stored in one or more of the aforementioned memories, and the instructions are transmitted to the device. Transmitting to a user device (UE) via a carrier-mediated communication link an indication of one or more encoding operations to be used by the UE to encode a compressed dataset, wherein the one or more encoding operations include a differential encoding operation. From the aforementioned UE, after the compressed dataset has been encoded following a quantization operation based at least in part on the one or more encoding operations, the compressed, quantized, and encoded dataset is received. Decoding the compressed, quantized, and encoded dataset using a differential decoding operation in order to generate the compressed dataset based on at least part of the one or more encoding operations, The neural network decodes the compressed dataset in order to generate a dataset based on at least a portion of the compressed, quantized, and encoded dataset to be decoded. A device that can be executed by one or more processors to perform the following actions.

15. A computer program that includes instructions, When the command is executed by a device for wireless communication in a user device (UE), the command causes the device to perform the method according to any one of claims 1 to 6. A computer program that, when the instruction is executed by a device for wireless communication in a device, causes the device to perform the method according to any one of claims 7 to 12.

Images