Method and device for training artificial intelligence model in wireless communication system

Abstract

The present disclosure relates to a 5G or 6G communication system for supporting higher data transmission rates beyond 4G communication systems such as LTE. A base station according to an embodiment may operate in the order of: transmitting, to a server, capability information of the base station for replacing at least one block included in a receiving end with a first artificial intelligence model; transmitting, to the server, block indication information indicating at least one block to be replaced with the first artificial intelligence model; receiving, from the server, inference values acquired from a second artificial intelligence model determined by the server on the basis of the capability information and the block indication information; and training the first artificial intelligence model on the basis of the inference values.

Description

Method and device for training an artificial intelligence model in a wireless communication system

[0001] The present disclosure relates to the operation of a base station and a server in a wireless communication system. Specifically, the present disclosure relates to an apparatus and method for training an artificial intelligence model included in a base station based on the inference value of an artificial intelligence model included in a server.

[0002] Looking back at the evolution of wireless communication through successive generations, technologies have been developed primarily for human-oriented services, such as voice, multimedia, and data. Following the commercialization of 5G (5th-generation) communication systems, connected devices, which have been increasing explosively, are expected to be connected to communication networks. Examples of networked objects include vehicles, robots, drones, home appliances, displays, smart sensors installed in various infrastructures, construction machinery, and factory equipment. Mobile devices are expected to evolve into various form factors, such as augmented reality glasses, virtual reality headsets, and holographic devices. In the 6G (6th-generation) era, efforts are underway to develop improved 6G communication systems to connect hundreds of billions of devices and objects to provide diverse services. For this reason, 6G communication systems are being referred to as "beyond 5G" systems.

[0003] In the 6G communication system predicted to be realized around 2030, the maximum transmission speed is tera (i.e., 1,000 gigabit) bps, and the wireless latency is 100 microseconds (μsec). In other words, compared to the 5G communication system, the transmission speed in the 6G communication system is 50 times faster, and the wireless latency is reduced to one-tenth.

[0004] To achieve such high data transmission speeds and ultra-low latency, 6G communication systems are being considered for implementation in the terahertz band (e.g., the 95 GHz to 3 terahertz (3 THz) band). In the terahertz band, due to more severe path loss and atmospheric absorption compared to the millimeter wave (mmWave) band introduced in 5G, the importance of technology capable of guaranteeing signal reach, or coverage, is expected to increase. As key technologies to ensure coverage, radio frequency (RF) devices, antennas, new waveforms that offer better coverage than orthogonal frequency division multiplexing (OFDM), beamforming, and multi-antenna transmission technologies such as massive multiple-input and multiple-output (massive MIMO), full-dimensional MIMO (FD-MIMO), array antennas, and large-scale antennas must be developed. In addition, new technologies such as metamaterial-based lenses and antennas, high-dimensional spatial multiplexing technology using orbital angular momentum (OAM), and reconfigurable intelligent surface (RIS) are being discussed to improve coverage of terahertz band signals.

[0005] In addition, to improve frequency efficiency and system network, development is underway in 6G communication systems for full duplex technology, in which uplink and downlink simultaneously utilize the same frequency resources at the same time; network technology that integrates satellites and HAPS (high-altitude platform stations); network structure innovation technology that supports mobile base stations and enables network operation optimization and automation; dynamic spectrum sharing technology through collision avoidance based on spectrum usage prediction; AI-based communication technology that utilizes AI (artificial intelligence) from the design stage and internalizes end-to-end AI support functions to realize system optimization; and next-generation distributed computing technology that realizes services of complexity exceeding the limits of terminal computing capabilities by utilizing ultra-high performance communication and computing resources (mobile edge computing (MEC), cloud, etc.). In addition, attempts are continuing to further strengthen connectivity between devices, further optimize networks, promote the softwareization of network entities, and increase the openness of wireless communication through the design of new protocols to be used in 6G communication systems, the implementation of hardware-based security environments, the development of mechanisms for the safe utilization of data, and the development of technologies regarding privacy maintenance methods.

[0006] Due to the research and development of such 6G communication systems, it is expected that a new dimension of hyper-connected experience will become possible through the hyper-connectivity of 6G communication systems, which encompasses not only connections between objects but also connections between people and objects. Specifically, it is projected that 6G communication systems will enable the provision of services such as truly immersive extended reality (truly immersive XR), high-fidelity mobile holograms, and digital replicas. Furthermore, services such as remote surgery, industrial automation, and emergency response, which are provided through 6G communication systems with enhanced security and reliability, will be applied in various fields including industry, healthcare, automotive, and home appliances.

[0007] A method of operation of a base station according to one embodiment may include the step of transmitting capability information of the base station to a server for replacing at least one block included in a receiving end with a first artificial intelligence model.

[0008] A method of operation of a base station according to one embodiment may include the step of transmitting block instruction information to the server that indicates the at least one block that is replaced by the first artificial intelligence model.

[0009] A method of operation of a base station according to one embodiment may include the step of receiving from the server an inference value obtained from a second artificial intelligence model determined by the server based on the capability information and the block instruction information.

[0010] A method of operating a base station according to one embodiment may include the step of training the first artificial intelligence model based on the inference value.

[0011] A method of operation of a server according to one embodiment may include the step of receiving capability information of the base station from the base station for replacing at least one block included in the receiving end of the base station with a first artificial intelligence model.

[0012] A method of operation of a server according to one embodiment may include the step of receiving block instruction information from the base station that indicates the at least one block to be replaced by the first artificial intelligence model.

[0013] A method of operation of a server according to one embodiment may include the step of determining a second artificial intelligence model for training the first artificial intelligence model based on the capability information and the block instruction information.

[0014] A method of operation of a server according to one embodiment may include the step of receiving input data from the base station for obtaining an inference value from the second artificial intelligence model.

[0015] A method of operation of a server according to one embodiment may include the step of obtaining an inference value of the second artificial intelligence model by inputting the input data into the second artificial intelligence model.

[0016] A method of operation of a server according to one embodiment may include the step of transmitting the inference value to the base station.

[0017] A base station according to one embodiment may include a transceiver and at least one processor connected to the transceiver.

[0018] A processor according to one embodiment can transmit capability information of the base station to a server for replacing at least one block included in the receiving end with a first artificial intelligence model.

[0019] A processor according to one embodiment can transmit block instruction information to the server that indicates the at least one block to be replaced by the first artificial intelligence model.

[0020] A processor according to one embodiment can receive from the server an inference value obtained from a second artificial intelligence model determined by the server based on the capability information and the block instruction information.

[0021] A processor according to one embodiment can train the first artificial intelligence model based on the inference value.

[0022] A server according to one embodiment may include a transceiver and at least one processor connected to the transceiver.

[0023] A processor according to one embodiment can receive capability information of the base station from the base station for replacing at least one block included in the receiving end of the base station with a first artificial intelligence model.

[0024] A processor according to one embodiment can receive block instruction information from the base station that indicates the at least one block that is replaced by the first artificial intelligence model.

[0025] A processor according to one embodiment can determine a second artificial intelligence model for training the first artificial intelligence model based on the capability information and the block instruction information.

[0026] A processor according to one embodiment can receive input data from the base station for obtaining an inference value from the second artificial intelligence model.

[0027] A processor according to one embodiment can obtain an inference value of the second artificial intelligence model by inputting the input data into the second artificial intelligence model.

[0028] A processor according to one embodiment can transmit the inference value to the base station.

[0029] FIG. 1 is a flowchart illustrating a method of operation of a base station according to one embodiment.

[0030] FIG. 2 is a diagram illustrating an example of a receiving unit included in a base station according to one embodiment.

[0031] FIG. 3 is a diagram illustrating a first artificial intelligence model learned at a base station according to one embodiment.

[0032] FIG. 4 is a diagram illustrating capability information that a base station according to one embodiment transmits to a server according to one embodiment.

[0033] FIG. 5 is a flowchart illustrating the operation method of a base station according to one embodiment.

[0034] FIG. 6 is a flowchart illustrating the operation method of a base station according to one embodiment.

[0035] FIG. 7 is a flowchart illustrating the operation method of a base station according to one embodiment.

[0036] FIG. 8 is a flowchart illustrating the operation method of a base station according to one embodiment.

[0037] FIG. 9 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0038] FIG. 10 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0039] FIG. 11 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0040] FIG. 12 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0041] FIG. 13 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0042] FIG. 14 is a block diagram illustrating the components of a base station according to one embodiment.

[0043] FIG. 15 is a block diagram illustrating the components of a server according to one embodiment.

[0044] The terms used in this specification will be briefly explained, and the invention will be described in detail.

[0045] The terms used in this invention have been selected based on currently widely used general terms, taking into account their functions within the invention; however, these terms may vary depending on the intent of those skilled in the art, case law, the emergence of new technologies, etc. Additionally, in specific cases, terms have been arbitrarily selected by the applicant, and in such cases, their meanings will be described in detail in the relevant description of the invention. Therefore, the terms used in this invention should be defined not merely by their names, but based on their meanings and the overall content of the invention.

[0046] In the present disclosure, it will be understood that each block of the process flow diagrams and combinations of the flow diagrams may be performed based on computer program instructions. Since these computer program instructions may be optionally loaded into at least one processor of a general-purpose computer, a computer for special purposes, or other programmable data processing equipment, the instructions performed through any one or any combination of at least one processor of the computer or other programmable data processing equipment create means for performing the functions described in the flow diagram block(s). Since these computer program instructions may also be stored in computer-available or computer-readable memory that can be directed toward the computer or other programmable data processing equipment to implement the functions in a specific manner, the instructions stored in computer-available or computer-readable memory may also produce a manufactured item containing means of instruction for performing the functions described in the flow diagram block(s). Since computer program instructions can be loaded onto a computer or other programmable data processing equipment, instructions that perform a series of operation steps on the computer or other programmable data processing equipment to create a process executed by the computer can also provide steps for executing the functions described in the flowchart block(s).

[0047] Additionally, each block may represent a module, segment, or part of code containing one or more executable instructions for executing a specified logical function(s). It should also be noted that in some alternative execution examples, the functions mentioned in the blocks may occur out of order. For example, two blocks (or functions) described in succession may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order according to the corresponding function.

[0048] As stated above, it should be noted that the blocks of each flowchart and combinations of flowcharts described in this disclosure may be executed by one or more computer programs including instructions. The entirety of one or more computer programs may be stored in a single memory device, or one or more computer programs may be divided into different parts and stored across multiple memory devices.

