Methods for interoperable ai / ML model training

Abstract

A wireless transmit / receive unit (WTRU) may receive configuration that indicates one or more (e.g., a plurality of) datasets and / or a set of anchor vectors. The WTRU may encode the set of anchor vectors using an encoder to generate encoded anchor vectors. The WTRU may encode channel state information (CSI) using the encoder to generate encoded CSI. The WTRU may determine at least one relative representation based on the encoded anchor vector(s) and / or the encoded CSI. For example, the WTRU may determine a plurality of relative representations based on the encoded anchor vector(s) and / or the encoded CSI. The relative representation(s) may be indicative of similarity between the encoded anchor vector(s) and the encoded CSI. The WTRU may send an indication of the at least one relative representation to a base station, for example. For example, the WTRU may send an indication of the plurality of relative representation(s) to a base station.

Description

METHODS FOR INTEROPERABLE AI / ML MODEL TRAINING CROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims priority to United States Provisional Patent Application No.63 / 642,934 filed in the United States of America on May 6, 2024, the entire contents of which are incorporated herein by reference. BACKGROUND

[0002] Channel State Information may include at least one of the following: channel quality index (CQI), rank indicator (RI), precoding matrix index (PMI), an L1 channel measurement (e.g., RSRP such as L1- RSRP, or SINR), CSI-RS resource indicator (CRI), SS / PBCH block resource indicator (SSBRI), layer indicator (LI), and / or one or more (e.g., any) other measurement quantity(ies) measured by the wireless transmit / receive unit (WTRU) from the configured reference signals (e.g. CSI-RS or SS / PBCH block and / or one or more (e.g., any) other reference signals). SUMMARY

[0003] Methods for training compatible / interoperable two-sided models with minimum interaction and / or coordination using relative representations and / or transformation functions are described herein. The wireless transmit / receive unit (WTRU) may train the WTRU-sided model (the network (NW) may train the NW-sided model) independently to satisfy a preconfigured performance condition and / or the WTRU may achieve interoperability for inference by mapping the WTRU-side model output (e.g., encoded channel state information (CSI)) to a relative representation based on anchor vector configuration. The WTRU-side model (NW may train the NW-side model) independently to satisfy a preconfigured performance condition and / or the WTRU may achieve interoperability for inference by transmission of interoperability report characterizing the WTRU-side latent space – wherein the characterization may enable a transformation to NW-side latent space.

[0004] A wireless transmit / receive unit (WTRU) may receive configuration information indicating a plurality of datasets, training configuration, and a quantization of a relative representation. The WTRU may send a first report indicating WTRU capability associated with one or more artificial intelligence (AI) / machine learning (ML) models. The WTRU may encode one or more anchor vectors using an encoder model. The WTRU may generate the one or more encoded anchor vectors in latent space. The WTRU may determinean encoded input in the latent space, wherein determining the encoded input comprises deriving relative representation of input as function of encoded latent representation of input, encoded latent representation of anchors, and / or a similarity metric. The WTRU may apply the quantization to the derived relative representation. The WTRU may send a second report. The second report including the quantized relative representation.

[0005] The WTRU may train one or more artificial intelligence (AI) / machine learning (ML) models on one or more configured datasets of the plurality of datasets using the training configuration. The training configuration may include a set of anchor vectors, performance metrics, configuration to map from latent representation to relative representation, and / or dimensionality of relative representation.

[0006] A wireless transmit / receive unit (WTRU) may receive configuration information indicating a plurality of datasets, training configuration, and / or interoperability reporting configuration, The interoperability reporting configuration may include interoperability reporting resources. The WTRU may train one or more artificial intelligence (AI) / machine learning (ML) models on one or more configured datasets of the plurality of datasets using the training configuration. The WTRU may send a first report indicating WTRU capability. The WTRU may activate an encoder model to encode one or more anchor vectors. The WTRU may, upon a determination that a preconfigured trigger condition is satisfied, transmit a dataset identification (ID), anchor set ID, and / or the one or more anchor vectors, on the interoperability reporting resources. The WTRU may send a second report indicating quantized latent representation.

[0007] A WTRU may receive configuration. The configuration information may indicate one or more (e.g., a plurality of) datasets and / or a set of anchor vectors. The WTRU may encode the set of anchor vectors using an encoder to generate encoded anchor vectors. The WTRU may encode channel state information (CSI) using the encoder to generate encoded CSI. The WTRU may determine at least one relative representation based on the encoded anchor vector(s) and / or the encoded CSI. For example, the WTRU may determine a plurality of relative representations based on the encoded anchor vector(s) and / or the encoded CSI. The relative representation(s) may be indicative of similarity between the encoded anchor vector(s) and the encoded CSI. The WTRU may send an indication of the at least one relative representation to a base station, for example. For example, the WTRU may send an indication of the plurality of relative representation(s) to a base station. For example, the WTRU may send an indication of the quantized relative representation(s).

[0008] The set of anchor vectors may be associated with at least one dataset of a plurality of datasets. Each relative representation of the at least one relative representation may include and / or be a scalar valueand / or may be defined by the dimensionality of each anchor vector I the set of anchor vectors. The at least one relative representation and / or the plurality of relative representations may be indicative of compressed channel state information (CSI) feedback.

[0009] The WTRU may send an indication that indicates a capability of the WTRU associated with one or more artificial intelligence or machine learning (AI / ML) models.

[0010] The configuration information may include a plurality of datasets, training criteria, and / or training information. The WTRU may be configured to use each dataset of the plurality of datasets and / or the set of anchor vectors to train the encoder based on the training criteria and / or the training information such that the encoder is trained to be interoperable with at least one model at the base station.

[0011] The WTRU may quantize the at least one relative representation and / or the plurality of relative representations. The indication of the at least one relative representation and / or the plurality of relative representations may be an indication of the quantized at least one relative representation and / or the quantized plurality of relative representations.

[0012] The set of anchor vectors may be a subset of anchor vectors. The WTRU may determine the subset of anchor vectors by excluding at least one anchor vector included in the set of anchor vectors. The WTRU may encode the subset of anchor vectors. For example, the WTRU being configured to encode the set of anchor vectors may include the WTRU being configured to encode the subset of anchor vectors.

[0013] The configuration information may include a plurality of datasets. Each dataset of the plurality of datasets may be associated with a respective dataset identification (ID). The set of anchor vectors may be associated with an anchor set ID. The WTRU being configured to send the indication of the at least one relative representation and / or the indication of the plurality of relative representations may include the WTRU being configured to send an indication of each dataset ID and / or the anchor set ID.

[0014] The WTRU may receive a command. The configuration information may include interoperability criteria. The encoded anchor vectors may be generated based on the command and / or the interoperability criteria. The command may include a radio resource configuration (RRC) message and / or a system information block (SIB) message that indicates one or more dataset IDs and / or one or more anchor set IDs. The encoded anchor vectors may be generated based on the one or more dataset IDs and / or the one or more anchor set IDs. The interoperability criteria may be indicative of interoperability between a first model at the WTRU and a second model at the base station. The interoperability criteria may include one or more of: a bottle-neck dimension size, a quantization type, a number of quantization bits, and / or an input data type.

[0015] The configuration information may include an indication of a plurality of datasets. The plurality of datasets may be associated with the set of anchor vectors.

[0016] The encoded CSI may include one or more data points in a latent space. The WTRU may determine a distance between each data point in the latent space to each encoded anchor vector to determine the similarity between the encoded CSI and the encoded anchor vectors to determine the at least one relative representation.

[0017] The WTRU may receive an indication that indicates one or more datasets. The encoder may be trained based on the indication of the one or more datasets. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] FIG.1A is a system diagram illustrating an example communications system in which one or more disclosed embodiments may be implemented.

[0019] FIG.1B is a system diagram illustrating an example wireless transmit / receive unit (WTRU) that may be used within the communications system illustrated in FIG.1A according to an embodiment.

[0020] FIG.1C is a system diagram illustrating an example radio access network (RAN) and an example core network (CN) that may be used within the communications system illustrated in FIG.1A according to an embodiment.

[0021] FIG.1D is a system diagram illustrating a further example RAN and a further example CN that may be used within the communications system illustrated in FIG.1A according to an embodiment.

[0022] FIG.2 depicts an example of channel state information (CSI) measurement setting.

[0023] FIG.3 depicts an example representation of the relative representation-based interoperable encoding-decoding framework.

[0024] FIG.4 depicts an example setup for training network (NW) side decoder model to handle relative representations. DETAILED DESCRIPTION

[0025] FIG.1A is a diagram illustrating an example communications system 100 in which one or more disclosed embodiments may be implemented. The communications system 100 may be a multiple access system that provides content, such as voice, data, video, messaging, broadcast, etc., to multiple wireless users. The communications system 100 may enable multiple wireless users to access such content through the sharing of system resources, including wireless bandwidth. For example, the communications systems100 may employ one or more channel access methods, such as code division multiple access (CDMA), time division multiple access (TDMA), frequency division multiple access (FDMA), orthogonal FDMA (OFDMA), single-carrier FDMA (SC-FDMA), zero-tail unique-word DFT-Spread OFDM (ZT UW DTS-s OFDM), unique word OFDM (UW-OFDM), resource block-filtered OFDM, filter bank multicarrier (FBMC), and the like.

[0026] As shown in FIG.1A, the communications system 100 may include wireless transmit / receive units (WTRUs) 102a, 102b, 102c, 102d, a RAN 104 / 113, a CN 106 / 115, a public switched telephone network (PSTN) 108, the Internet 110, and other networks 112, though it will be appreciated that the disclosed embodiments contemplate any number of WTRUs, base stations, networks, and / or network elements. Each of the WTRUs 102a, 102b, 102c, 102d may be any type of device configured to operate and / or communicate in a wireless environment. By way of example, the WTRUs 102a, 102b, 102c, 102d, any of which may be referred to as a “station” and / or a “STA”, may be configured to transmit and / or receive wireless signals and may include a user equipment (UE), a mobile station, a fixed or mobile subscriber unit, a subscription-based unit, a pager, a cellular telephone, a personal digital assistant (PDA), a smartphone, a laptop, a netbook, a personal computer, a wireless sensor, a hotspot or Mi-Fi device, an Internet of Things (IoT) device, a watch or other wearable, a head-mounted display (HMD), a vehicle, a drone, a medical device and applications (e.g., remote surgery), an industrial device and applications (e.g., a robot and / or other wireless devices operating in an industrial and / or an automated processing chain contexts), a consumer electronics device, a device operating on commercial and / or industrial wireless networks, and the like. Any of the WTRUs 102a, 102b, 102c and 102d may be interchangeably referred to as a WTRU. Further, any description herein that is described with reference to a UE may be equally applicable to a WTRU (or vice versa). For example, a WTRU may be configured to perform any of the processes or procedures described herein as being performed by a UE (or vice versa).