[0049] Additionally, any / any function or operation described in this disclosure may be processed by a single processor or a combination of processors. The single processor or combination of processors is a circuitry that performs processing and may include an application processor (AP, e.g., a central processing unit (CPU)), a communication processor (CP, e.g., a modem), a graphics processing unit (GPU), a neural network processing unit (NPU) (e.g., an artificial intelligence (AI) chip), a Wi-Fi chip, a Bluetooth® chip, a global positioning system (GPS) chip, a near-field communication (NFC) chip, a connectivity chip, a sensor controller, a touch controller, a fingerprint sensor controller, a display driver integrated circuit (IC), an audio codec (CODEC) chip, a universal serial bus (USB) controller, a camera controller, an image processing IC, a microprocessor unit (MPU), a system-on-chip (SoC), an IC, or similar circuitry.

[0050] Additionally, it should be noted that various embodiments in the claims and description of the present disclosure may be implemented in the form of hardware, software, or a combination of hardware and software.

[0051] When a part of a specification is described as "including" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "...part" or "module" as used in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware or software, or as a combination of hardware and software.

[0052] Additionally, the description 'at least one of A, B, and C' means that it may be any one of 'A', 'B', 'C', 'A and B', 'A and C', 'B and C', and 'A, B, and C'.

[0053] It should be understood that the combinations of blocks and flowcharts in each flowchart can be executed by one or more computer programs containing computer-executable instructions. One or more computer programs may be stored entirely in a single memory or may be partitioned and stored in multiple different memories.

[0054] All functions or operations described in this document may be processed by a single processor or a combination of processors. A single processor or a combination of processors is a circuitry that performs processing and may include circuitry such as an AP (Application Processor), CP (Communication Processor), GPU (Graphical Processing Unit), NPU (Neural Processing Unit), MPU (Microprocessor Unit), SoC (System on Chip), IC (Integrated Chip), etc.

[0055] A processor may include various processing circuits and / or multiple processors. For example, the term “processor” as used herein, including in the claims, may include at least one processor and various processing circuits. In the at least one processor, one or more processors may be configured to perform the various functions described herein individually and / or collectively in a distributed manner. As used herein, “processor,” “at least one processor,” and “one or more processors” may be configured to perform various functions. However, these terms cover, without limitation, situations where one processor performs some of the functions and other processor(s) perform other parts of the functions, and situations where a single processor can perform all functions. Additionally, at least one processor may include a combination of processors performing various functions of the disclosed functions in a distributed manner. At least one processor may execute program instructions to achieve or perform various functions.

[0056] Furthermore, the expressions 'if' and 'in case that' described in this disclosure or claims may be interpreted, depending on the context, as meaning 'when or upon,' 'in response to,' 'based on,' or 'according to,' and these expressions may be used interchangeably. In addition, other expressions having substantially the same meaning may be used as substitutes, provided that they do not impair the technical features of this disclosure.

[0057] Additionally, the term "not perform" as used in this disclosure or claims may be understood, depending on the context, to mean to omit or skip the corresponding step. Such a term may be replaced with other terms having the same or substantially similar meaning.

[0058] Additionally, the phrase "transmitting a message containing A and B" as described in this specification may be interpreted to include not only (i) cases where A and B are transmitted as a single message, but also (ii) cases where A and B are transmitted individually through multiple messages (e.g., transmitting a first message containing A and a second message containing B). This interpretation may also apply to cases where messages containing two or more items, such as A, B, and C, are transmitted together or individually.

[0059] In addition, 'transmitting a message containing A and transmitting a message containing B' can also be interpreted as transmitting a single message containing A and B.

[0060] The drawings or flowcharts described below illustrate exemplary methods that may be implemented in accordance with the principles of the present disclosure, and various modifications may be made to the methods illustrated in the flowcharts of the present disclosure. For example, although illustrated as a series of steps, the various steps of each drawing or flowchart may overlap, occur in parallel, occur in a different order, or occur multiple times. In other examples, any step may be omitted or replaced with another step.

[0061] The methods and devices proposed in the embodiments of the present disclosure below are not limited to each embodiment and may be utilized as a combination of all or part of the embodiments proposed in the disclosure. Accordingly, the embodiments of the present disclosure may be applied with some modifications within the scope that does not deviate significantly from the scope of the present disclosure, at the judgment of a person skilled in the art.

[0062] In this case, any wording mentioned in different embodiments may be used interchangeably, combined, or substituted if the concepts correspond. For example, regarding the same or corresponding concepts, even if the expression 'A' is used in one embodiment and the expression 'B' is used in another embodiment, they may be understood by interchangeably, substituted, or combined.

[0063] Hereinafter, the expression in the present disclosure or claims that information can be configured from a base station may mean that, depending on the context, a terminal receives said information from a base station through physical layer signaling or upper layer signaling, and such expression may be replaced with other terms having the same or substantially similar meaning.

[0064] The following describes embodiments with reference to the attached drawings so that those skilled in the art can easily implement the present invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0065] FIG. 1 is a flowchart illustrating a method of operation of a base station according to one embodiment.

[0066] In step 110, the base station may transmit information on the base station's capability to replace at least one block included in the receiver with the first artificial intelligence model to the server.

[0067] A receiving unit included in a base station according to one embodiment will be described with reference to FIG. 2.

[0068] FIG. 2 is a diagram illustrating an example of a receiving unit included in a base station according to one embodiment.

[0069] A base station according to one embodiment may include a receiving unit (200).

[0070] The receiving unit (200) may include at least one block (210, 220, ..., 280, 290) for converting the input (205) into an output (295).

[0071] At least one block (210, 220, ..., 280, 290) included in the receiving unit (200) is merely an example, and some blocks may be omitted or added depending on the operation of the base station according to one embodiment.

[0072] The input (200) may mean an analog signal received from an external device.

[0073] The external device may refer to a terminal that communicates with a base station according to one embodiment and / or a base station that communicates with a base station according to one embodiment, but is not limited thereto.

[0074] The output (295) may represent a digital signal obtained by inputting the input (200) to the receiving unit (200).

[0075] The OFDM (orthogonal frequency division multiplexing) demodulation block (210) can convert the OFDM signal into a baseband signal, detect the offset of the symbol timing to set the synchronization of the symbol, and perform a Fourier transform.

[0076] The STO (symbol time offset) & CFO (carrier frequency offset) correction block (220) can compensate for the symbol timing offset and carrier frequency offset of the received signal to set the synchronization of the symbol and frequency.

[0077] The Layer & RE (resource element) demapping block (230) can convert the modulated data into the original bit sequence and restore the interleaved data to the original order.

[0078] The channel estimator block (240) can estimate the characteristics of the channel through the received signal. The characteristics of the channel may include, but are not limited to, a demodulation reference signal (DMRS), a sounding reference signal (SRS), and / or a positioning reference signal (PRS).

[0079] The channel equalizer block (250) can restore the original data by correcting the influence of the channel using the channel estimation result.

[0080] The soft demodulator block (260) can convert the demodulated signal into soft bits to obtain probability information necessary for error correction.

[0081] The descrambling block (270) can descramble the scrambled data to restore the data to its original order.

[0082] The rate dematching block (280) can perform dematching to match the redundant data added through channel coding to the original data rate.

[0083] The Channel decoding block (290) can correct errors that occur in the received data.

[0084] Referring again to Fig. 1,

[0085] A base station according to one embodiment may include a first artificial intelligence model.

[0086] A first artificial intelligence model included in a base station according to one embodiment will be described in detail with reference to FIG. 3.

[0087] FIG. 3 is a diagram illustrating a first artificial intelligence model learned at a base station according to one embodiment.

[0088] A base station according to one embodiment may include a first artificial intelligence model that performs computation through a neural network.

[0089] The neural network (320) can be trained by receiving training data. The trained neural network (320) receives input data (310) through the input terminal (330), and the output terminal (350) can perform operations to analyze the input data (310) and output output data (360), which is the desired result. Operations through the neural network can be performed through a hidden layer (340). In FIG. 3, for convenience, the hidden layer (340) is simplified to be formed as a single layer, but the hidden layer (340) can be formed as multiple layers.

[0090] Referring again to Fig. 1,

[0091] A base station according to one embodiment can transmit capability information to a server.

[0092] A base station according to one embodiment may include a plurality of artificial intelligence models, including a first artificial intelligence model.

[0093] Capability information refers to information related to at least one artificial intelligence model supported by a base station according to one embodiment for the base station according to one embodiment to replace at least one block included in the receiving end with a first artificial intelligence model.

[0094] Here, "supported" may mean that it is operable at a base station according to one embodiment.

[0095] For example, in relation to artificial intelligence computing, capability information may include I / O memory bandwidth information and / or FLOPS (floating operations per second) of at least one artificial intelligence model supported at a base station according to one embodiment. However, it is not limited thereto.

[0096] For example, regarding the size of the artificial intelligence model, the capability information may include, but is not limited to, memory information of at least one artificial intelligence model supported by a base station according to one embodiment.

[0097] For example, capability information may include at least one of layer information, backbone information, or structural information of at least one artificial intelligence model supported at a base station according to one embodiment, but is not limited thereto.

[0098] Layer information may include information on whether an artificial intelligence model supported by a base station according to one embodiment supports an FC layer (fully connected layer) and / or a convolution layer, but is not limited thereto.

[0099] Backbone information may include information on whether an artificial intelligence model supported at a base station according to one embodiment is based on a Convolutional Neural Network (CNN) and / or a Transformer, but is not limited thereto.

[0100] The structure information may include information on whether an artificial intelligence model supported by a base station according to one embodiment has a ResNet structure and / or a SwinIR structure, but is not limited thereto.

[0101] A base station according to one embodiment can transmit capability information to a server in the form of an index.

[0102] The capability information in the form of an index transmitted by a base station to a server according to one embodiment will be explained with reference to FIG. 4.

[0103] FIG. 4 is a diagram illustrating capability information that a base station according to one embodiment transmits to a server according to one embodiment.

[0104] A base station according to one embodiment can assign an index to each of at least one artificial intelligence model supported by the base station according to one embodiment and transmit capability information of each artificial intelligence model to a server.

[0105] For example, a base station can transmit capability information of an artificial intelligence model assigned index 0 to a base station by assigning index 0 to an artificial intelligence model that has FLOPS of 10^6, memory of 100MB, and a supported layer of fc layer, to a server. However, it is not limited to this.

[0106] For example, a base station can transmit capability information of an artificial intelligence model assigned index 1 to an artificial intelligence model having FLOPS of 10^7, memory of 100MB, and a supported layer of fc layer to a server. However, it is not limited to this.