[0027] The communications systems 100 may also include a base station 114a and / or a base station 114b. Each of the base stations 114a, 114b may be any type of device configured to wirelessly interface with at least one of the WTRUs 102a, 102b, 102c, 102d to facilitate access to one or more communication networks, such as the CN 106 / 115, the Internet 110, and / or the other networks 112. By way of example, the base stations 114a, 114b may be a base transceiver station (BTS), a Node-B, an eNode B, a Home Node B, a Home eNode B, a gNB, a NR NodeB, a site controller, an access point (AP), a wireless router, and the like. While the base stations 114a, 114b are each depicted as a single element, it will beappreciated that the base stations 114a, 114b may include any number of interconnected base stations and / or network elements.

[0028] The base station 114a may be part of the RAN 104 / 113, which may also include other base stations and / or network elements (not shown), such as a base station controller (BSC), a radio network controller (RNC), relay nodes, etc. The base station 114a and / or the base station 114b may be configured to transmit and / or receive wireless signals on one or more carrier frequencies, which may be referred to as a cell (not shown). These frequencies may be in licensed spectrum, unlicensed spectrum, or a combination of licensed and unlicensed spectrum. A cell may provide coverage for a wireless service to a specific geographical area that may be relatively fixed or that may change over time. The cell may further be divided into cell sectors. For example, the cell associated with the base station 114a may be divided into three sectors. Thus, in one embodiment, the base station 114a may include three transceivers, i.e., one for each sector of the cell. In an embodiment, the base station 114a may employ multiple-input multiple output (MIMO) technology and may utilize multiple transceivers for each sector of the cell. For example, beamforming may be used to transmit and / or receive signals in desired spatial directions.

[0029] The base stations 114a, 114b may communicate with one or more of the WTRUs 102a, 102b, 102c, 102d over an air interface 116, which may be any suitable wireless communication link (e.g., radio frequency (RF), microwave, centimeter wave, micrometer wave, infrared (IR), ultraviolet (UV), visible light, etc.). The air interface 116 may be established using any suitable radio access technology (RAT).

[0030] More specifically, as noted above, the communications system 100 may be a multiple access system and may employ one or more channel access schemes, such as CDMA, TDMA, FDMA, OFDMA, SC-FDMA, and the like. For example, the base station 114a in the RAN 104 / 113 and the WTRUs 102a, 102b, 102c may implement a radio technology such as Universal Mobile Telecommunications System (UMTS) Terrestrial Radio Access (UTRA), which may establish the air interface 115 / 116 / 117 using wideband CDMA (WCDMA). WCDMA may include communication protocols such as High-Speed Packet Access (HSPA) and / or Evolved HSPA (HSPA+). HSPA may include High-Speed Downlink (DL) Packet Access (HSDPA) and / or High-Speed UL Packet Access (HSUPA).

[0031] In an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement a radio technology such as Evolved UMTS Terrestrial Radio Access (E-UTRA), which may establish the air interface 116 using Long Term Evolution (LTE) and / or LTE-Advanced (LTE-A) and / or LTE-Advanced Pro (LTE-A Pro).

[0032] In an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement a radio technology such as NR Radio Access , which may establish the air interface 116 using New Radio (NR).

[0033] In an embodiment, the base station 114a and the WTRUs 102a, 102b, 102c may implement multiple radio access technologies. For example, the base station 114a and the WTRUs 102a, 102b, 102c may implement LTE radio access and NR radio access together, for instance using dual connectivity (DC) principles. Thus, the air interface utilized by WTRUs 102a, 102b, 102c may be characterized by multiple types of radio access technologies and / or transmissions sent to / from multiple types of base stations (e.g., a eNB and a gNB).

[0034] In other embodiments, the base station 114a and the WTRUs 102a, 102b, 102c may implement radio technologies such as IEEE 802.11 (i.e., Wireless Fidelity (WiFi), IEEE 802.16 (i.e., Worldwide Interoperability for Microwave Access (WiMAX)), CDMA2000, CDMA20001X, CDMA2000 EV-DO, Interim Standard 2000 (IS-2000), Interim Standard 95 (IS-95), Interim Standard 856 (IS-856), Global System for Mobile communications (GSM), Enhanced Data rates for GSM Evolution (EDGE), GSM EDGE (GERAN), and the like.

[0035] The base station 114b in FIG.1A may be a wireless router, Home Node B, Home eNode B, or access point, for example, and may utilize any suitable RAT for facilitating wireless connectivity in a localized area, such as a place of business, a home, a vehicle, a campus, an industrial facility, an air corridor (e.g., for use by drones), a roadway, and the like. In one embodiment, the base station 114b and the WTRUs 102c, 102d may implement a radio technology such as IEEE 802.11 to establish a wireless local area network (WLAN). In an embodiment, the base station 114b and the WTRUs 102c, 102d may implement a radio technology such as IEEE 802.15 to establish a wireless personal area network (WPAN). In yet another embodiment, the base station 114b and the WTRUs 102c, 102d may utilize a cellular-based RAT (e.g., WCDMA, CDMA2000, GSM, LTE, LTE-A, LTE-A Pro, NR etc.) to establish a picocell or femtocell. As shown in FIG.1A, the base station 114b may have a direct connection to the Internet 110. Thus, the base station 114b may not be required to access the Internet 110 via the CN 106 / 115.

[0036] The RAN 104 / 113 may be in communication with the CN 106 / 115, which may be any type of network configured to provide voice, data, applications, and / or voice over internet protocol (VoIP) services to one or more of the WTRUs 102a, 102b, 102c, 102d. The data may have varying quality of service (QoS) requirements, such as differing throughput requirements, latency requirements, error tolerance requirements, reliability requirements, data throughput requirements, mobility requirements, and the like.The CN 106 / 115 may provide call control, billing services, mobile location-based services, pre-paid calling, Internet connectivity, video distribution, etc., and / or perform high-level security functions, such as user authentication. Although not shown in FIG.1A, it will be appreciated that the RAN 104 / 113 and / or the CN 106 / 115 may be in direct or indirect communication with other RANs that employ the same RAT as the RAN 104 / 113 or a different RAT. For example, in addition to being connected to the RAN 104 / 113, which may be utilizing a NR radio technology, the CN 106 / 115 may also be in communication with another RAN (not shown) employing a GSM, UMTS, CDMA 2000, WiMAX, E-UTRA, or WiFi radio technology.

[0037] The CN 106 / 115 may also serve as a gateway for the WTRUs 102a, 102b, 102c, 102d to access the PSTN 108, the Internet 110, and / or the other networks 112. The PSTN 108 may include circuit- switched telephone networks that provide plain old telephone service (POTS). The Internet 110 may include a global system of interconnected computer networks and devices that use common communication protocols, such as the transmission control protocol (TCP), user datagram protocol (UDP) and / or the internet protocol (IP) in the TCP / IP internet protocol suite. The networks 112 may include wired and / or wireless communications networks owned and / or operated by other service providers. For example, the networks 112 may include another CN connected to one or more RANs, which may employ the same RAT as the RAN 104 / 113 or a different RAT.

[0038] Some or all of the WTRUs 102a, 102b, 102c, 102d in the communications system 100 may include multi-mode capabilities (e.g., the WTRUs 102a, 102b, 102c, 102d may include multiple transceivers for communicating with different wireless networks over different wireless links). For example, the WTRU 102c shown in FIG.1A may be configured to communicate with the base station 114a, which may employ a cellular-based radio technology, and with the base station 114b, which may employ an IEEE 802 radio technology.

[0039] FIG.1B is a system diagram illustrating an example WTRU 102. As shown in FIG.1B, the WTRU 102 may include a processor 118, a transceiver 120, a transmit / receive element 122, a speaker / microphone 124, a keypad 126, a display / touchpad 128, non-removable memory 130, removable memory 132, a power source 134, a global positioning system (GPS) chipset 136, and / or other peripherals 138, among others. It will be appreciated that the WTRU 102 may include any sub-combination of the foregoing elements while remaining consistent with an embodiment.

[0040] The processor 118 may be a general purpose processor, a special purpose processor, a conventional processor, a digital signal processor (DSP), a plurality of microprocessors, one or more microprocessors in association with a DSP core, a controller, a microcontroller, Application SpecificIntegrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) circuits, any other type of integrated circuit (IC), a state machine, and the like. The processor 118 may perform signal coding, data processing, power control, input / output processing, and / or any other functionality that enables the WTRU 102 to operate in a wireless environment. The processor 118 may be coupled to the transceiver 120, which may be coupled to the transmit / receive element 122. While FIG.1B depicts the processor 118 and the transceiver 120 as separate components, it will be appreciated that the processor 118 and the transceiver 120 may be integrated together in an electronic package or chip.

[0041] The transmit / receive element 122 may be configured to transmit signals to, or receive signals from, a base station (e.g., the base station 114a) over the air interface 116. For example, in one embodiment, the transmit / receive element 122 may be an antenna configured to transmit and / or receive RF signals. In an embodiment, the transmit / receive element 122 may be an emitter / detector configured to transmit and / or receive IR, UV, or visible light signals, for example. In yet another embodiment, the transmit / receive element 122 may be configured to transmit and / or receive both RF and light signals. It will be appreciated that the transmit / receive element 122 may be configured to transmit and / or receive any combination of wireless signals.

[0042] Although the transmit / receive element 122 is depicted in FIG.1B as a single element, the WTRU 102 may include any number of transmit / receive elements 122. More specifically, the WTRU 102 may employ MIMO technology. Thus, in one embodiment, the WTRU 102 may include two or more transmit / receive elements 122 (e.g., multiple antennas) for transmitting and receiving wireless signals over the air interface 116.

[0043] The transceiver 120 may be configured to modulate the signals that are to be transmitted by the transmit / receive element 122 and to demodulate the signals that are received by the transmit / receive element 122. As noted above, the WTRU 102 may have multi-mode capabilities. Thus, the transceiver 120 may include multiple transceivers for enabling the WTRU 102 to communicate via multiple RATs, such as NR and IEEE 802.11, for example.

[0044] The processor 118 of the WTRU 102 may be coupled to, and may receive user input data from, the speaker / microphone 124, the keypad 126, and / or the display / touchpad 128 (e.g., a liquid crystal display (LCD) display unit or organic light-emitting diode (OLED) display unit). The processor 118 may also output user data to the speaker / microphone 124, the keypad 126, and / or the display / touchpad 128. In addition, the processor 118 may access information from, and store data in, any type of suitable memory, such as the non-removable memory 130 and / or the removable memory 132. The non-removable memory 130 mayinclude random-access memory (RAM), read-only memory (ROM), a hard disk, or any other type of memory storage device. The removable memory 132 may include a subscriber identity module (SIM) card, a memory stick, a secure digital (SD) memory card, and the like. In other embodiments, the processor 118 may access information from, and store data in, memory that is not physically located on the WTRU 102, such as on a server or a home computer (not shown).

[0045] The processor 118 may receive power from the power source 134, and may be configured to distribute and / or control the power to the other components in the WTRU 102. The power source 134 may be any suitable device for powering the WTRU 102. For example, the power source 134 may include one or more dry cell batteries (e.g., nickel-cadmium (NiCd), nickel-zinc (NiZn), nickel metal hydride (NiMH), lithium-ion (Li-ion), etc.), solar cells, fuel cells, and the like.

[0046] The processor 118 may also be coupled to the GPS chipset 136, which may be configured to provide location information (e.g., longitude and latitude) regarding the current location of the WTRU 102. In addition to, or in lieu of, the information from the GPS chipset 136, the WTRU 102 may receive location information over the air interface 116 from a base station (e.g., base stations 114a, 114b) and / or determine its location based on the timing of the signals being received from two or more nearby base stations. It will be appreciated that the WTRU 102 may acquire location information by way of any suitable location- determination method while remaining consistent with an embodiment.

[0047] The processor 118 may further be coupled to other peripherals 138, which may include one or more software and / or hardware modules that provide additional features, functionality and / or wired or wireless connectivity. For example, the peripherals 138 may include an accelerometer, an e-compass, a satellite transceiver, a digital camera (for photographs and / or video), a universal serial bus (USB) port, a vibration device, a television transceiver, a hands free headset, a Bluetooth® module, a frequency modulated (FM) radio unit, a digital music player, a media player, a video game player module, an Internet browser, a Virtual Reality and / or Augmented Reality (VR / AR) device, an activity tracker, and the like. The peripherals 138 may include one or more sensors, the sensors may be one or more of a gyroscope, an accelerometer, a hall effect sensor, a magnetometer, an orientation sensor, a proximity sensor, a temperature sensor, a time sensor; a geolocation sensor; an altimeter, a light sensor, a touch sensor, a magnetometer, a barometer, a gesture sensor, a biometric sensor, and / or a humidity sensor.

[0048] The WTRU 102 may include a full duplex radio for which transmission and reception of some or all of the signals (e.g., associated with particular subframes for both the UL (e.g., for transmission) and downlink (e.g., for reception) may be concurrent and / or simultaneous. The full duplex radio may include aninterference management unit 139 to reduce and or substantially eliminate self-interference via either hardware (e.g., a choke) or signal processing via a processor (e.g., a separate processor (not shown) or via processor 118). In an embodiment, the WRTU 102 may include a half-duplex radio for which transmission and reception of some or all of the signals (e.g., associated with particular subframes for either the UL (e.g., for transmission) or the downlink (e.g., for reception)).

[0049] FIG.1C is a system diagram illustrating the RAN 104 and the CN 106 according to an embodiment. As noted above, the RAN 104 may employ an E-UTRA radio technology to communicate with the WTRUs 102a, 102b, 102c over the air interface 116. The RAN 104 may also be in communication with the CN 106.

[0050] The RAN 104 may include eNode-Bs 160a, 160b, 160c, though it will be appreciated that the RAN 104 may include any number of eNode-Bs while remaining consistent with an embodiment. The eNode-Bs 160a, 160b, 160c may each include one or more transceivers for communicating with the WTRUs 102a, 102b, 102c over the air interface 116. In one embodiment, the eNode-Bs 160a, 160b, 160c may implement MIMO technology. Thus, the eNode-B 160a, for example, may use multiple antennas to transmit wireless signals to, and / or receive wireless signals from, the WTRU 102a.

[0051] Each of the eNode-Bs 160a, 160b, 160c may be associated with a particular cell (not shown) and may be configured to handle radio resource management decisions, handover decisions, scheduling of users in the UL and / or DL, and the like. As shown in FIG.1C, the eNode-Bs 160a, 160b, 160c may communicate with one another over an X2 interface.

[0052] The CN 106 shown in FIG.1C may include a mobility management entity (MME) 162, a serving gateway (SGW) 164, and a packet data network (PDN) gateway (or PGW) 166. While each of the foregoing elements are depicted as part of the CN 106, it will be appreciated that any of these elements may be owned and / or operated by an entity other than the CN operator.

[0053] The MME 162 may be connected to each of the eNode-Bs 162a, 162b, 162c in the RAN 104 via an S1 interface and may serve as a control node. For example, the MME 162 may be responsible for authenticating users of the WTRUs 102a, 102b, 102c, bearer activation / deactivation, selecting a particular serving gateway during an initial attach of the WTRUs 102a, 102b, 102c, and the like. The MME 162 may provide a control plane function for switching between the RAN 104 and other RANs (not shown) that employ other radio technologies, such as GSM and / or WCDMA.

[0054] The SGW 164 may be connected to each of the eNode Bs 160a, 160b, 160c in the RAN 104 via the S1 interface. The SGW 164 may generally route and forward user data packets to / from the WTRUs102a, 102b, 102c. The SGW 164 may perform other functions, such as anchoring user planes during inter- eNode B handovers, triggering paging when DL data is available for the WTRUs 102a, 102b, 102c, managing and storing contexts of the WTRUs 102a, 102b, 102c, and the like.

[0055] The SGW 164 may be connected to the PGW 166, which may provide the WTRUs 102a, 102b, 102c with access to packet-switched networks, such as the Internet 110, to facilitate communications between the WTRUs 102a, 102b, 102c and IP-enabled devices.

[0056] The CN 106 may facilitate communications with other networks. For example, the CN 106 may provide the WTRUs 102a, 102b, 102c with access to circuit-switched networks, such as the PSTN 108, to facilitate communications between the WTRUs 102a, 102b, 102c and traditional land-line communications devices. For example, the CN 106 may include, or may communicate with, an IP gateway (e.g., an IP multimedia subsystem (IMS) server) that serves as an interface between the CN 106 and the PSTN 108. In addition, the CN 106 may provide the WTRUs 102a, 102b, 102c with access to the other networks 112, which may include other wired and / or wireless networks that are owned and / or operated by other service providers.

[0057] Although the WTRU is described in FIGS.1A-1D as a wireless terminal, it is contemplated that in certain representative embodiments that such a terminal may use (e.g., temporarily or permanently) wired communication interfaces with the communication network.

[0058] In representative embodiments, the other network 112 may be a WLAN.

[0059] A WLAN in Infrastructure Basic Service Set (BSS) mode may have an Access Point (AP) for the BSS and one or more stations (STAs) associated with the AP. The AP may have an access or an interface to a Distribution System (DS) or another type of wired / wireless network that carries traffic in to and / or out of the BSS. Traffic to STAs that originates from outside the BSS may arrive through the AP and may be delivered to the STAs. Traffic originating from STAs to destinations outside the BSS may be sent to the AP to be delivered to respective destinations. Traffic between STAs within the BSS may be sent through the AP, for example, where the source STA may send traffic to the AP and the AP may deliver the traffic to the destination STA. The traffic between STAs within a BSS may be considered and / or referred to as peer-to- peer traffic. The peer-to-peer traffic may be sent between (e.g., directly between) the source and destination STAs with a direct link setup (DLS). In certain representative embodiments, the DLS may use an 802.11e DLS or an 802.11z tunneled DLS (TDLS). A WLAN using an Independent BSS (IBSS) mode may not have an AP, and the STAs (e.g., all of the STAs) within or using the IBSS may communicatedirectly with each other. The IBSS mode of communication may sometimes be referred to herein as an “ad- hoc” mode of communication.

[0060] When using the 802.11ac infrastructure mode of operation or a similar mode of operations, the AP may transmit a beacon on a fixed channel, such as a primary channel. The primary channel may be a fixed width (e.g., 20 MHz wide bandwidth) or a dynamically set width via signaling. The primary channel may be the operating channel of the BSS and may be used by the STAs to establish a connection with the AP. In certain representative embodiments, Carrier Sense Multiple Access with Collision Avoidance (CSMA / CA) may be implemented, for example in in 802.11 systems. For CSMA / CA, the STAs (e.g., every STA), including the AP, may sense the primary channel. If the primary channel is sensed / detected and / or determined to be busy by a particular STA, the particular STA may back off. One STA (e.g., only one station) may transmit at any given time in a given BSS.

[0061] High Throughput (HT) STAs may use a 40 MHz wide channel for communication, for example, via a combination of the primary 20 MHz channel with an adjacent or nonadjacent 20 MHz channel to form a 40 MHz wide channel.

[0062] Very High Throughput (VHT) STAs may support 20MHz, 40 MHz, 80 MHz, and / or 160 MHz wide channels. The 40 MHz, and / or 80 MHz, channels may be formed by combining contiguous 20 MHz channels. A 160 MHz channel may be formed by combining 8 contiguous 20 MHz channels, or by combining two non-contiguous 80 MHz channels, which may be referred to as an 80+80 configuration. For the 80+80 configuration, the data, after channel encoding, may be passed through a segment parser that may divide the data into two streams. Inverse Fast Fourier Transform (IFFT) processing, and time domain processing, may be done on each stream separately. The streams may be mapped on to the two 80 MHz channels, and the data may be transmitted by a transmitting STA. At the receiver of the receiving STA, the above described operation for the 80+80 configuration may be reversed, and the combined data may be sent to the Medium Access Control (MAC).

[0063] Sub 1 GHz modes of operation are supported by 802.11af and 802.11ah. The channel operating bandwidths, and carriers, are reduced in 802.11af and 802.11ah relative to those used in 802.11n, and 802.11ac.802.11af supports 5 MHz, 10 MHz and 20 MHz bandwidths in the TV White Space (TVWS) spectrum, and 802.11ah supports 1 MHz, 2 MHz, 4 MHz, 8 MHz, and 16 MHz bandwidths using non-TVWS spectrum. According to a representative embodiment, 802.11ah may support Meter Type Control / Machine- Type Communications, such as MTC devices in a macro coverage area. MTC devices may have certain capabilities, for example, limited capabilities including support for (e.g., only support for) certain and / orlimited bandwidths. The MTC devices may include a battery with a battery life above a threshold (e.g., to maintain a very long battery life).

[0064] WLAN systems, which may support multiple channels, and channel bandwidths, such as 802.11n, 802.11ac, 802.11af, and 802.11ah, include a channel which may be designated as the primary channel. The primary channel may have a bandwidth equal to the largest common operating bandwidth supported by all STAs in the BSS. The bandwidth of the primary channel may be set and / or limited by a STA, from among all STAs in operating in a BSS, which supports the smallest bandwidth operating mode. In the example of 802.11ah, the primary channel may be 1 MHz wide for STAs (e.g., MTC type devices) that support (e.g., only support) a 1 MHz mode, even if the AP, and other STAs in the BSS support 2 MHz, 4 MHz, 8 MHz, 16 MHz, and / or other channel bandwidth operating modes. Carrier sensing and / or Network Allocation Vector (NAV) settings may depend on the status of the primary channel. If the primary channel is busy, for example, due to a STA (which supports only a 1 MHz operating mode), transmitting to the AP, the entire available frequency bands may be considered busy even though a majority of the frequency bands remains idle and may be available.