[0107] For example, a base station can transmit capability information of an artificial intelligence model assigned index 2 to a base station by assigning index 2 to an artificial intelligence model that has FLOPS of 10^8, memory of 200MB, and supports layers including an fc layer and a convolution layer, to a server. However, it is not limited to this.

[0108] For example, a base station can transmit capability information of an artificial intelligence model assigned index 3 to a base station, which has FLOPS of 10^9, memory of 200MB, and supports layers including an fc layer and a convolution layer, to a server. However, it is not limited to this.

[0109] For example, a base station can transmit capability information of an artificial intelligence model assigned index 4 to a base station, which has FLOPS of 10^10, memory of 500MB, and supports layers including an fc layer, a convolution layer, and a multi-head attention layer, to a server. However, it is not limited to this.

[0110] Referring again to Fig. 1,

[0111] In step 120, the base station may transmit block instruction information to the server indicating at least one block to be replaced by the first artificial intelligence model.

[0112] A base station according to one embodiment can select at least one block to be replaced with a first artificial intelligence model.

[0113] A base station according to one embodiment may select at least one block included in the receiving end as a block to be replaced with a first artificial intelligence model.

[0114] For example, the base station may select the channel estimation block and channel equalization block included in the receiver as blocks to replace with the first artificial intelligence model. However, it is not limited thereto.

[0115] For example, the base station may select the Layer & RE demapping block, channel estimation block, and channel equalization block included in the receiver as blocks to replace with the first artificial intelligence model.

[0116] For example, the base station may select a channel equalization block included in the receiver as a block to replace with the first artificial intelligence model.

[0117] For example, the base station may select the Layer & RE demapping block, channel estimation block, and / or channel equalization block included in the receiver as blocks to replace the first artificial intelligence model. However, it is not limited thereto.

[0118] A base station according to one embodiment can transmit block instruction information indicating at least one selected block to a server.

[0119] A base station according to one embodiment can transmit block instruction information to a server in the form of an index.

[0120] For example, the base station may transmit a table of types for at least one block included in the receiving end to the server. The base station may transmit block indication information to the server by assigning "1" to the transmitted table for at least one block that is replaced by the first artificial intelligence model, and assigning "0" to the transmitted table for at least one block that is not replaced by the first artificial intelligence model.

[0121] A specific operation in which a base station according to one embodiment transmits block instruction information to a server will be explained with reference to FIG. 5.

[0122] FIG. 5 is a flowchart illustrating the operation method of a base station according to one embodiment.

[0123] Steps 110, 120, and 130 of Fig. 5 correspond to steps 110, 120, and 130 of Fig. 1, respectively.

[0124] In step 122, the base station may receive a directive message from the server indicating whether training of the artificial intelligence model is necessary based on the inference value.

[0125] Based on capability information received from a base station according to one embodiment, the server can determine whether learning of an artificial intelligence model based on an inference value obtained from a second artificial intelligence model included in the server is necessary.

[0126] Here, the artificial intelligence model may include at least one artificial intelligence model including a first artificial intelligence model supported at a base station according to one embodiment.

[0127] The second artificial intelligence model may have a structure identical or similar to the first artificial intelligence model described in Figure 3.

[0128] In step 124, the base station can transmit block instruction information to the server in response to the instruction message.

[0129] Referring again to Fig. 1,

[0130] In step 130, the base station can receive from the server an inference value obtained from a second artificial intelligence model determined by the server based on capability information and block instruction information.

[0131] A specific operation in which a base station according to one embodiment receives an inference value from a server will be explained with reference to FIGS. 6 and FIGS. 7.

[0132] FIG. 6 is a flowchart illustrating the operation method of a base station according to one embodiment.

[0133] Steps 120, 130, and 140 of Fig. 6 correspond to steps 120, 130, and 140 of Fig. 1, respectively.

[0134] In step 132, the base station can transmit environmental information of the cell containing the base station to the server.

[0135] Environmental information may include, but is not limited to, at least one of: information about a cell containing a base station and adjacent cells; information on the connection status between a base station and a terminal; information on the number of terminals located in a cell containing a base station; network information; protocol information; Doppler information; information on the signal-to-noise ratio (SNR); information on line of sight (LoS) / non-line of sight (NLos); information on delay spread; information on bandwidth; or information on scheduling.

[0136] A base station according to one embodiment can receive a request message from a server requesting the transmission of environment information.

[0137] For example, a base station may receive a request message from a server requesting cell environment information for channel estimation. However, it is not limited thereto.

[0138] A base station according to one embodiment can transmit environment information to a server in response to a request message.

[0139] For example, a base station may transmit Doppler information related to channel estimation and / or information regarding Loss / Nlos to a server in response to a request message requesting cell environment information for channel estimation. However, it is not limited thereto.

[0140] In step 134, the base station can receive from the server the inference value of the second artificial intelligence model determined by the server based on environmental information.

[0141] FIG. 7 is a flowchart illustrating the operation method of a base station according to one embodiment.

[0142] Steps 120, 130, and 140 in Fig. 7 correspond to steps 120, 130, and 140 in Fig. 1, respectively.

[0143] In step 136, the base station can transmit input data to the server for the second artificial intelligence model to obtain an inference value.

[0144] The input data may refer to the input (205) of Fig. 2.

[0145] The second artificial intelligence model can be included in the server.

[0146] The server can obtain an inference value by inputting the received input data as input to a second artificial intelligence model. The inference value may be in the form of a soft label.

[0147] In step 138, the base station can receive an inference value obtained from the second artificial intelligence model based on input data from the server.

[0148] Referring again to Fig. 1,

[0149] In step 140, the base station can train the first artificial intelligence model based on the inference value.

[0150] A base station according to one embodiment can train a first artificial intelligence model based on comparing an inference value obtained by inputting input data into a first artificial intelligence model with an inference value received from a server.

[0151] A base station according to one embodiment can train a first artificial intelligence model based on comparing an inference value obtained by inputting input data into the first artificial intelligence model with a ground truth value. Here, the ground truth value may be in the form of a hard label.

[0152] Accordingly, by the base station receiving inference values ​​from the server and training an artificial intelligence model based on them, the base station, which has hardware constraints compared to the server, can not only simplify the training of the artificial intelligence model but also lighten the artificial intelligence model.

[0153] A specific method for training a first artificial intelligence model based on inference values ​​in a base station according to one embodiment will be explained with reference to FIG. 8.

[0154] FIG. 8 is a flowchart illustrating the operation method of a base station according to one embodiment.

[0155] Steps 130 and 140 of Fig. 8 correspond to steps 130 and 140 of Fig. 1, respectively.

[0156] In step 142, the base station may send a request message to the server requesting at least one hyperparameter for learning the first artificial intelligence model.

[0157] Hyperparameters may refer to initial input parameters used by a base station according to one embodiment to train a first artificial intelligence model.

[0158] Hyperparameters can be set by the server based on capability information and / or block instruction information.

[0159] A loss function for training a first artificial intelligence model at a base station according to one embodiment can be expressed by the following [Equation 1].

[0160] [Mathematical Formula 1]

[0161]

[0162] Loss can mean total loss.

[0163] x can represent the input.

[0164] W can represent the weights of the first artificial intelligence model.

[0165] can mean a balancing parameter that adjusts the ratio of the first and second terms on the right-hand side of [Equation 1].

[0166] CE can stand for cross entropy.

[0167] y can represent the ground truth.

[0168] can mean the softmax function.

[0169] can mean the inference value of the first artificial intelligence model.

[0170] T can represent a temperature parameter that determines the degree of smoothing of the inference values ​​of the second artificial intelligence model.

[0171] can mean the inference value of the second artificial intelligence model.

[0172] Hyperparameters may include, but are not limited to, balancing parameters and / or temperature parameters.

[0173] A base station according to one embodiment may transmit a request message to a server requesting information on at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model. However, it is not limited thereto.

[0174] In step 144, the base station may receive at least one hyperparameter from the server in response to the request message.

[0175] Hyperparameters can be obtained by the server based on capability information and / or block instruction information.

[0176] A base station according to one embodiment may receive information from a server regarding at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model. However, it is not limited thereto.

[0177] In step 146, the base station can train the first artificial intelligence model based on at least one hyperparameter.

[0178] A base station according to one embodiment can train a first artificial intelligence model so that the loss of [Equation 1] described above is minimized.

[0179] FIG. 9 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0180] In step 910, the server can receive from the base station information on the base station's capability to replace at least one block included in the base station's receiving end with the first artificial intelligence model.

[0181] Capability information refers to information related to at least one artificial intelligence model supported by the base station for the base station to replace at least one block included in the receiving end with the first artificial intelligence model.

[0182] Here, "supported" may mean that it is operable at a base station according to one embodiment.

[0183] In step 920, the server can receive block instruction information from the base station indicating at least one block to be replaced by the first artificial intelligence model.

[0184] A specific method for a server according to one embodiment to receive block instruction information from a base station that indicates at least one block to be replaced by a first artificial intelligence model will be described with reference to FIG. 10.

[0185] FIG. 10 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0186] Steps 910, 920, and 930 of Fig. 10 correspond to steps 910, 920, and 930 of Fig. 9, respectively.

[0187] In step 922, the server can determine whether training of an artificial intelligence model based on inference values ​​is necessary based on capability information.

[0188] For example, the server can determine whether training of an AI model based on inference values ​​is necessary based on FLOPS information of the AI ​​model supported by the base station included in the capability information, memory information of the AI ​​model supported by the base station, and layer information of the AI ​​model supported by the base station.

[0189] Referring to FIG. 4, a server according to one embodiment may determine that learning based on inference values ​​is required for an artificial intelligence model assigned index 0 and index 1, considering that the FLOPS is small and the memory is not large.

[0190] Referring to FIG. 4, a server according to one embodiment can determine that learning based on inference values ​​is required for an artificial intelligence model assigned index 2, considering that the FLOPS is not small but the memory is not large and there are two types of supported layers (fc layer and convolution layer).

[0191] Referring to FIG. 4, a server according to one embodiment can determine that, for an artificial intelligence model assigned index 3, there are two types of supported layers (fc layer and convolution layer), but considering that FLOPS is not small and memory is not small, it can determine that learning based on inference values ​​is not required for the artificial intelligence model assigned index 3.

[0192] Referring to FIG. 4, a server according to one embodiment can determine that, for an artificial intelligence model assigned index 4, there are three types of supported layers (fc layer, convolution layer, and multi-head attention layer), but considering that FLOPS is not small and memory is not small, it can determine that learning based on inference values ​​is not required for the artificial intelligence model assigned index 4.