[0065] In the United States, the available frequency bands, which may be used by 802.11ah, are from 902 MHz to 928 MHz. In Korea, the available frequency bands are from 917.5 MHz to 923.5 MHz. In Japan, the available frequency bands are from 916.5 MHz to 927.5 MHz. The total bandwidth available for 802.11ah is 6 MHz to 26 MHz depending on the country code.

[0066] FIG.1D is a system diagram illustrating the RAN 113 and the CN 115 according to an embodiment. As noted above, the RAN 113 may employ an NR radio technology to communicate with the WTRUs 102a, 102b, 102c over the air interface 116. The RAN 113 may also be in communication with the CN 115.

[0067] The RAN 113 may include gNBs 180a, 180b, 180c, though it will be appreciated that the RAN 113 may include any number of gNBs while remaining consistent with an embodiment. The gNBs 180a, 180b, 180c may each include one or more transceivers for communicating with the WTRUs 102a, 102b, 102c over the air interface 116. In one embodiment, the gNBs 180a, 180b, 180c may implement MIMO technology. For example, gNBs 180a, 108b may utilize beamforming to transmit signals to and / or receive signals from the gNBs 180a, 180b, 180c. Thus, the gNB 180a, for example, may use multiple antennas to transmit wireless signals to, and / or receive wireless signals from, the WTRU 102a. In an embodiment, the gNBs 180a, 180b, 180c may implement carrier aggregation technology. For example, the gNB 180a may transmit multiple component carriers to the WTRU 102a (not shown). A subset of these component carriersmay be on unlicensed spectrum while the remaining component carriers may be on licensed spectrum. In an embodiment, the gNBs 180a, 180b, 180c may implement Coordinated Multi-Point (CoMP) technology. For example, WTRU 102a may receive coordinated transmissions from gNB 180a and gNB 180b (and / or gNB 180c).

[0068] The WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c using transmissions associated with a scalable numerology. For example, the OFDM symbol spacing and / or OFDM subcarrier spacing may vary for different transmissions, different cells, and / or different portions of the wireless transmission spectrum. The WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c using subframe or transmission time intervals (TTIs) of various or scalable lengths (e.g., containing varying number of OFDM symbols and / or lasting varying lengths of absolute time).

[0069] The gNBs 180a, 180b, 180c may be configured to communicate with the WTRUs 102a, 102b, 102c in a standalone configuration and / or a non-standalone configuration. In the standalone configuration, WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c without also accessing other RANs (e.g., such as eNode-Bs 160a, 160b, 160c). In the standalone configuration, WTRUs 102a, 102b, 102c may utilize one or more of gNBs 180a, 180b, 180c as a mobility anchor point. In the standalone configuration, WTRUs 102a, 102b, 102c may communicate with gNBs 180a, 180b, 180c using signals in an unlicensed band. In a non-standalone configuration WTRUs 102a, 102b, 102c may communicate with / connect to gNBs 180a, 180b, 180c while also communicating with / connecting to another RAN such as eNode-Bs 160a, 160b, 160c. For example, WTRUs 102a, 102b, 102c may implement DC principles to communicate with one or more gNBs 180a, 180b, 180c and one or more eNode-Bs 160a, 160b, 160c substantially simultaneously. In the non-standalone configuration, eNode-Bs 160a, 160b, 160c may serve as a mobility anchor for WTRUs 102a, 102b, 102c and gNBs 180a, 180b, 180c may provide additional coverage and / or throughput for servicing WTRUs 102a, 102b, 102c.

[0070] Each of the gNBs 180a, 180b, 180c may be associated with a particular cell (not shown) and may be configured to handle radio resource management decisions, handover decisions, scheduling of users in the UL and / or DL, support of network slicing, dual connectivity, interworking between NR and E-UTRA, routing of user plane data towards User Plane Function (UPF) 184a, 184b, routing of control plane information towards Access and Mobility Management Function (AMF) 182a, 182b and the like. As shown in FIG.1D, the gNBs 180a, 180b, 180c may communicate with one another over an Xn interface.

[0071] The CN 115 shown in FIG.1D may include at least one AMF 182a, 182b, at least one UPF 184a,184b, at least one Session Management Function (SMF) 183a, 183b, and possibly a Data Network(DN) 185a, 185b. While each of the foregoing elements are depicted as part of the CN 115, it will be appreciated that any of these elements may be owned and / or operated by an entity other than the CN operator.

[0072] The AMF 182a, 182b may be connected to one or more of the gNBs 180a, 180b, 180c in the RAN 113 via an N2 interface and may serve as a control node. For example, the AMF 182a, 182b may be responsible for authenticating users of the WTRUs 102a, 102b, 102c, support for network slicing (e.g., handling of different PDU sessions with different requirements), selecting a particular SMF 183a, 183b, management of the registration area, termination of NAS signaling, mobility management, and the like. Network slicing may be used by the AMF 182a, 182b in order to customize CN support for WTRUs 102a, 102b, 102c based on the types of services being utilized WTRUs 102a, 102b, 102c. For example, different network slices may be established for different use cases such as services relying on ultra-reliable low latency (URLLC) access, services relying on enhanced massive mobile broadband (eMBB) access, services for machine type communication (MTC) access, and / or the like. The AMF 162 may provide a control plane function for switching between the RAN 113 and other RANs (not shown) that employ other radio technologies, such as LTE, LTE-A, LTE-A Pro, and / or non-3GPP access technologies such as WiFi.

[0073] The SMF 183a, 183b may be connected to an AMF 182a, 182b in the CN 115 via an N11 interface. The SMF 183a, 183b may also be connected to a UPF 184a, 184b in the CN 115 via an N4 interface. The SMF 183a, 183b may select and control the UPF 184a, 184b and configure the routing of traffic through the UPF 184a, 184b. The SMF 183a, 183b may perform other functions, such as managing and allocating WTRU IP address, managing PDU sessions, controlling policy enforcement and QoS, providing downlink data notifications, and the like. A PDU session type may be IP-based, non-IP based, Ethernet-based, and the like.

[0074] The UPF 184a, 184b may be connected to one or more of the gNBs 180a, 180b, 180c in the RAN 113 via an N3 interface, which may provide the WTRUs 102a, 102b, 102c with access to packet-switched networks, such as the Internet 110, to facilitate communications between the WTRUs 102a, 102b, 102c and IP-enabled devices. The UPF 184, 184b may perform other functions, such as routing and forwarding packets, enforcing user plane policies, supporting multi-homed PDU sessions, handling user plane QoS, buffering downlink packets, providing mobility anchoring, and the like.

[0075] The CN 115 may facilitate communications with other networks. For example, the CN 115 may include, or may communicate with, an IP gateway (e.g., an IP multimedia subsystem (IMS) server) that serves as an interface between the CN 115 and the PSTN 108. In addition, the CN 115 may provide theWTRUs 102a, 102b, 102c with access to the other networks 112, which may include other wired and / or wireless networks that are owned and / or operated by other service providers. In one embodiment, the WTRUs 102a, 102b, 102c may be connected to a local Data Network (DN) 185a, 185b through the UPF 184a, 184b via the N3 interface to the UPF 184a, 184b and an N6 interface between the UPF 184a, 184b and the DN 185a, 185b.

[0076] In view of Figures 1A-1D, and the corresponding description of Figures 1A-1D, one or more, or all, of the functions described herein with regard to one or more of: WTRU 102a-d, Base Station 114a-b, eNode-B 160a-c, MME 162, SGW 164, PGW 166, gNB 180a-c, AMF 182a-ab, UPF 184a-b, SMF 183a-b, DN 185a-b, and / or any other device(s) described herein, may be performed by one or more emulation devices (not shown). The emulation devices may be one or more devices configured to emulate one or more, or all, of the functions described herein. For example, the emulation devices may be used to test other devices and / or to simulate network and / or WTRU functions.

[0077] The emulation devices may be designed to implement one or more tests of other devices in a lab environment and / or in an operator network environment. For example, the one or more emulation devices may perform the one or more, or all, functions while being fully or partially implemented and / or deployed as part of a wired and / or wireless communication network in order to test other devices within the communication network. The one or more emulation devices may perform the one or more, or all, functions while being temporarily implemented / deployed as part of a wired and / or wireless communication network. The emulation device may be directly coupled to another device for purposes of testing and / or may performing testing using over-the-air wireless communications.

[0078] The one or more emulation devices may perform the one or more, including all, functions while not being implemented / deployed as part of a wired and / or wireless communication network. For example, the emulation devices may be utilized in a testing scenario in a testing laboratory and / or a non-deployed (e.g., testing) wired and / or wireless communication network in order to implement testing of one or more components. The one or more emulation devices may be test equipment. Direct RF coupling and / or wireless communications via RF circuitry (e.g., which may include one or more antennas) may be used by the emulation devices to transmit and / or receive data.

[0079] A wireless transmit / receive unit (WTRU) may be configured to report the CSI through the uplink (UL) control channel on physical uplink control channel (PUCCH), and / or per the gNBs’ request on an UL physical uplink shared channel (PUSCH) grant. Depending on the configuration, for example, channel state information reference signal (CSI-RS) can cover the (e.g., full) bandwidth of a BandWidth Part (BWP)and / or just a fraction of the BWP. Within the CSI-RS bandwidth, CSI-RS can be configured in each physical resource block (PRB) and / or each (e.g., every other) PRB. In the time domain, CSI-RS resources can be configured either periodic, semi-persistent, and / or aperiodic. Semi-persistent CSI-RS may be similar to periodic CSI-RS, except that the resource can be (de)-activated by medium access control control entities (MAC CEs); and / or the WTRU may report related measurements (e.g., only) when the resource is activated. For Aperiodic CSI-RS, the WTRU may be triggered to report measured CSI-RS on PUSCH by request in a downlink control information (DCI). Periodic reports may be carried over the PUCCH, while semi-persistent reports can be carried (e.g., either) on PUCCH and / or PUSCH. The reported CSI may be used by the scheduler when allocating (e.g., optimal) resource blocks possibly based on channel’s time- frequency selectivity, determining precoding matrices, beams, transmission mode, and / or selecting (e.g., suitable) modulation and coding schemes (MCSs). WTRU CSI reports may be reliable, accurate, and / or timely to meet ultra-reliable and low latency communications (URLLC) service requirements.

[0080] A WTRU may be configured with a CSI measurement setting which may include one or more CSI reporting settings, resource settings, and / or a link between one or more CSI reporting settings and one or more resource settings.

[0081] Figure 2 shows an example of a configuration for CSI reporting settings, resource settings, and / or one or more links.

[0082] In a CSI measurement setting, one or more of the following configuration parameters may be provided. In a CSI measurement setting, one or more configuration parameters may include N≥1 CSI reporting settings, M≥1 resource settings, and / or a CSI measurement setting which links the N CSI reporting settings with the M resource settings. In a CSI measurement setting, one or more configuration parameters may include a CSI reporting setting that includes at least one of the following: Time-domain behavior - aperiodic and / or periodic and / or semi-persistent; Frequency-granularity, at least for precoding matrix indicator (PMI) and / or channel quality indicator (CQI); CSI reporting type (e.g., PMI, CQI, rank indicator (RI), CSI-RS RI (CRI), etc.); and / or, if a PMI is reported, PMI Type (e.g., Type I and / or II) and / or codebook configuration. In a CSI measurement setting, one or more configuration parameters may include a resource setting that includes at least one of the following: Time-domain behavior – aperiodic and / or periodic and / or semi-persistent; RS type (e.g., for channel measurement or interference measurement); and / or S≥1 resource set(s) and each resource set can include Ks resources. In a CSI measurement setting, one or more configuration parameters may include a CSI measurement setting that includes at least one of the following: one CSI reporting setting; One resource setting; and / or, for CQI, a referencetransmission scheme setting. In a CSI measurement setting, for CSI reporting for a component carrier, at one or more of the following frequency granularities may be supported: Wideband CSI; Partial band CSI; and / or Sub band CSI.

[0083] Artificial intelligence may be defined as the behavior exhibited by machines. Such behavior may, for example, mimic cognitive functions to sense, reason, adapt, and / or act. The terms Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), and / or deep neural networks (DNNs) may be used interchangeably herein. Methods described herein may be exemplified based on learning in wireless communication systems, for example. The methods may not be limited to such scenarios, systems, and / or services and / or may be applicable to each type of transmissions, communication systems, and / or services, etc.

[0084] Auto-encoders (AE) may be a specific class of deep neural networks (DNNs) that arise in context of un-supervised machine learning setting wherein the high-dimensional data is non-linearly transformed to a lower dimensional latent vector using the DNN based encoder. The lower dimensional latent vector may be used to re-produce the high-dimensional data using a non-linear decoder. The encoder may be represented as ^^^^^;^^^^ where x is the high-dimensional data and / or ^^^may represent the parameters of the encoder. The may be represented as ^^^^ ^^;^^ௗwhere z is the low-dimensional latentrepresentation and / or ^^ௗ may represent the parameters of the decoder. Using training data ^ ^^^,⋯ , ^^ே^the auto-encoder can be trained by solving the following optimization problem ே ^^^^௧^, ^^௧^ௗ ^ ൌ arg^min ^ || ^^^ െ ^^^^^^^^^; ^^^^; ^^ௗ^ ||ଶଶ . ^,^^^ୀ^

[0085] The above problem can be approximately solved using backpropagation algorithm. The trained encoder ^^^^^;^^^௧^^ can be used to compress the high-dimensional data and / or the trained decoder^^^^^; ^^ௗ௧^^can be used to decompress the latent representation.

[0086] Existing methods for training of 2 sided models (e.g., like auto-encoders) either require that the 2 models (e.g., encoder and / or decoder) are both present at the same compute entity for the training process and / or require high amounts of data transfer to accomplish the training. The first set of methods that rely on training at the same compute entity may be restrictive as the 2-sided models (e.g., like encoders and / or decoders) may be deployed at 2 different entities (e.g., WTRU and gNB, respectively). Ensuring that the model is available for training at a single entity may end up disclosing the proprietary model details (e.g., like architecture, type of layers, etc.) of (e.g., either) the WTRU and / or gNB. In the second set of training methods, for example, the 2 models may be hosted at their respective entities and / or large amount of datatransfer may required (e.g., either) for the gradient flow at each training step and / or for sharing the raw and / or encoded data pairs. Such large quantities of data transfer may be prohibitive for the training process and / or may make the training process (e.g., virtually) impractical. This training method may have the restriction that the training is done sequentially (e.g., first at gNB and then at WTRU or vice versa), which may not be easily extendible to a multi-vendor setting.

[0087] Embodiments described herein relate to a method for training 2-sided model (e.g., for CSI compression) where the WTRU trains the UE-side model (NW trains the NW-side model) independently to satisfy a preconfigured performance condition and / or the WTRU may achieve interoperability for inference by mapping the WTRU-side model output (e.g., encoded channel state information (CSI)) to a relative representation based on anchor vector configuration.

[0088] A WTRU may receive configuration information. The configuration information may indicate one or more (e.g., a plurality of datasets) and / or a set of anchor vectors. For example, the WTRU may be configured with a plurality of datasets. Each dataset may be associated with one or more of deployments, vendors, WTRU / NW additional condition(s), etc. For example, each dataset may be associated with Dataset ID. For example, each dataset may be associated with one or more applicable conditions. For example, each dataset may be configured as distribution of parameters (e.g., expressed as fundamental channel parameters (e.g., decomposed channel), expressed in terms of (e.g., 3GPP) channel model parameters 3D-urban macro (UMa) / clustered delay line (CDL) / etc., doppler, delay, etc.). For example, each anchor vector of the set of anchor vectors may be associated with at least one dataset of a plurality of datasets. The configuration information may include an indication of a plurality of datasets. The plurality of datasets may be associated with the set of anchor vectors. The configuration information may include a plurality of datasets, training criteria, and / or training information. Each dataset of the plurality of datasets may be associated with a respective dataset ID. The set of anchor vectors may be associated with an anchor set ID. The configuration information may include interoperability criteria. The interoperability criteria may be indicative of interoperability between a (e.g., first) model (e.g., at the WTRU) and another (e.g., second) model at the base station.

[0089] The WTRU may be configured with a set of anchor vectors (Ax1, Ax2, … AxN) for each dataset. For example, the WTRU may be configured with [A11, A12…. A1N] for dataset 1, [A21, A22…A2N] for dataset 2, and so on. For example, the WTRU may be configured with a set of anchor vectors that may be identified by a logical ID. For example, anchor vectors may be determined based on: Clustering-based selection- K- Means, etc; Convex hull selection- vertex component analysis (VCA); Random selection; and / orEigenvectors of original and / or encoded space. The WTRU may choose anchor vectors based on preconfigured condition(s). Different subsets of anchor vector within an anchor vector set may be configured with different priority.

[0090] The WTRU may be configured with performance metric(s) and / or requirement(s), for example for each dataset. Metric(s) for performance may include normalized means square error (NMSE), squared generalized cosine similarity (SGCS), etc. Metric(s) for performance may include key performance indicators (KPIs) (e.g., minimum performance thresholds).

[0091] The WTRU may be configured with rules and / or configuration and / or function(s) to map from latent representation to relative representation. For example, the WTRU may be configured with a similarity metric e.g., similarity metric S(.,.) – cosine similarity. For example, the WTRU may be configured with a distance metric - L1 / L2 / Lp norm. For example, the configuration information may include a similarity metric. The at least one (e.g., a plurality of) relative representation(s) may be determined based on the similarity metric. The similarity metric may be and / or include one or more of a cosine similarity and / or a distance metric.

[0092] The WTRU may be configured with dimensionality of relative representation and / or quantization of relative representation. The WTRU may be configured with latent dimension and / or quantization. The WTRU may be configured with training hyperparameters (e.g., random seed, optimizer, etc.). The WTRU may be configured with model hyperparameters, architecture / backbone, model complexity (e.g. number of parameters, FLOPS), number of layers etc. The WTRU may be configured with a training loss function.

[0093] The WTRU may train one or more AI / ML model(s) on one or more configured datasets using training configuration until the performance requirements are satisfied. It may be up to WTRU implementation to train one encoder for more than one dataset. For example, the WTRU may have one or more encoders E1, E2...EM. The WTRU may use each dataset of the plurality of datasets and / or the set of anchor vectors to train the encoder, for example, based on the training criteria and / or the training information such that the encoder (e.g., at the WTRU) is trained to be interoperable with at least one model at the base station. The WTRU may receive an indication that indicates one or more datasets. The encoder may be trained based on the indication of the one or more datasets.

[0094] The WTRU may report WTRU capability associated to one or more trained AI / ML model(s). For example, the WTRU may report support of Dataset ID(s) and / or Anchor vectors set. For example, the WTRU may send an indication that indicates a capability of the WTRU associated with one or more AI / ML models.

[0095] The WTRU may activate the model Eu. The WTRU may encodes anchor vectors using the encoder Euand / or may generate the encoded anchor vectors in latent space. (e.g., [Eu(Ax1), Eu(Ax2) …, Eu(AxN)]. For example, the WTRU may encode the set of anchor vectors using an encoder to generate encoded anchor vectors. The WTRU may encode channel state information (CSI) using the encoder to generate encoded CSI. The WTRU being configured to encode the set of anchor vectors may include the WTRU being configured to encode the subset of anchor vectors. The encoded anchor vectors may be generated based on the command and / or the interoperability criteria. The encoded anchor vectors may be generated based on the one or more dataset IDs and / or the one or more anchor set IDs. The encoded CSI may include one or more data points in a latent space. The WTRU may determine a distance and / or similarity between each data point in the latent space to each encoded anchor vector, for example, to determine the similarity between the encoded CSI and the encoded anchor vectors to determine the at least one relative representation. For example, the WTRU may determine one or more (e.g., any) similarity metric (e.g., cosine similarity, a distance metric) between each data point in the latent space to each encoded anchor vector, for example, to determine the similarity between the encoded CSI and the encoded anchor vectors to determine the at least one relative representation. The relative representation may be one or more dimensions. The dimensionality of the relative representation(s) may be based on the number of anchor vectors. For example, if there are N anchor vectors, the relative representation may be N dimensional. There can be more than one relative representation. For example, if there are different sets of anchor vectors, there may be one relative representation corresponding to each set of anchor vectors. For example, a first relative representation may correspond to a first set of anchor vectors. For example, a second relative representation may correspond to a second set of anchor vectors. The WTRU may activate the model Eu and / or may generate the encoded anchor vectors in latent space based on NW command (e.g., indicating dataset ID and / or anchor set ID broadcast in system information block (SIB) and / or in dedicated radio resource control (RRC) configuration). For example, the WTRU may receive a command. The command may include a radio resource configuration (RRC) message and / or a system information block (SIB) message that indicates one or more dataset IDs and / or one or more anchor set IDs. The WTRU may activate the model Eu and / or may generate the encoded anchor vectors in latent space based on NW configured condition (e.g., determine dataset ID based on measurement / applicable conditions, etc.).