[0193] In step 924, the server can transmit a directive message to the base station indicating whether training of the artificial intelligence model based on the inference value is necessary based on the decision.

[0194] A server according to one embodiment may transmit an instruction message to a base station with a value of 0 or 1 indicating whether training of an artificial intelligence model based on an inference value is necessary. An instruction message having a value of 1 may mean that training of an artificial intelligence model based on an inference value is necessary. An instruction message having a value of 0 may mean that training of an artificial intelligence model based on an inference value is not necessary.

[0195] In step 926, the server can receive block instruction information from the base station in response to the instruction message.

[0196] Referring again to Fig. 9,

[0197] In step 930, the server can determine a second artificial intelligence model to train the first artificial intelligence model based on capability information and block instruction information.

[0198] A specific method for a server to determine a second artificial intelligence model according to one embodiment will be explained with reference to FIG. 11.

[0199] FIG. 11 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0200] Steps 920, 930, and 940 of Fig. 11 correspond to steps 920, 930, and 940 of Fig. 9, respectively.

[0201] In step 932, the server can determine a group of artificial intelligence models including a plurality of artificial intelligence models corresponding to at least one block based on block instruction information.

[0202] The server can store a group of artificial intelligence models composed of multiple artificial intelligence models capable of replacing at least one block included in the receiving end of the base station.

[0203] For example, the server may store an AI model group A including AI model a, AI model b, AI model c, and AI model d that replace the Layer & RE demapping block, the channel estimation block, and the channel equalization block. However, it is not limited thereto.

[0204] For example, the server may store a group of AI models B including AI models d, e, and f that replace the channel estimation block and the channel equalization block. However, it is not limited thereto.

[0205] For example, the server may store an AI model group C including AI model g, AI model h, and AI model i that replace a channel equalization block. However, it is not limited thereto.

[0206] A server according to one embodiment can determine an artificial intelligence model group based on block instruction information.

[0207] For example, when a server receives block instruction information from a base station instructing it to replace the Layer & RE demapping block, the channel estimation block, and the channel equalization block with a first artificial intelligence model, it may determine an artificial intelligence model group A, which includes artificial intelligence model a, artificial intelligence model b, artificial intelligence model c, and artificial intelligence model d that replace the Layer & RE demapping block, the channel estimation block, and the channel equalization block among a plurality of artificial intelligence model groups, as an artificial intelligence model group for training the first artificial intelligence model.

[0208] For example, when a server receives block instruction information from a base station instructing it to replace a channel estimation block and a channel equalization block with a first artificial intelligence model, it may determine an artificial intelligence model group B, which includes artificial intelligence model d, artificial intelligence model e, and artificial intelligence model f that replace the channel estimation block and the channel equalization block among a plurality of artificial intelligence model groups, as an artificial intelligence model group for training the first artificial intelligence model.

[0209] For example, when a server receives block instruction information from a base station instructing it to replace a channel equalization block with a first artificial intelligence model, it may determine an artificial intelligence model group C, which includes artificial intelligence model g, artificial intelligence model h, and artificial intelligence model i among a plurality of artificial intelligence model groups that replace the channel equalization block, as an artificial intelligence model group for training the first artificial intelligence model.

[0210] In step 934, the server can receive environment information of the cell containing the base station from the base station.

[0211] A specific operation for receiving environment information of a cell including a base station according to one embodiment from a base station will be explained with reference to FIG. 12.

[0212] FIG. 12 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0213] Steps 932, 934, and 936 of Fig. 12 correspond to steps 932, 934, and 936 of Fig. 11, respectively.

[0214] In step 934a, the server may send a request message to the base station requesting the transmission of environment information based on block instruction information.

[0215] Environmental information may include, but is not limited to, at least one of: information about a cell containing a base station and adjacent cells; information on the connection status between a base station and a terminal; information on the number of terminals located in a cell containing a base station; network information; protocol information; Doppler information; information on the signal-to-noise ratio (SNR); information on line of sight (LoS) / non-line of sight (NLos); information on delay spread; information on bandwidth; or information on scheduling.

[0216] A server according to one embodiment can recognize, based on block instruction information, that a block to be replaced by a first artificial intelligence model at a base station is a block related to channel estimation.

[0217] For example, if the block instruction information includes information related to "replacing the channel estimation block included in the receiving end with the first artificial intelligence model," the server may recognize that the block to be replaced with the first artificial intelligence model at the base station is a block related to channel estimation. However, it is not limited thereto.

[0218] According to one embodiment, if the server recognizes that the block to be replaced by the first artificial intelligence model at the base station is a block related to channel estimation, it may transmit a request message to the base station requesting the transmission of environment information related to channel estimation.

[0219] Environmental information related to channel estimation may include, but is not limited to, Doppler information related to the channel and information regarding Los (line of sight) / NLoS (non-line of sight).

[0220] A server according to one embodiment can recognize, based on block instruction information, that a block to be replaced by a first artificial intelligence model at a base station is a block related to detection.

[0221] For example, if the block indication information includes information related to "replacing the channel equalization block and soft demodulation block included in the receiving end with the first artificial intelligence model," the server may recognize that the block to be replaced with the first artificial intelligence model at the base station is a block related to detection. However, it is not limited thereto.

[0222] According to one embodiment, if the server recognizes that the block to be replaced by the first artificial intelligence model at the base station is a block related to detection, it may send a request message to the base station requesting the transmission of environment information related to detection.

[0223] Environmental information related to detection may include, but is not limited to, information on the signal-to-noise ratio (SNR).

[0224] In step 934b, the server can receive environment information from the base station in response to the request message.

[0225] The base station can transmit environment information corresponding to the request message received from the server to the server.

[0226] For example, if a base station receives a request message requesting the transmission of environmental information related to channel estimation, the base station can transmit the environmental information related to channel estimation to the server.

[0227] For example, if a base station receives a request message requesting the transmission of environmental information related to detection, the base station can transmit the environmental information related to detection to the server.

[0228] A server according to one embodiment can receive environment information corresponding to a request message from a base station.

[0229] Referring again to Fig. 11,

[0230] In step 936, the server may determine, based on environmental information, a second artificial intelligence model included in the artificial intelligence model group as an artificial intelligence model for training a first artificial intelligence model.

[0231] A server according to one embodiment may determine an artificial intelligence model learned in the same environment as the environmental information received from a base station as an artificial intelligence model for training a first artificial intelligence model.

[0232] For example, if the server determines, based on block instruction information, an AI model group A including AI model a, AI model b, AI model c, and AI model d as an AI model group for training a first AI model, the server may determine, among the plurality of AI models included in AI model group A, AI model a trained in the same environment as the environment information received from the base station as an AI model for training the first AI model. In this case, the second AI model may be AI model a. However, it is not limited thereto.

[0233] In one embodiment, if the artificial intelligence model learned in the same environment as the environmental information received from the base station is not included in the determined artificial intelligence model group, the server can learn and / or acquire a generalized artificial intelligence model based on the data set used to learn a plurality of artificial intelligence models included in the artificial intelligence model group.

[0234] For example, if the server determines an AI model group A, which includes AI model a, AI model b, AI model c, and AI model d, as an AI model group for training a first AI model based on block instruction information, but among the multiple AI models included in AI model group A, there is no AI model trained in the same environment as the environment information received from the base station, the server can train and / or acquire a generalized AI model x based on the input data set used for training AI model a, AI model b, AI model c, and AI model d.

[0235] A server according to one embodiment may determine a generalized artificial intelligence model as an artificial intelligence model for training a first artificial intelligence model.

[0236] For example, if the server learns and / or acquires a generalized artificial intelligence model x based on the input data set used to learn artificial intelligence model a, artificial intelligence model b, artificial intelligence model c, and artificial intelligence model d, the server may determine the generalized artificial intelligence model x as an artificial intelligence model for learning the first artificial intelligence model. In this case, the second artificial intelligence model may be artificial intelligence model x.

[0237] A server according to one embodiment may determine a second artificial intelligence model included in an artificial intelligence model group as an artificial intelligence model for training a first artificial intelligence model based on capability information and environment information.

[0238] For example, in a case where the server determines an AI model group B including AI model d, AI model e, and AI model f as an AI model group for training a first AI model, if among the multiple AI models included in AI model group B, the AI ​​models trained in the same environment as the environment information received from the base station are AI model d and AI model e, the server may determine AI model d as an AI model for training the first AI model by considering the capability information of the base station (FLOPS, supported layers, etc.).

[0239] In one embodiment, if the artificial intelligence model learned in the same environment as the environmental information received from the base station is not included in the artificial intelligence model group, the server may determine a model using the result value of MMSE (minimum mean square error), maximum likelihood detection, or sphere decoding as a model for training the first artificial intelligence model.

[0240] Referring again to Fig. 9,

[0241] In step 940, the server can receive input data from the base station to obtain an inference value from the second artificial intelligence model.

[0242] In step 950, the server can obtain the inference value of the second artificial intelligence model by inputting the input data into the second artificial intelligence model.

[0243] In step 960, the server can transmit the inference value to the base station.

[0244] FIG. 13 is a flowchart illustrating a method of operation of a server according to one embodiment.

[0245] Step 960 of Fig. 13 corresponds to Step 960 of Fig. 9.

[0246] In step 1310, the server may receive a request message from the base station requesting at least one hyperparameter for training the first artificial intelligence model.

[0247] Hyperparameters may refer to initial input parameters used by the base station to train the first artificial intelligence model.

[0248] The server may receive a request message from a base station requesting information on at least one of a batch size, epoch, learning rate, or filter size required for training the first artificial intelligence model. However, it is not limited thereto.

[0249] In step 1320, the server can obtain at least one hyperparameter based on the request message.

[0250] Hyperparameters may include, but are not limited to, balancing parameters and / or temperature parameters.

[0251] A server according to one embodiment can set hyperparameters based on environment information.

[0252] For example, if the server has a high degree of similarity between the learning environment of the second artificial intelligence model and the environment of the cell of the base station, it can set the balancing parameter to a small value.

[0253] For example, if the similarity between the learning environment of the second artificial intelligence model and the environment of the base station cell is low, the server can set the balancing parameter high.

[0254] A server according to one embodiment can set hyperparameters based on capability information.

[0255] For example, if the difference between the model size of the second AI model and the model size of the first AI model is large, the server can set the temperature parameter to a large value.

[0256] For example, the server can set the temperature parameter to a small value if the difference between the model size of the second AI model and the model size of the first AI model is small.

[0257] A server according to one embodiment can acquire hyperparameters based on at least one artificial intelligence model included in the server according to one embodiment.

[0258] A server according to one embodiment may obtain information regarding at least one of a batch size, epoch, learning rate, or filter size required for training a first artificial intelligence model based on at least one artificial intelligence model included in the server according to one embodiment. However, it is not limited thereto.

[0259] A server according to one embodiment may obtain information regarding at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model based on capability information and / or environment information of a base station. However, it is not limited thereto.

[0260] In step 1330, the server can transmit at least one hyperparameter to the base station.

[0261] A server according to one embodiment can transmit information regarding at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model to a base station.

[0262] A server according to one embodiment can receive a first artificial intelligence model from a base station.

[0263] A server according to one embodiment can train a received first artificial intelligence model based on an inference value obtained from a second artificial intelligence model.

[0264] A server according to one embodiment can transmit a first artificial intelligence model, upon completion of training, to a base station.

[0265] A server according to one embodiment can transmit a second artificial intelligence model to a base station.

[0266] A base station according to one embodiment can train a first artificial intelligence model based on an inference value obtained from a received second artificial intelligence model.

[0267] FIG. 14 is a block diagram illustrating the components of a base station according to one embodiment.

[0268] A base station (1400) according to one embodiment may include a transceiver (1420) and at least one processor (1410) connected to the transceiver (1420).

[0269] A base station (1400) according to one embodiment may include a memory (not shown).

[0270] The memory may store various data, programs, or applications for driving and controlling a base station (1400) according to one embodiment. The memory may include non-volatile memory such as RAM (Random Access Memory) or SRAM (Static Random Access Memory), and volatile memory such as non-volatile memory including at least one of a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), and PROM (Programmable Read-Only Memory).

[0271] Instructions, data structures, and program code that can be read by the processor (1410) may be stored in the memory. In the following embodiments, the processor (1410) may be implemented by executing the instructions or code of the program stored in the memory.

[0272] The processor (1410) is configured to control a series of processes to enable the base station (1400) to operate according to the embodiments described below, and may be composed of one or more processors.

[0273] The processor (1410) may be composed of hardware components that perform arithmetic, logic, and input / output operations and signal processing. One or more processors included in the processor (1410) may be circuitry such as a System on Chip (SoC) or an Integrated Circuit (IC). The processor (210) may be composed of at least one of, for example, a Central Processing Unit, a microprocessor, a Graphic Processing Unit, Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), and Field Programmable Gate Arrays (FPGAs), but is not limited thereto.

[0274] The processor (1410) can write data to memory or read data stored in memory, and in particular, can process data according to predefined operation rules by executing a program or at least one instruction stored in memory.

[0275] The transceiver (1420) can communicate with an external device or server through at least one wired or wireless communication network by means of the processor (1410).

[0276] The transceiver (1420) may include at least one short-range communication module that performs communication according to a communication standard such as Bluetooth, Wi-Fi, BLE (Bluetooth Low Energy), NFC / RFID, Wi-Fi Direct, UWB, or ZIGBEE, and a long-range communication module that performs communication with a server to support long-range communication according to a long-range communication standard. The long-range communication module may perform communication through a communication network according to 3G, 4G, 5G and / or 6G communication standards, or a network for internet communication.

[0277] A processor (1410) according to one embodiment can transmit capability information of a base station to a server for replacing at least one block included in a receiving end with a first artificial intelligence model.

[0278] A base station (1400) according to one embodiment may include a first artificial intelligence model.

[0279] A base station (1400) according to one embodiment may include a first artificial intelligence model that performs computation through a neural network.

[0280] A neural network can be trained by receiving training data as input. The trained neural network receives input data at its input layer, and at its output layer, it can analyze the input data to perform operations aimed at producing the desired output data. Operations performed by the neural network can be carried out through hidden layers. Hidden layers can be formed from multiple layers.

[0281] A processor (1410) according to one embodiment can transmit capability information to a server.

[0282] A base station (1400) according to one embodiment may include a plurality of artificial intelligence models, including a first artificial intelligence model.

[0283] Capability information refers to information related to at least one artificial intelligence model supported at a base station (1400) according to one embodiment, for a processor (1410) according to one embodiment to replace at least one block included in a receiving end with a first artificial intelligence model.

[0284] Here, "supported" may mean that it is operable at a base station (1400) according to one embodiment.

[0285] For example, in relation to artificial intelligence computing, capability information may include I / O memory bandwidth information and / or FLOPS (floating operations per second) of at least one artificial intelligence model supported at a base station (1400) according to one embodiment. However, it is not limited thereto.

[0286] For example, regarding the size of the artificial intelligence model, the capability information may include memory information of at least one artificial intelligence model supported by a base station (1400) according to one embodiment, but is not limited thereto.

[0287] For example, capability information may include at least one of layer information, backbone information, or structural information of at least one artificial intelligence model supported by a base station (1400) according to one embodiment, but is not limited thereto.

[0288] Layer information may include information on whether an artificial intelligence model supported by a base station (1400) according to one embodiment supports an FC layer (fully connected layer) and / or a convolution layer, but is not limited thereto.

[0289] Backbone information may include information on whether the artificial intelligence model supported by the base station (1400) according to one embodiment is based on a Convolutional Neural Network (CNN) and / or a Transformer, but is not limited thereto.

[0290] The structure information may include information on whether the artificial intelligence model supported by the base station (1400) according to one embodiment has a ResNet structure and / or a SwinIR structure, but is not limited thereto.

[0291] A processor (1410) according to one embodiment can transmit capability information to a server in the form of an index.

[0292] A processor (1410) according to one embodiment can assign an index to each of at least one artificial intelligence model supported by a base station according to one embodiment and transmit capability information of each artificial intelligence model to a server.

[0293] For example, the processor (1410) can transmit capability information of the AI ​​model to which index 0 is assigned to a server by assigning index 0 to an AI model that has FLOPS of 10^6, memory of 100MB, and a supported layer of fc layer. However, it is not limited to this.

[0294] For example, the processor (1410) can transmit capability information of the AI ​​model to which index 1 is assigned to a server by assigning index 1 to an AI model that has FLOPS of 10^7, memory of 100MB, and a supported layer of fc layer. However, it is not limited to this.

[0295] For example, the processor (1410) can transmit capability information of the AI ​​model assigned index 2 to the server by assigning index 2 to an AI model having FLOPS of 10^8, memory of 200MB, and supported layers of fc layer and convolution layer. However, it is not limited to this.

[0296] For example, the processor (1410) can transmit capability information of the AI ​​model assigned index 3 to the server by assigning index 3 to an AI model having FLOPS of 10^9, memory of 200MB, and supported layers of fc layer and convolution layer. However, it is not limited to this.

[0297] For example, the processor (1410) can transmit capability information of the AI ​​model assigned index 4 to a server by assigning index 4 to an AI model having FLOPS of 10^10, memory of 500MB, and supporting layers of an fc layer, a convolution layer, and a multi-head attention layer. However, it is not limited thereto.

[0298] A processor (1410) according to one embodiment can transmit block instruction information to a server indicating at least one block that is replaced by a first artificial intelligence model.

[0299] A processor (1410) according to one embodiment can select at least one block to be replaced with a first artificial intelligence model.

[0300] A processor (1410) according to one embodiment may select at least one block included in the receiving end as a block to replace with the first artificial intelligence model.

[0301] For example, the processor (1410) may select the channel estimation block and channel equalization block included in the receiver as blocks to replace with the first artificial intelligence model. However, it is not limited thereto.

[0302] For example, the processor (1410) may select the Layer & RE demapping block, channel estimation block, and channel equalization block included in the receiver as blocks to replace the first artificial intelligence model.

[0303] For example, the processor (1410) can select a channel equalization block included in the receiver as a block to replace with the first artificial intelligence model.

[0304] For example, the processor (1410) may select the Layer & RE demapping block, channel estimation block and / or channel equalization block included in the receiver as blocks to replace the first artificial intelligence model. However, it is not limited thereto.

[0305] A processor (1410) according to one embodiment can transmit block instruction information indicating at least one selected block to a server.

[0306] A processor (1410) according to one embodiment can transmit block instruction information to a server in the form of an index.

[0307] For example, the processor (1410) can transmit a table of types for at least one block included in the receiving end to the server. The base station can transmit block instruction information to the server by assigning "1" to the transmitted table for at least one block that is replaced by the first artificial intelligence model and assigning "0" to the transmitted table for at least one block that is not replaced by the first artificial intelligence model.

[0308] A processor (1410) according to one embodiment may receive a directive message from a server indicating whether training of an artificial intelligence model is necessary based on an inference value.

[0309] Based on capability information received from a processor (1410) according to one embodiment, the server can determine whether learning of an artificial intelligence model based on an inference value obtained from a second artificial intelligence model included in the server is necessary.

[0310] Here, the artificial intelligence model may include at least one artificial intelligence model including a first artificial intelligence model supported at a base station (1400) according to one embodiment.

[0311] A processor (1410) according to one embodiment can transmit block instruction information to a server in response to an instruction message.

[0312] A processor (1410) according to one embodiment can receive an inference value obtained from a second artificial intelligence model determined by a server based on capability information and block instruction information from the server.

[0313] A processor (1410) according to one embodiment can transmit environmental information of a cell including a base station to a server.

[0314] Environmental information may include at least one of the following: information about a cell containing the base station (1400) and an adjacent cell; information about the connection status between the base station (1400) and a terminal; information about the number of terminals located in the cell containing the base station (1400); network information; protocol information; Doppler information; information about the signal-to-noise ratio (SNR); information about the line of sight (LoS) / non-line of sight (NLos); information about the delay spread; information about bandwidth; or information about scheduling, but is not limited thereto.

[0315] A processor (1410) according to one embodiment may receive a request message from a server requesting the transmission of environment information.

[0316] For example, the processor (1410) may receive a request message from the server requesting environment information of the cell for channel estimation. However, it is not limited thereto.

[0317] A processor (1410) according to one embodiment can transmit environment information to a server in response to a request message.

[0318] For example, the processor (1410) may transmit Doppler information and / or Loss / Nlos information related to channel estimation to the server in response to a request message requesting environment information of the cell for channel estimation. However, it is not limited thereto.

[0319] A processor (1410) according to one embodiment can receive from the server an inference value of a second artificial intelligence model determined by the server based on environmental information.

[0320] A processor (1410) according to one embodiment can transmit input data to a server for the second artificial intelligence model to obtain an inference value.

[0321] The input data may refer to the input (205) of Fig. 2.

[0322] The second artificial intelligence model can be included in the server.