[0096] For a reporting instance (e.g., CSI feedback), for example, the WTRU may derive encoded input Hi(e.g., Channel matrix) in latent space – Eu(Hi). For a reporting instance (e.g., CSI feedback), the WTRU may derives relative representation of input as a function of encoded latent representation of input,encoded latent representation of anchors and / or similarity metric (e.g., [S(Eu(Hi), Eu(Ax1)), S(Eu(Hi), Eu(Ax2)), …S(Eu(Hi), Eu(AxN))]). For example, the WTRU may determine at least one relative representation based on the encoded anchor vector(s) and / or the encoded CSI. The at least one relative representation may be indicative of similarity between the encoded anchor vector(s) and the encoded CSI. For example, one relative representation can mean that one representation was found using one or more (e.g., multiple) anchor vectors. A relative representation may be and / or include one or more dimensions. A relative representation may correspond to (e.g., only) one anchor vector. A relative representation can be multi- dimensional; each dimension may correspond to one of the anchor vectors (e.g., in the set of anchor vectors). For example, if there are one or more (e.g., multiple) sets of anchor vectors, there may be one relative representation for each set of anchor vectors.

[0097] The WTRU may apply preconfigured quantization to the relative representation. For example, the WTRU may apply prioritization rules to determine if one or more anchor vectors can be dropped for more compression. For example, the set of anchor vectors may be a subset of anchor vectors. The WTRU may determine the subset of anchor vectors, for example, by excluding at least one anchor vector included in the set of anchor vectors.

[0098] The WTRU may send an indication of the at least one relative representation, for example, to a base station. The WTRU may report the quantized relative representation (e.g., as compressed CSI feedback). For example, the indication of the at least one (e.g., a plurality of) relative representation(s) may be indicative of the quantized at least one (e.g., a plurality of) relative representation(s). The WTRU may report the anchor set ID and / or dataset ID. The WTRU may report the anchor subset ID. For example, the WTRU being configured to send the indication of the at least one (e.g., a plurality of) relative representation(s) may include the WTRU being configured to send an indication of each dataset ID and / or the anchor set ID and / or the anchor subset ID.

[0099] The gNB may perform gNB side training. The gNB may train autoencoder model E2 -> D2. Training may be completed to a pre-defined performance bound / threshold. The gNB may pass one or more sets of anchor vectors through the encoder E2 and / or may generate representation of anchor vectors in latent space (e.g., (E2(A1) …. E2(AN))). For each datapoint in the dataset, ^^^^^^, may build relativerepresentation [^^൫^^2^^^1^,^^2^^^^^^൯, … , ^^^^^2^^^^^^, ^^2^^^^^^^]. The gNB may train another (e.g., new)decoder to go^^^^ and / or to a space compatible with decoder D2.

[0100] Embodiments described herein may avoid drawbacks of the (e.g., current) training methods, where the encoder-decoder models may trained together, and / or large amounts training data and / or gradient data transfer has to be undertaken. The WTRU side and / or NW side models can be independently trained and / or interoperability between the 2 models can be ensured with (e.g., only) minimal amount of coordination / data exchange and / or data processing. There may be no constraints on the model architecture, loss functions and / or one or more (e.g., any) training parameters. For a 2-sided model (like an encoder-decoder), no prior knowledge (and / or coordination) about the model architecture, and / or other training parameters may be required. Minimal data exchange to ensure operability may (e.g., only) be required. Different WTRU vendors and / or NW vendors could have their own trained models and / or interoperability can be achieved (e.g., easily) on the fly by transmitting a small amount of data synchronizing data.

[0101] Embodiments described herein relate to configuration for interoperability based on relative representation. A WTRU may be configured with an AI / ML encoder to be used for the CSI generation part. The configured encoder may satisfy one or more interoperability conditions / parameters. This may be to ensure (e.g., proper) operation / pairing between the WTRU AI / ML encoder model and the NW AI / ML decoder model, with limited share of information about the models at both communication nodes. For example, the conditions / parameters may include the bottle-neck dimension size, quantization type, number of quantization bits, the input data type (e.g., CSI and / or possibly preprocessed version). For example, the interoperability criteria may include one or more of: a bottle-neck dimension size, a quantization type, a number of quantization bits, and / or an input data type. In examples, the WTRU encoder model may be explicitly configured (e.g., using model ID), wherein each model may be defined by one or more parameters (e.g., bottleneck size, quantization, input data type). In examples, the WTRU encoder model may be implicitly configured (e.g., as a function of the allocated uplink CSI payload, channel conditions, number of configured MIMO layers, number of transmit / receive antennas, configured BWP, etc.).

[0102] While CSI compression may be used herein as an example use-case, embodiments herein may not be limited to CSI compression and / or may be (e.g., broadly) considered applicable to each use-case in which two-sided model framework is used, with separate training of the encoder and / or decoder models and / or with limited share of information about the encoder and / or decoder models (e.g., architecture, training procedure, etc.) at the two nodes.

[0103] A WTRU may be configured with a plurality of datasets in support of AI / ML encoder model training while ensuring interoperability with a separately trained AI / ML decoder model. Each dataset may beassociated with at least one of the following. Each dataset may be associated with a dataset index / ID, where the dataset ID may be used to identify a pre-existing dataset along with one or more associated parameters (e.g., defined format, CSI type (full CSI, eigenvector-domain samples, beam-domain samples, delay-domain samples, etc.). Each dataset may be associated with dataset applicable conditions, where the conditions may include a set of possible CSI measurements range at which the dataset is generated. For example, the measurements type may include reference signal received power (RSRP), reference signal received quality (RSRQ), signal-to-noise ratio (SNR), received signal strength indicator (RSSI), rank indicator (RI), delay spread, Doppler shift, angle of arrival (AoA) range, and / or angel of departure (AoD) range. Each dataset may be associated with dataset configuration parameters. Dataset configuration parameters may include at least one of a configuration parameter (e.g., number of transmit / receive antenna ports, BWP, frequency region, number of beams in case of beam domain samples, and / or number of delay tabs in case of delay-domain samples). In examples, the dataset configuration may include at least one of a scenario (e.g., link-type (line of sight (Los) and / or non-LOS (NLOS)), indoor and / or outdoor, moving and / or non-moving WTRU and / or associated speed, cell-edge users, and / or cell-center users). The dataset may include one or more functions (e.g., preprocessing / post-processing, transformation applied to data, etc.). The dataset may include channel model parameters (e.g., channel model (UMa, 3D-urban micro (Umi), ray tracing, field data, etc.).

[0104] Configuration of anchor vectors may be associated with each dataset. A WTRU may be configured with a set of anchor vectors, where the anchor vectors are used as means for satisfying interoperability requirements between the separately trained WTRU-sided AI / ML encoder model and the NW-sided AI / ML decoder model. The anchor vectors may be denoted as (Ax1, Ax2, … AxN), where the first subscript ‘x’ may denote the dataset index, the second subscript index may denote the anchor sample index within the x-th dataset, and ^^௫may denote the number of anchor vectors associated with the x-th dataset. In examples, the set of anchor vectors may be (e.g., implicitly) configured based on the configured dataset. For example, the set of anchor vectors may be configured based on known samples from the configured dataset (e.g., first N samples and / or N may be configured). In examples, the set of anchor vectors may be configured separately from the dataset using a logical ID.

[0105] In examples, the WTRU may be configured with one or more parameters associated with the anchor vectors. For example, the parameters may include the number of anchor vectors, the type of anchor vectors (e.g., EV, beam-domain, full CSI, and the format of the anchor vectors, vector, matrix, tensor, etc.). The number of anchor vectors may vary across different datasets and / or across the different AI / ML encodermodels. The number of anchor vectors may be configured as a function of the bottleneck size of the AI / ML encoder model. For example, AI / ML encoder models with large bottleneck sizes may require larger number of anchor vectors relative to AI / ML encoder models with small bottleneck sizes. The anchor vectors may be in one of the following types: eigenvectors samples, beam-domain samples, full CSI samples, and / or delay- domain samples. The anchor vectors may be represented in vector format, matrix format, and / or tensor format.

[0106] The WTRU may be configured to determine the set of anchor vectors based on the configured dataset. The WTRU may be configured with and / or indicated to determine the number of anchor vectors. For example, if the WTRU is configured with the number of anchor vectors (e.g., N), the WTRU may use a method to determine the candidate anchor vectors from the dataset. In examples, the method can be clustering based selection using K means where the determined N candidate anchor vectors may represent the N closest samples within the dataset to the N identified K-means centroids; and / or (e.g., probably) the identified centroid can serve as the candidate anchor vectors. The selection may be performed in the latent domain (e.g., on the encoded representation of the entire samples) and / or on the original space. The same determined anchor vectors can be (e.g., easily) identified at the NW side since the dataset is shared across the two nodes. In example, it may (e.g., only) be the way of determination that is to be aligned across the two nodes. In examples, the WTRU may use convex hull based selection to determine the anchor vectors, and / or random selection operation with the same random seed being used at both WTRU and NW.

[0107] In examples, the WTRU may be configured to select the set of anchor vectors based on one or more preconfigured conditions. For example, the WTRU may be configured to select and / or update the set of anchors samples if the performance and / or metric associated with the relative representation using the current set of anchor vectors drops below a configured threshold. For example, the WTRU may select and / or update the anchor vectors satisfying one or more conditions (e.g., associated rank of each anchor sample needs to exceed a certain configured threshold). For example, the WTRU may select the set of anchor vectors satisfying a correlation criteria in the encoded space (e.g., pair-wise correlation in the encoded space across anchor vectors are below a configured threshold). In examples, the WTRU may be configured with a different subsets of anchor vectors within the same dataset, where the different subsets may be configured with different priority. For example, the WTRU may select and / or update the anchor vectors subset when a priority is changed (e.g., compared to previous transmission, and / or based on explicit request from the NW).

[0108] In examples, the WTRU may be configured with a performance metric and / or requirement associated with each dataset. For example, the metric may be configured as cosine similarity and / or variations thereof. In examples, the metric may be normalized mean square error. In examples, the metric may be Euclidean distance. In examples, the metric can be a projection of encoded test samples on the encoded anchor vectors and / or mat compare the resulting projection with a configured threshold. The WTRU may measure the configured metric between the encoded anchor vectors and each encoded test samples from the dataset, where the encoded test samples and / or the encoded anchor vectors are different.

[0109] The WTRU may be configured and / or indicated to perform a transformation and / or relative representation of the encoded samples and / or latent representation using the identified set of anchor vectors in the latent domain. The terms relative representation, transformation, and / or mapping may be used interchangeably herein to indicate a post processing operation on the latent representation of the WTRU AI / ML encoder model. For example, the WTRU may be configured with a relative representation function to perform the mapping from the latent representation to the relative representation. For example, the configured function may be expressed as ^^^^^;^^^, where ^^ may be the latent representation associated with the target CSI samples, ^^ may be a matrix including the set of identified anchor vectors inthe latent domain, (e.g., after passing through the AI / ML encoder mode), and / or ^^^. ^ may indicate theconfigured function used to transform z to a relative representation ^^^. For example, ^^^^^;^^^can be asimilarity metric and / or a linear transformation of z, e.g., ^^^ ൌ ^^^^. For example, ^^^^^; ^^^ can be a non-linear transformation (e.g., distance metric ( ^^ଶ െ ^^^^^^^^^^, where each element of ^^^ may be constructedas the ^^^^ ൌ ห|^^ െ ^^^|หଶ ଶ , where ^^^may be the ^^-th row of the matrix ^^. The WTRU may be (e.g., explicitly)function metric in a DCI field and / or MAC CE.