[0323] The server can obtain an inference value by inputting the received input data as input to a second artificial intelligence model. The inference value may be in the form of a soft label.

[0324] A processor (1410) according to one embodiment can receive an inference value obtained from a second artificial intelligence model based on input data from a server.

[0325] A processor (1410) according to one embodiment can train a first artificial intelligence model based on inference values.

[0326] A processor (1410) according to one embodiment can train a first artificial intelligence model based on comparing an inference value obtained by inputting input data into the first artificial intelligence model with an inference value received from a server.

[0327] A processor (1410) according to one embodiment can train a first artificial intelligence model based on comparing an inference value obtained by inputting input data into the first artificial intelligence model with a ground-truth value. Here, the ground-truth value may be in the form of a hard label.

[0328] A processor (1410) according to one embodiment can send a request message to a server requesting at least one hyperparameter for learning a first artificial intelligence model.

[0329] Hyperparameters may refer to initial input parameters used by a processor (1410) according to one embodiment to train a first artificial intelligence model.

[0330] Hyperparameters can be set by the server based on capability information and / or block instruction information.

[0331] Hyperparameters may include, but are not limited to, balancing parameters and / or temperature parameters.

[0332] A processor (1410) according to one embodiment may send a request message to a server requesting information on at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model. However, it is not limited thereto.

[0333] A processor (1410) according to one embodiment may receive at least one hyperparameter from a server in response to a request message.

[0334] Hyperparameters can be obtained by the server based on capability information and / or block instruction information.

[0335] A processor (1410) according to one embodiment may receive information from a server regarding at least one of a batch size, epoch, learning rate, or filter size required for training a first artificial intelligence model. However, it is not limited thereto.

[0336] A processor (1410) according to one embodiment can train a first artificial intelligence model based on at least one hyperparameter.

[0337] A processor (1410) according to one embodiment can train a first artificial intelligence model so that the Loss of the above-described [Equation 1] is minimized.

[0338] FIG. 15 is a block diagram illustrating the components of a server according to one embodiment.

[0339] A server (1500) according to one embodiment may include a transceiver (1520) and at least one processor (1510) connected to the transceiver (1520).

[0340] A server (1500) according to one embodiment may include memory (not shown).

[0341] The memory may store various data, programs, or applications for operating and controlling a server (1500) according to one embodiment. The memory may include non-volatile memory such as RAM (Random Access Memory) or SRAM (Static Random Access Memory), and volatile memory such as non-volatile memory including at least one of a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), and PROM (Programmable Read-Only Memory).

[0342] Instructions, data structures, and program code that can be read by the processor (1510) may be stored in the memory. In the following embodiments, the processor (1510) may be implemented by executing the instructions or code of the program stored in the memory.

[0343] The processor (1510) is configured to control a series of processes to enable the server (1500) to operate according to the embodiments described below, and may be composed of one or more processors.

[0344] The processor (1510) may be composed of hardware components that perform arithmetic, logic, and input / output operations and signal processing. One or more processors included in the processor (1510) may be circuitry such as a System on Chip (SoC) or an Integrated Circuit (IC). The processor (210) may be composed of at least one of, for example, a Central Processing Unit, a microprocessor, a Graphic Processing Unit, Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), and Field Programmable Gate Arrays (FPGAs), but is not limited thereto.

[0345] The processor (1510) can write data to memory or read data stored in memory, and in particular, can process data according to predefined operation rules by executing a program or at least one instruction stored in memory.

[0346] The transceiver (1520) can communicate with an external device or server through at least one wired or wireless communication network by means of the processor (1510).

[0347] The transceiver (1520) may include at least one short-range communication module that performs communication according to communication standards such as Bluetooth, Wi-Fi, BLE (Bluetooth Low Energy), NFC / RFID, Wi-Fi Direct, UWB, or ZIGBEE, and a long-range communication module that performs communication with a server to support long-range communication according to long-range communication standards. The long-range communication module may perform communication through a communication network according to 3G, 4G, 5G and / or 6G communication standards, or a network for internet communication.

[0348] A processor (1510) according to one embodiment can receive capability information from a base station for replacing at least one block included in the receiving end of a base station with a first artificial intelligence model.

[0349] Capability information refers to information related to at least one artificial intelligence model supported by the base station for the base station to replace at least one block included in the receiving end with the first artificial intelligence model.

[0350] Here, "supported" may mean that it is operable at a base station according to one embodiment.

[0351] A processor (1510) according to one embodiment can receive block instruction information from a base station that indicates at least one block to be replaced by a first artificial intelligence model.

[0352] A processor (1510) according to one embodiment can determine whether learning of an artificial intelligence model based on inference values ​​is necessary based on capability information.

[0353] For example, the processor (1510) can determine whether training of an artificial intelligence model based on inference values ​​is necessary based on FLOPS information of an artificial intelligence model supported by a base station included in capability information, memory information of an artificial intelligence model supported by a base station, and layer information of an artificial intelligence model supported by a base station.

[0354] Referring to FIG. 4, a processor (1510) according to one embodiment may determine that learning based on inference values ​​is required for an artificial intelligence model assigned index 0 and index 1, considering that the FLOPS is small and the memory is not large.

[0355] Referring to FIG. 4, a processor (1510) according to one embodiment can determine that learning based on inference values ​​is required for an artificial intelligence model assigned index 2, considering that the FLOPS is not small but the memory is not large and there are two types of supported layers (fc layer and convolution layer).

[0356] Referring to FIG. 4, a processor (1510) according to one embodiment can determine that, for an artificial intelligence model assigned index 3, learning based on inference values ​​is not required for the artificial intelligence model assigned index 3, considering that there are two types of supported layers (fc layer and convolution layer) but FLOPS is not small and memory is not small.

[0357] Referring to FIG. 4, a processor (1510) according to one embodiment can determine that, for an artificial intelligence model assigned index 4, there are three types of supported layers (fc layer, convolution layer, and multi-head attention layer), but considering that FLOPS is not small and memory is not small, it can determine that learning based on inference values ​​is not required for the artificial intelligence model assigned index 4.

[0358] A processor (1510) according to one embodiment can transmit a command message to a base station indicating whether learning of an artificial intelligence model based on an inference value is necessary based on a decision.

[0359] A processor (1510) according to one embodiment may transmit an instruction message to a base station with a value of 0 or 1 indicating whether training of an artificial intelligence model based on an inference value is necessary. An instruction message having a value of 1 may mean that training of an artificial intelligence model based on an inference value is necessary. An instruction message having a value of 0 may mean that training of an artificial intelligence model based on an inference value is not necessary.

[0360] A processor (1510) according to one embodiment can receive block instruction information from a base station in response to an instruction message.

[0361] A processor (1510) according to one embodiment can determine a second artificial intelligence model for training a first artificial intelligence model based on capability information and block instruction information.

[0362] A processor (1510) according to one embodiment can determine an artificial intelligence model group including a plurality of artificial intelligence models corresponding to at least one block based on block instruction information.

[0363] A processor (1510) according to one embodiment can store in memory a group of artificial intelligence models composed of a plurality of artificial intelligence models that can replace at least one block included in the receiving end of a base station.

[0364] For example, the processor (1510) may store an AI model group A including AI model a, AI model b, AI model c, and AI model d that replace the Layer & RE demapping block, the channel estimation block, and the channel equalization block. However, it is not limited thereto.

[0365] For example, the processor (1510) may store a group of artificial intelligence models B including artificial intelligence models d, e, and f that replace the channel estimation block and the channel equalization block. However, it is not limited thereto.

[0366] For example, the processor (1510) may store an artificial intelligence model group C including artificial intelligence model g, artificial intelligence model h, and artificial intelligence model i that replace a channel equalization block. However, it is not limited thereto.

[0367] A processor (1510) according to one embodiment can determine an artificial intelligence model group based on block instruction information.

[0368] For example, when the processor (1510) receives block instruction information from the base station instructing to replace the Layer & RE demapping block, the channel estimation block, and the channel equalization block with the first artificial intelligence model, the processor (1510) may determine, among a plurality of artificial intelligence model groups, an artificial intelligence model group A including artificial intelligence model a, artificial intelligence model b, artificial intelligence model c, and artificial intelligence model d that replace the Layer & RE demapping block, the channel estimation block, and the channel equalization block, as the artificial intelligence model group for training the first artificial intelligence model.

[0369] For example, when the processor (1510) receives block instruction information from the base station instructing to replace the channel estimation block and the channel equalization block with the first artificial intelligence model, the processor (1510) may determine an artificial intelligence model group B, which includes artificial intelligence model d, artificial intelligence model e, and artificial intelligence model f that replace the channel estimation block and the channel equalization block among a plurality of artificial intelligence model groups, as the artificial intelligence model group for training the first artificial intelligence model.

[0370] For example, when the processor (1510) receives block instruction information from the base station instructing the replacement of a channel equalization block with a first artificial intelligence model, the processor (1510) may determine an artificial intelligence model group C, which includes artificial intelligence model g, artificial intelligence model h, and artificial intelligence model i that replace the channel equalization block among a plurality of artificial intelligence model groups, as an artificial intelligence model group for training the first artificial intelligence model.

[0371] A processor (1510) according to one embodiment can receive environment information of a cell including a base station from the base station.

[0372] A processor (1510) according to one embodiment can transmit a request message to a base station requesting the transmission of environment information based on block instruction information.

[0373] Environmental information may include, but is not limited to, at least one of: information about a cell containing a base station and adjacent cells; information on the connection status between a base station and a terminal; information on the number of terminals located in a cell containing a base station; network information; protocol information; Doppler information; information on the signal-to-noise ratio (SNR); information on line of sight (LoS) / non-line of sight (NLos); information on delay spread; information on bandwidth; or information on scheduling.

[0374] A processor (1510) according to one embodiment can recognize that a block to be replaced by a first artificial intelligence model at a base station based on block instruction information is a block related to channel estimation.

[0375] For example, if the block instruction information includes information related to "replacing the channel estimation block included in the receiving end with the first artificial intelligence model," the processor (1510) can recognize that the block to be replaced with the first artificial intelligence model at the base station is a block related to channel estimation. However, it is not limited thereto.

[0376] A processor (1510) according to one embodiment may transmit a request message to a base station requesting the transmission of environment information related to channel estimation when it is recognized that the block to be replaced by the first artificial intelligence model at the base station is a block related to channel estimation.

[0377] Environmental information related to channel estimation may include, but is not limited to, Doppler information related to the channel and information regarding Los (line of sight) / NLoS (non-line of sight).

[0378] A processor (1510) according to one embodiment can recognize, based on block instruction information, that a block to be replaced by a first artificial intelligence model at a base station is a block related to detection.

[0379] For example, if the block instruction information includes information related to "replacing the channel equalization block and soft demodulation block included in the receiving end with the first artificial intelligence model," the processor (1510) can recognize that the block to be replaced with the first artificial intelligence model at the base station is a block related to detection. However, it is not limited thereto.

[0380] A processor (1510) according to one embodiment may transmit a request message to a base station requesting the transmission of environment information related to detection when it is recognized that the block to be replaced by the first artificial intelligence model at the base station is a block related to detection.

[0381] Environmental information related to detection may include, but is not limited to, information on the signal-to-noise ratio (SNR).

[0382] A processor (1510) according to one embodiment can receive environment information from a base station in response to a request message.

[0383] The base station can transmit environment information corresponding to the request message received from the server (1500) to the server (1500).

[0384] For example, if a base station receives a request message requesting that it transmit environmental information related to channel estimation, the base station can transmit environmental information related to channel estimation to a server (1500).

[0385] For example, if a base station receives a request message requesting that it transmit environmental information related to detection, the base station can transmit environmental information related to detection to a server (1500).

[0386] A processor (1510) according to one embodiment can receive environment information corresponding to a request message from a base station.

[0387] A processor (1510) according to one embodiment may determine, based on environmental information, a second artificial intelligence model included in an artificial intelligence model group as an artificial intelligence model for training a first artificial intelligence model.

[0388] A processor (1510) according to one embodiment may determine an artificial intelligence model learned in the same environment as the environmental information received from the base station as an artificial intelligence model for training a first artificial intelligence model.

[0389] For example, if the processor (1510) determines, based on block instruction information, an AI model group A including AI model a, AI model b, AI model c, and AI model d as an AI model group for training a first AI model, the processor (1510) may determine AI model a, which is trained in the same environment as the environment information received from the base station among the plurality of AI models included in AI model group A, as an AI model for training a first AI model. In this case, the second AI model may be AI model a. However, it is not limited thereto.

[0390] In one embodiment, the processor (1510) can learn and / or acquire a generalized artificial intelligence model based on a data set used to learn a plurality of artificial intelligence models included in the artificial intelligence model group, if the artificial intelligence model learned in the same environment as the environmental information received from the base station is not included in the determined artificial intelligence model group.

[0391] For example, if the processor (1510) determines, based on block instruction information, an AI model group A including AI model a, AI model b, AI model c, and AI model d as an AI model group for training a first AI model, but among the multiple AI models included in AI model group A, there is no AI model trained in the same environment as the environment information received from the base station, the processor (1510) can train and / or acquire a generalized AI model x based on the input data set used for training AI model a, AI model b, AI model c, and AI model d.

[0392] A processor (1510) according to one embodiment may determine a generalized artificial intelligence model as an artificial intelligence model for training a first artificial intelligence model.

[0393] For example, if the processor (1510) learns and / or acquires a generalized artificial intelligence model x based on the input data set used to learn artificial intelligence model a, artificial intelligence model b, artificial intelligence model c, and artificial intelligence model d, the processor (1510) may determine the generalized artificial intelligence model x as an artificial intelligence model for learning the first artificial intelligence model. In this case, the second artificial intelligence model may be artificial intelligence model x.

[0394] A processor (1510) according to one embodiment may determine a second artificial intelligence model included in an artificial intelligence model group as an artificial intelligence model for training a first artificial intelligence model based on capability information and environment information.

[0395] For example, in the case where the processor (1510) determines an AI model group B including AI model d, AI model e, and AI model f as an AI model group for training the first AI model, if among the multiple AI models included in AI model group B, the AI ​​models that are trained in the same environment as the environment information received from the base station are AI model d and AI model e, the processor (1510) may determine AI model d as an AI model for training the first AI model by considering the capability information (FLOPS, supported layers, etc.) of the base station.

[0396] A processor (1510) according to one embodiment may determine, when an artificial intelligence model learned in the same environment as the environmental information received from a base station is not included in the artificial intelligence model group, a model using the result value of MMSE (minimum mean square error), maximum likelihood detection, or sphere decoding as a model for training a first artificial intelligence model.

[0397] A processor (1510) according to one embodiment can receive input data from a base station to obtain an inference value from a second artificial intelligence model.

[0398] A processor (1510) according to one embodiment can obtain an inference value of a second artificial intelligence model by inputting input data into a second artificial intelligence model.

[0399] A processor (1510) according to one embodiment can transmit an inference value to a base station.

[0400] A processor (1510) according to one embodiment can receive a request message from a base station requesting at least one hyperparameter for learning a first artificial intelligence model.

[0401] Hyperparameters may refer to initial input parameters used by the base station to train the first artificial intelligence model.

[0402] A processor (1510) according to one embodiment may receive a request message from a base station requesting information on at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model. However, it is not limited thereto.

[0403] A processor (1510) according to one embodiment can obtain at least one hyperparameter based on a request message.

[0404] Hyperparameters may include, but are not limited to, balancing parameters and / or temperature parameters.

[0405] A processor (1510) according to one embodiment can set hyperparameters based on environmental information.

[0406] For example, the processor (1510) can set the balancing parameter to a small value when the similarity between the learning environment of the second artificial intelligence model and the environment of the cell of the base station is high.

[0407] For example, the processor (1510) can set the balancing parameter high when the similarity between the learning environment of the second artificial intelligence model and the environment of the cell of the base station is low.

[0408] A processor (1510) according to one embodiment can set hyperparameters based on capability information.

[0409] For example, the processor (1510) can set the temperature parameter to be large when the difference between the model size of the second artificial intelligence model and the model size of the first artificial intelligence model is large.

[0410] For example, the processor (1510) can set the temperature parameter to be small when the difference between the model size of the second artificial intelligence model and the model size of the first artificial intelligence model is small.

[0411] A processor (1510) according to one embodiment can obtain hyperparameters based on at least one artificial intelligence model included in a server (1500) according to one embodiment.

[0412] A processor (1510) according to one embodiment may obtain information regarding at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model based on at least one artificial intelligence model included in a server (1500) according to one embodiment. However, it is not limited thereto.

[0413] A processor (1510) according to one embodiment may obtain information regarding at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model based on capability information and / or environment information of a base station. However, it is not limited thereto.

[0414] A processor (1510) according to one embodiment can transmit at least one hyperparameter to a base station.

[0415] A processor (1510) according to one embodiment can transmit information about at least one of a batch size, epoch, learning rate, or filter size required for learning a first artificial intelligence model to a base station.

[0416] A processor (1510) according to one embodiment can receive a first artificial intelligence model from a base station.

[0417] A processor (1510) according to one embodiment can train a received first artificial intelligence model based on inference values ​​obtained from a second artificial intelligence model.

[0418] A processor (1510) according to one embodiment can transmit a first artificial intelligence model that has completed learning to a base station.

[0419] A processor (1510) according to one embodiment can transmit a second artificial intelligence model to a base station.

[0420] A base station according to one embodiment can train a first artificial intelligence model based on an inference value obtained from a received second artificial intelligence model.

[0421] A method of operation of a base station according to one embodiment may include the step of transmitting capability information of the base station to a server for replacing at least one block included in a receiving end with a first artificial intelligence model.

[0422] A method of operation of a base station according to one embodiment may include the step of transmitting block instruction information to the server that indicates the at least one block that is replaced by the first artificial intelligence model.

[0423] A method of operation of a base station according to one embodiment may include the step of receiving from the server an inference value obtained from a second artificial intelligence model determined by the server based on the capability information and the block instruction information.

[0424] A method of operating a base station according to one embodiment may include the step of training the first artificial intelligence model based on the inference value.

[0425] A receiving end according to one embodiment may include at least one of a channel estimation block or a channel equalization block.

[0426] A method of operating a base station according to one embodiment may include the step of transmitting environment information of a cell containing the base station to the server.

[0427] A method of operation of a base station according to one embodiment may include the step of receiving from the server an inference value of the second artificial intelligence model determined by the server based on the environment information.

[0428] A method of operation of a base station according to one embodiment may include the step of transmitting input data to the server for the second artificial intelligence model to obtain an inference value.

[0429] A method of operation of a base station according to one embodiment may include the step of receiving an inference value obtained from the second artificial intelligence model based on the input data from the server.

[0430] A method of operation of a base station according to one embodiment may include the step of transmitting a request message to the server requesting at least one hyperparameter for learning the first artificial intelligence model.

[0431] A method of operation of a base station according to one embodiment may include the step of receiving at least one hyperparameter from the server in response to the request message.

[0432] A method of operating a base station according to one embodiment may include the step of training the first artificial intelligence model based on the at least one hyperparameter.

[0433] A hyperparameter according to one embodiment may include a temperature parameter.

[0434] A method of operation of a base station according to one embodiment may include the step of receiving a command message from the server indicating whether training of the first artificial intelligence model based on the inference value is necessary.

[0435] A method of operation of a base station according to one embodiment may include the step of transmitting the block instruction information to the server in response to the instruction message.

[0436] A method of operation of a server according to one embodiment may include the step of receiving capability information of the base station from the base station for replacing at least one block included in the receiving end of the base station with a first artificial intelligence model.

[0437] A method of operation of a server according to one embodiment may include the step of receiving block instruction information from the base station that indicates the at least one block to be replaced by the first artificial intelligence model.

[0438] A method of operation of a server according to one embodiment may include the step of determining a second artificial intelligence model for training the first artificial intelligence model based on the capability information and the block instruction information.

[0439] A method of operation of a server according to one embodiment may include the step of receiving input data from the base station to obtain an inference value from the second artificial intelligence model.

[0440] A method of operation of a server according to one embodiment may include the step of obtaining an inference value of the second artificial intelligence model by inputting the input data into the second artificial intelligence model.

[0441] A method of operation of a server according to one embodiment may include the step of transmitting the inference value to the base station.

[0442] A method of operation of a server according to one embodiment may include the step of determining an artificial intelligence model group comprising a plurality of artificial intelligence models corresponding to at least one block based on the block instruction information.

[0443] A method of operation of a server according to one embodiment may include the step of receiving environment information of a cell containing the base station from the base station.

[0444] A method of operation of a server according to one embodiment may include the step of determining, based on the environment information, the second artificial intelligence model included in the artificial intelligence model group as an artificial intelligence model for training the first artificial intelligence model.

[0445] A method of operation of a server according to one embodiment may include the step of transmitting a request message to the base station requesting the transmission of environment information based on the block instruction information.