[0110] The WTRU may be configured to report one or more parameters associated with the relative representation. The parameters may include one or more of the following.

[0111] The parameters associated with the relative representation may include the dimension of the relative representation. For example, the WTRU may be (e.g., explicitly) configured with the dimension of the relative representation. In examples, the WTRU may be configured to use the number of anchor vectors as the default dimension of the relative representation. In examples, the WTRU may be configured with the dimension of the relative representation in an implicit way (e.g., as a function of the configured latent of the AI / ML encoder model, and / or as a function of the configured dataset, and / or as a function of the configured anchor vectors.

[0112] The parameters associated with the relative representation may include the quantization information associated with the relative representation. Quantization information associated with the relative representation may enable for example, (e.g., proper) reconstruction at the NW. In examples, the WTRU may be configured to determine and / or report the quantization information to the NW.

[0113] The parameters associated with the relative representation may include training and / or model parameters. The WTRU may be configured to report one or more training parameters. Training parameters may include, for example, random seed, optimizer, model complexity in terms of number of floating point operations per second (FLOPs) and / or number of model parameters, number of layers, model backbone (e.g., Transformer, CNN), and / or training loss function.

[0114] The parameters associated with the relative representation may include a performance metric threshold. For example, the metric threshold may be associated with the relative representation (e.g., the L2 / L1-norm of the relative representation vector prior quantization and / or maximum / minimum value per each element in the relative representation vector).

[0115] In examples, the WTRU may be configured with resources to indicate the determined relative representation and one or more of the associated parameters (e.g., performance metric associated with the relative representation, quantization info, etc.). The WTRU may be configured to indicate the relative representation and / or its associated parameters in one or more (e.g., two) parts. For example, a first part may include the parameters and / or a second part may include the relative representation. The first part may be configured to be transmitted aperiodically and / or semi-persistently (e.g., when a change occurs in quantization and / or performance metric associated with the relative representation exceeds a certain threshold). The second part may be configured to be transmitted periodically within each CSI feedback report. The WTRU may be configured to transmit the first part based on request from the network.

[0116] Embodiments described herein may include WTRU side model training, capability, and / or model selection and / or activation. One or more methods herein may be described in terms of encoder and / or decoder of an autoencoder architecture as examples; embodiments described herein may be applicable to each type of AI / ML model architecture. Herein the terms encoder and / or decoder may be used interchangeably with AI / ML model. One or more methods herein may be described in terms of gNB / NW; one or more methods described herein may be applicable to each type of transmission / reception node (e.g., WTRU).

[0117] The WTRU may be configured with one or more datasets DS1, DS2…DSN. The WTRU may be configured to train one or more models on the datasets such that the performance requirements are satisfied.

[0118] In examples, the WTRU may train an encoder decoder pair for each dataset. For example, the WTRU may train a first encoder (E1) decoder (D1) pair for a first dataset DS1. For example, the WTRU may train a second encoder (E2) decoder (D2) pair for a second dataset DS2 and so on leading to an encoder (EN) decoder (DN) pair for the dataset DSN.

[0119] In examples, the WTRU may train an encoder decoder pair for two or more datasets. For example, the WTRU may train a first encoder (E1) decoder (D1) pair for the datasets DS1, DS2...DSk. For example, the WTRU may train an encoder (Ex) decoder (Dx) pair for the datasets DSk+1, DSk+2 …DSm. And so, on leading to an encoder (EN) decoder (DN) pair for the datasets DSm+1, DSm+2…. DSN.

[0120] The WTRU may have a set of encoders (E1, E2….EU) where U >=1. In examples, the WTRU may indicate the capability associated with the one or more encoders (e.g., possibly with explicit identification methods). For example, the WTRU may indicate a logical identity associated with one or more encoders. The logical identity may be the model ID. The logical identity may be a functionality ID. In examples, the WTRU may indicate the capability associated with the one or more encoders with implicit identification methods. For example, the WTRU may indicate a logical identity of encoder as a function of datasets upon which the encoder is trained. For example, the WTRU may indicate the identity of datasets upon which the encoder(s) are trained. For example, the WTRU may indicate the support of scenarios and / or applicable conditions by indicating the availability of encoders trained on the datasets associated with those scenarios and / or applicable conditions.

[0121] In examples, the WTRU may be configured with anchor vector set(s). For example, each anchor vector set may be linked to a specific dataset. For example, each anchor vector set may include N anchor vectors. The value of N may be >=1. In examples, an anchor vector set may be identified by a logical ID. In examples, each anchor vector within the anchor vector set may be identified by a logical ID. The identity of anchor vector may be derived from the identity of associated anchor vector set. In examples, the identity of anchor vector set may be derived from identity of the dataset.

[0122] In examples, the WTRU may indicate the capability associated with one or more encoders implicitly via the identity of anchor vector set. For example, the WTRU may have encoder Ex trained on a dataset DSx associated with anchor vector set Ax1, Ax2…AxN. The anchor vector set may be associated with anidentity AIDx. The WTRU may indicate the support of Ex by including the identity AIDx in the capability report.

[0123] In examples, the WTRU may select and / or activate the encoder based on preconfigured conditions. Preconfigured conditions may be a function of dataset, anchor vector set and / or associated applicable conditions. For example, the WTRU may receive an applicability indication from the NW where the applicable indication may be associated with dataset and / or anchor vector set. For example, the WTRU may receive an applicable dataset ID and / or applicable anchor vector set ID for operation in the cell. For example, such applicable dataset ID and / or applicable anchor vector set ID may be indicated in the broadcast signaling like master information block (MIB) and / or system information block (SIB), and / or the like. For example, such applicable dataset ID and / or applicable anchor vector set ID may be indicated in the RRC signaling including RRC configuration, RRC reconfiguration, RRC setup, etc. For example, such applicable dataset ID and / or applicable anchor vector set ID may be indicated in the MAC CE signaling and / or L1 / DCI signaling. Upon receiving the applicable dataset ID and / or anchor vector set ID, for example, the WTRU may select the associated encoder and / or may activate the encoder for inference. In examples, the WTRU may be configured to measure and / or monitor channel measurements and / or may identify a dataset and / or anchor vector set that (e.g., closely) matches those measurements. The WTRU may select the encoder associated with the identified dataset and / or anchor vector set for inference.

[0124] In examples, the WTRU may select and / or activate an encoder (Eu) based on one or more embodiments described herein. In examples, the WTRU may generate the latent space representation for one or more anchor vectors in anchor vector set associated with the selected encoder (Eu). In examples, the WTRU may generate the latent space representations for each anchor vector in the anchor vector set. For a selected encoder Eu and / or associated anchor vector set [Ax1, Ax2…AxN], for example, the WTRU may pass the anchor vectors through the encoder individually and / or may obtain the representation of anchor vectors in latent space as [Eu(Ax1), Eu(Ax2)…Eu(AxN)]. If more than one anchor vector set is linked to the selected encoder, for example, the WTRU may generate latent space representations for each anchor vector in each those anchor vector sets. For example, if more than one anchor vector set is linked to the selected encoder, the WTRU may generate latent space representations for all anchor vectors in all those anchor vector sets.

[0125] In examples, the WTRU may determine the dataset associated with the selected encoder (Eu) and / or may generate the latent space representation for one or more anchor vectors in anchor vector set linked to the determined dataset. If more than one dataset is linked to the selected encoder, for example,the WTRU may generate latent space representations for the anchor vector sets associated with each of those datasets. For example, if more than one dataset is linked to the selected encoder, the WTRU may generate latent space representations for the anchor vector sets associated with all those datasets.

[0126] Embodiments are described herein for determination and / or reporting of relative representation.

[0127] FIG.2 depicts an example representation of the relative representation-based interoperable encoding-decoding framework 200. In the example setup, the channel matrix ^^^may be compressed using the WTRU side encoder ^^^. The compressed channel may be transformed to a relative representation,[^^൫^^^^^^^^, ^^^^^^^^൯, … , ^^^^^^^^^ே^, ^^^^^^^^^], using the encoded anchor vectors. The relative(e.g., with quantization) and / or may be fed back to themay to decompress the channel.

[0128] In examples, where a WTRU is configured for model interoperability through relative representation, the WTRU may derive a first latent representation by passing the input data through a 1st model (e.g. a first encoder model). For example, in a CSI compression setup using 2 sided AI / ML basedautoencoder models, the WTRU may pass a channel matrix ^^^ through an encoder model ^^^^. ^, toproduce a compressed representation and / or an encoded representation and / or a latent representation^^^^^^^^.

[0129] The WTRU may process the anchor vectors through the first model (e.g. a first encoder model) to obtain the encoded anchor vectors and / or the anchor vectors in the first latent space. For example, in a CSI compression setup using 2 sided AI / ML based autoencoder models, the WTRU may pass anchor vectors^^^ through an encoder model ^^^^. ^, to produce a compressed representation and / or an encodedrepresentation and / or a latent representation ^^^^^^^^.

[0130] The WTRU may utilize the encoded anchor vectors (e.g. ^^^^^^^^.) and / or the encoded data vectors (^^^^^^^^), for example, to obtain a relative representation.

[0131] In examples, the WTRU may utilize a similarity metric ^^^. , . ^ (e.g., for cosine similarity, L1 / L2 / Lpdistance, NMSE, mean squared error (MSE), etc.) and / or N encoded anchor vectors to obtain the Ndimensional relative representation [^^൫^^^^^^^^, ^^^^^^^^൯, … , ^^^^^^^^^ே^,^^^^^^^^^]. For example, in thecase of Lp distance,

[0132] The N^|^^^^^^^^^ െ ^^^^^^^^|^, |^^^^^^^ଶ^ െ ^^^^^^^^|^, … , |^^^^^^^ே^ െ ^^^^^^^^|^ ^

[0133] In examples where a cosine similarity based similarity metric is used, the relative representation vector can be computed as:^^^^^^^^்^^^^^^^^ ^^൫^^^^^^^^, ^^^^^^^^൯ ൌ ^^ ||^^^^^^^^| | ||^^^^^^^^||

[0134] The Nே^்^^^^^^^ ^ ||^^^^^^|^^^^^^^^||^, ^||^^^^^^ଶ^| | ||^^^^^^^^||^, … , |^ ^^^ ^ ^| | | ||^^^^^^ே^| | ||^^^^^^^^||

[0135] In examples, the relative representation may undergo additional post-processing steps. For example, the relative representation vector may be quantized. For example, the WTRU may quantize the at least one (e.g., a plurality of) relative representation(s). A scalar quantization strategy where each element is individually quantized may be adopted. For example, each relative representation of the at least one relative representation may be a scalar value and / or may be defined by the dimensionality of each anchor vector in the set of anchor vectors. The at least one relative representation may be indicative of compressed channel state information (CSI) feedback. Additionally or alternatively, a vector quantization strategy, where the entire vector and / or parts of the vector may be quantized together. In examples, the anchor vectors may have a priority and / or ranking associated with them. Each of the dimensions of the relative representation associated with the lower priority anchor vectors may be dropped. In examples, the dimensions associated with higher priority may be quantized with more bits and / or the dimensions associated with lower priority may be quantized with fewer bits.