[0446] A method of operation of a server according to one embodiment may include the step of receiving environment information from the base station in response to the request message.

[0447] A method of operation of a server according to one embodiment may include a step of determining whether learning of the first artificial intelligence model based on the inference value is necessary based on the capability information.

[0448] A method of operation of a server according to one embodiment may include the step of transmitting a command message to the base station indicating whether training of the first artificial intelligence model based on the inference value is necessary, based on the above determination.

[0449] A method of operation of a server according to one embodiment may include the step of receiving block instruction information from the base station in response to the instruction message.

[0450] A method of operation of a server according to one embodiment may include the step of receiving a request message from the base station requesting at least one hyperparameter for learning the first artificial intelligence model.

[0451] A method of operation of a server according to one embodiment may include the step of obtaining at least one hyperparameter including a temperature parameter based on the request message.

[0452] A method of operation of a server according to one embodiment may include the step of transmitting the at least one hyperparameter to the base station.

[0453] A base station according to one embodiment may include a transceiver and at least one processor connected to the transceiver.

[0454] A processor according to one embodiment can transmit capability information of the base station to a server for replacing at least one block included in the receiving end with a first artificial intelligence model.

[0455] A processor according to one embodiment can transmit block instruction information to the server that indicates the at least one block to be replaced by the first artificial intelligence model.

[0456] A processor according to one embodiment can receive from the server an inference value obtained from a second artificial intelligence model determined by the server based on the capability information and the block instruction information.

[0457] A processor according to one embodiment can train the first artificial intelligence model based on the inference value.

[0458] A processor according to one embodiment can transmit environmental information of a cell including the base station to the server.

[0459] A processor according to one embodiment can receive from the server an inference value of the second artificial intelligence model determined by the server based on the environment information.

[0460] A processor according to one embodiment can transmit input data to the server for the second artificial intelligence model to obtain an inference value.

[0461] A processor according to one embodiment can receive an inference value obtained from the second artificial intelligence model based on the input data from the server.

[0462] A processor according to one embodiment can transmit a request message to the server requesting at least one hyperparameter for learning the first artificial intelligence model.

[0463] A processor according to one embodiment may receive at least one hyperparameter from the server in response to the request message.

[0464] A processor according to one embodiment can train the first artificial intelligence model based on the at least one hyperparameter.

[0465] A processor according to one embodiment may receive an instruction message from the server indicating whether training of the first artificial intelligence model based on the inference value is necessary.

[0466] A processor according to one embodiment may transmit the block instruction information to the server in response to the instruction message.

[0467] A server according to one embodiment may include a transceiver and at least one processor connected to the transceiver.

[0468] A processor according to one embodiment can receive capability information of the base station from the base station for replacing at least one block included in the receiving end of the base station with a first artificial intelligence model.

[0469] A processor according to one embodiment can receive block instruction information from the base station that indicates the at least one block that is replaced by the first artificial intelligence model.

[0470] A processor according to one embodiment can determine a second artificial intelligence model for training the first artificial intelligence model based on the capability information and the block instruction information.

[0471] A processor according to one embodiment can receive input data from the base station for obtaining an inference value from the second artificial intelligence model.

[0472] A processor according to one embodiment can obtain an inference value of the second artificial intelligence model by inputting the input data into the second artificial intelligence model.

[0473] A processor according to one embodiment can transmit the inference value to the base station.

[0474] A processor according to one embodiment can determine an artificial intelligence model group including a plurality of artificial intelligence models corresponding to at least one block based on the block instruction information.

[0475] A processor according to one embodiment can receive environment information of a cell including the base station from the base station.

[0476] A processor according to one embodiment may determine, based on the environment information, the second artificial intelligence model included in the artificial intelligence model group as an artificial intelligence model for training the first artificial intelligence model.

[0477] A processor according to one embodiment can transmit a request message to the base station requesting the transmission of the environment information based on the block instruction information.

[0478] A processor according to one embodiment can receive environment information from the base station in response to the request message.

[0479] A processor according to one embodiment can determine whether learning of the first artificial intelligence model based on the inference value is necessary based on the capability information.

[0480] A processor according to one embodiment can transmit an instruction message to the base station indicating whether training of the first artificial intelligence model based on the inference value is necessary, based on the above determination.

[0481] A processor according to one embodiment can receive block instruction information from the base station in response to the instruction message.

[0482] A processor according to one embodiment can receive a request message from the base station requesting at least one hyperparameter for learning the first artificial intelligence model.

[0483] A processor according to one embodiment can obtain at least one hyperparameter including a temperature parameter based on the request message.

[0484] A processor according to one embodiment can transmit the at least one hyperparameter to the base station.

Claims

1. In a method of operation of a base station of a wireless communication system, A step (S110) of transmitting capability information of the base station to a server for replacing at least one block included in the receiving end with a first artificial intelligence model; A step (S120) of transmitting block instruction information indicating at least one block that is replaced by the first artificial intelligence model to the server; A step (S130) of receiving from the server an inference value obtained from a second artificial intelligence model determined by the server based on the above capability information and the above block instruction information; and The method includes the step (S140) of training the first artificial intelligence model based on the inference value above, and A method of operation in which the receiving end comprises at least one of a channel estimation block or a channel equalization block.

2. In paragraph 1, the step (S120) of transmitting the block instruction information to the server is, A step of receiving a command message from the server indicating whether training of an artificial intelligence model based on the above inference value is necessary (S122); and A method of operation comprising the step (S124) of transmitting the block instruction information to the server in response to the instruction message.

3. In paragraph 1, the step of receiving the inference value (S130) is, A step of transmitting environment information of a cell including the base station to the server (S132); and A method of operation further comprising the step (S134) of receiving from the server an inference value of the second artificial intelligence model determined by the server based on the above environment information.

4. In paragraph 1, the step of receiving the inference value (S130) is, The step of transmitting input data to the server for the second artificial intelligence model to obtain an inference value (S136); and A method of operation comprising the step (S138) of receiving an inference value obtained from the second artificial intelligence model based on the input data from the server.

5. In paragraph 1, the step (S140) of training the first artificial intelligence model is, A step (S142) of sending a request message to the server requesting at least one hyperparameter for learning the first artificial intelligence model; In response to the above request message, receiving at least one hyperparameter from the server (S144); and The method includes the step (S146) of training the first artificial intelligence model based on at least one hyperparameter. The above hyperparameters include a temperature parameter, a method of operation.

6. In the method of operation of a server of a wireless communication system, A step (S910) of receiving capability information of the base station from the base station for replacing at least one block included in the receiving end of the base station with a first artificial intelligence model; A step (S920) of receiving block instruction information from the base station that indicates the at least one block to be replaced by the first artificial intelligence model; A step (S930) of determining a second artificial intelligence model for training the first artificial intelligence model based on the above capability information and the above block instruction information; A step (S940) of receiving input data from the base station to obtain an inference value from the second artificial intelligence model; A step of obtaining an inference value of the second artificial intelligence model by inputting the above input data into the second artificial intelligence model (S950); and A method of operation comprising the step (S960) of transmitting the above inference value to the base station.

7. In paragraph 6, the step of receiving the block instruction information (S920) is, A step (S922) of determining whether training of an artificial intelligence model based on the inference value is necessary based on the above capability information; Based on the above decision, a step of transmitting a command message to the base station indicating whether training of an artificial intelligence model based on the inference value is necessary (S924); and A method of operation comprising the step (S926) of receiving block instruction information from the base station in response to the above instruction message.

8. In paragraph 6, the step of determining the second artificial intelligence model (S930) is, A step (S932) of determining an artificial intelligence model group including a plurality of artificial intelligence models corresponding to at least one block based on the block instruction information above; A step of receiving environment information of a cell including the base station from the base station (S934); and A method of operation comprising the step (S936) of determining the second artificial intelligence model included in the artificial intelligence model group based on the above environmental information as an artificial intelligence model for training the first artificial intelligence model.

9. In paragraph 8, the step of receiving the environment information (S934) is, A step of transmitting a request message to the base station requesting the transmission of the environment information based on the block instruction information (S934a); and A method of operation comprising the step (S934b) of receiving environment information from the base station in response to the above request message.

10. In paragraph 6, the above method of operation is, A step (S1310) of receiving a request message from the base station requesting at least one hyperparameter for learning the first artificial intelligence model; Based on the above request message, a step (S1320) of obtaining at least one hyperparameter including a temperature parameter; and A method of operation further comprising the step (S1330) of transmitting at least one hyperparameter to the base station.

11. In a base station (1400) of a wireless communication system, the base station (1400) is, Transmitter / receiver (1420); and It includes at least one processor (1410) connected to the above transceiver (1420), and The above at least one processor (1410) is, Transmitting capability information of the base station to a server for replacing at least one block included in the receiving end with a first artificial intelligence model, and Block instruction information indicating at least one block that is replaced by the first artificial intelligence model is transmitted to the server, and Based on the above capability information and the above block instruction information, an inference value obtained from a second artificial intelligence model determined by the server is received from the server, and Based on the above inference value, the first artificial intelligence model is trained, and The above-mentioned receiving unit is a base station comprising at least one of a channel estimation block or a channel equalization block.

12. In paragraph 11, the above at least one processor (1410) is, Receive a command message from the server indicating whether training of an artificial intelligence model based on the above inference value is necessary, and A base station that transmits the block instruction information to the server in response to the above instruction message.

13. In paragraph 11, the above at least one processor (1410) is, Transmitting environment information of the cell including the above base station to the above server, and A base station that receives from the server the inference value of the second artificial intelligence model determined by the server based on the above environment information.

14. In paragraph 11, the above at least one processor (1410) is, The above second artificial intelligence model transmits input data to the server to obtain an inference value, and A base station that receives an inference value obtained from the second artificial intelligence model based on the above input data from the server.

15. In a server (1500) of a wireless communication system, the server (1500) is, Transmitter / receiver (1520); and It includes at least one processor (1510) connected to the above transceiver (1520), and The above at least one processor (1510) is, Receiving capability information of the base station from the base station for replacing at least one block included in the receiving end of the base station with a first artificial intelligence model, and Receiving block instruction information from the base station that indicates the at least one block to be replaced by the first artificial intelligence model, and Based on the above capability information and the above block instruction information, a second artificial intelligence model for training the first artificial intelligence model is determined, and Input data for obtaining an inference value from the above-mentioned second artificial intelligence model is received from the above-mentioned base station, and By inputting the above input data into the above second artificial intelligence model, the inference value of the above second artificial intelligence model is obtained, and A server that transmits the above inference value to the above base station.

Images