[0136] After evaluating and / or processing the relative representation, for example, the WTRU may transmit and / or report the vector to the second entity which has the secondary interoperable model. For example, in the case of CSI feedback, the compressed and / or post processed relative channel may be transmitted by the WTRU to gNB for decoding and / or processing.

[0137] In examples, along with the relative representation, additional information pertaining to the identity of encoder model (e.g., in the case with multiple encoders) and / or corresponding identity of the anchor vectors may be sent along with the relative representation.

[0138] In examples, when the WTRU transmits the relative representation associated with (e.g., only) a subset of anchor vectors, the WTRU may, additionally or alternatively, include the identity of the included and / or dropped anchor vectors and / or may include the identity of the subset of the anchor vectors used.

[0139] Embodiments described here may relate to the training of NW models to handle relative representation.

[0140] Fig.3 depicts an example setup for training NW side decoder model to handle relative representations. In this example setup, a 2-stage training setup may be adopted, where in the firstprocedure, an encoder-decoder pair may be trained to reconstruct the input ^^^and / or the anchor vectors are mapped through the trained encoder model. For example, at 302, the WTRU encoder may encode each anchor vector (e.g., in the set of anchor vectors). For example, the WTRU encoder may encode anchor vector A(1) to obtain encoded anchor vector E1(A1). For example, the WTRU encoder may encode anchor vector A(2) to obtain encoded anchor vector E1(A2). For example, the WTRU may encode anchor vector A(N) to obtain encoded anchor vector E1(AN). The WTRU encoder may encode input ^^^to obtain encoded ^^^. The WTRU may encode channel state information (CSI) using the encoder toencoded CSI. The WTRU being configured to encode the set of anchor vectors may WTRU being configured to encode the subset of anchor vectors. The encoded anchor vectors may be generated based on the command and / or the interoperability criteria. The encoded anchor vectors may be generated based on the one or more dataset IDs and / or the one or more anchor set IDs. The encoded CSI may include one or more data points in a latent space. At 304, the WTRU may compute (e.g., as described herein) at least one relative representation based on the encoded CSI and / or the encoded anchor vector(s). At 306, the WTRU may perform a post-process (e.g., as described herein). At 308, a network node (e.g., associated with a base station) may receive an indication of the at least one relative representation. In the second procedure (e.g., at 310), a relative decoder may be trained to reconstruct the input ^^^from arelative representation vector [^^൫^^ଶ^^^^^, ^^ଶ^^^^^൯, … , ^^^^^ଶ^^^ே^,^^ଶ^^^^^^].

[0141] In examples, themodel (e.g. the decoder), may be hosted and / or operated at the gNB. For example, in the CSI feedback, to design the secondary interoperable model and / or the relative decoder, the gNB may employ a two-stage training process.

[0142] At the first training stage / procedure the gNB may train an encoder-decoder pair, ^^ଶ^. ^,^^ଶ^. ^. Theencoder decoder pair maybe trained to achieve the pre-defined performance thresholds on each of the defined metrics. For example, the encoder-decoder pair may achieve the pre-defined performance on the reconstruction normalized mean squared error (NMSE) and / or reconstruction cosine similarity. The anchorvectors ^^^, … , ^^ே, configured by the NW and / or defined by the WTRU may be passed to the encodermodel ^^ଶ^. ^ to generate the compressed or latent representation ^^ଶ^^^^^. Each data point ^^^ in thetraining dataset available at the gNB, may be passed through the ^^ଶ^. ^ Model to generate the compressedand / or latent representation ^^ଶ^^^^^. Wi . For each ^^^, the gNB may utilize the specified similarity metric^^^. , . ^ (e.g., for cosine similarity, L1 / L2 / Lp distance, NMSE, MSE, etc.) and / or N encoded anchor vectorsto obtain the N dimensional relative representation [^^൫^^ଶ^^^^^,^^ଶ^^^^^൯, … , ^^^^^ଶ^^^ே^, ^^ଶ^^^^^^]. Forexample, in the case of Lp distance, ^^൫^^ଶ^^^^^, ^^ଶ^^^^^൯ ൌ |^^^ଶ^^^^^ െ ^^ଶ^^^^^|^ And the Ndimensional relative vector is given as: ^|^^^ଶ^^^^^ െ ^^ଶ^^^^^|^, |^^^ଶ^^^ଶ^ െ ^^ଶ^^^^^|^, … , |^^^ଶ^^^ே^ െ ^^ଶ^^^^^|^ ^

[0143] At decoder model)the model may be trained with the dataset consisting of input-output pairs where the input may be therelative representation [^^൫^^ଶ^^^^^,^^ଶ^^^^^൯, … , ^^^^^ଶ^^^ே^,^^ଶ^^^^^^] and / or the corresponding outputlabel may be the channel matrix to achieve the pre-definedperformance on the predefined mean squared error (NMSE)and / or reconstruction cosine similarity). In the secondary model may be trained for indirect interoperability. The model may be trained with the dataset including input-output pairs, where the inputmay be the relative representation [^^൫^^ଶ^^^^^, ^^ଶ^^^^^൯, … , ^^^^^ଶ^^^ே^,^^ଶ^^^^^^] and / or thecorresponding output label may with decoder^^ଶ^. ^ to go from the relative representation to the channel matrix and / or CSI ^^^.

[0144] Embodiments described herein may relate to the configuration of report. Inexamples, where interoperability is to be achieved between 2-sided AI ML models, an interoperability report may be shared from one entity to the other. In a setup where a first model is hosted at the WTRU and a second is hosted at NW, for example, the interoperability report may be sent from the WTRU to the NW and / or from the NW to the WTRU.

[0145] The interoperability report may include information required to enable the alignment of the 2 models and / or their underlying latent spaces. In examples, the interoperability report may be sent from the WTRU to the NW, and / or may contain information about: the dataset used for training, the size of the anchor vector set, and / or identify of anchor vectors, the encoded anchor vectors, latent dimension of the encoder model, additional training related parameters, and / or model performance and / or KPI. In examples, the dataset used for training may be included in the interoperability report. This may include the parameters used for generating the dataset and / or (e.g., more explicitly) the dataset ID used for training. In examples, the size of the anchor vector set, N, and / or identity of anchor vectors may be included in the interoperability report. In examples, the encoded anchor vectors may be included in the report. This may include the anchor vectors mapped in the latent space having been passed through the encoder model. In examples, the latent dimension of the encoder model may be included in the interoperability report. One or more additional training related parameters, like training hyperparameters (e.g., random seed, optimizer etc.),model hyperparameters, architecture / backbone, encoder complexity information (e.g. number of parameters, FLOPS), number layers, training loss function, etc. may be included in the interoperability report. The interoperability report may include model performance and / or KPI. Model performance and / or KPI may include about the metrics used to gauge model performance and / or or the performance values achieved by the encoder.

[0146] Embodiments described herein may include triggers and / or transmission of interoperability report. One or more embodiments herein may enable the transmitter and / or receiver to train the models independently and / or use those models for joint inference. The WTRU may transmit interoperability report. The interoperability report may include one or more elements that characterize the latent space of the AI / ML model at the WTRU. In examples, the interoperability report may enable the receiver to determine a transformation from a first latent space to a second latent space. For example, the first latent space may be associated with the AI / ML model at the transmitter. For example, the second latent space may be associated with the AI / ML model at the receiver.

[0147] The WTRU may be preconfigured with plurality of datasets. Each dataset may be identified with a dataset ID. The WTRU may train one or more AI / ML models. Each AI / ML model may be associated with one or more datasets. The WTRU may include Dataset ID(s) associated with the selected AI / ML model (e.g., encoder) in the interoperability report. Such AI / ML model may be selected for inference. Such AI / ML model may be the model for which transmission of interoperability report is requested. The WTRU may be preconfigured with plurality of datasets and / or each dataset may be associated with anchor vector set. Each anchor vector set may be associated with an anchor vector set ID. The WTRU may train one or more AI / ML models where the latent space of each AI / ML model may be characterized by anchor vector set. The WTRU may include anchor vector set ID associated with the selected AI / ML model (e.g., encoder) in the interoperability report. Such AI / ML model may be selected for inference. Such AI / ML model may be the model for which transmission of interoperability report is requested.

[0148] The WTRU may be configured to transmit the latent representation of one or more anchor vectors within the anchor vector set associated with the selected AI / ML model. In examples, the WTRU may be configured with the size of anchor vector set. In examples, the WTRU may be configured to include a subset of the anchor vectors within the anchor vector set in the interoperability report. The WTRU may indicate the identity of anchor vectors in the interoperability report.

[0149] In examples, the WTRU may transmit the interoperability report in uplink control information (UCI), MAC CE and / or RRC signaling. In examples, the WTRU may transmit interoperability report duringcapability transfer. For example, the WTRU may include the interoperability report for each of the AI / ML models for which support is indicated in the capability transfer. For example, the WTRU may include the include the interoperability report for the AI / ML models for which the capability is requested. In examples, the WTRU may transmit the interoperability report during model identification. For example, the WTRU may include the interoperability report for each of the AI / ML models to be identified.

[0150] In examples, the WTRU may transmit the interoperability report when model selection is triggered. The WTRU may include the interoperability report for the selected model. In examples, the WTRU may transmit the interoperability report during model monitoring. For example, the WTRU may transmit the interoperability report when the model performance is below a threshold. For example, the WTRU may be configured to transmit the interoperability report for current active AI / ML model. In examples, the WTRU may be configured to transmit the interoperability report for one or more inactive AI / ML model.

[0151] In examples, the WTRU may transmit the interoperability report when model switching is triggered. For example, the WTRU may transmit the interoperability report associated with a second model when switching from a first model to a second model. For example, the WTRU may transmit the interoperability report when the interoperability report for the second model is not transmitted within the last T ms – where the value of T may be configured for a WTRU.

[0152] In examples, the WTRU may transmit the interoperability report when model transfer is triggered. For example, the WTRU may transmit interoperability report for a model successfully download and / or verified for execution by the WTRU. In examples, the WTRU may transmit the interoperability report upon successful completion of model update and / or fine-tuning. In examples, the WTRU may transmit the interoperability report upon model activation. In examples, the WTRU may be configured to transmit the interoperability report upon explicit request from NW. In examples, the WTRU may transmit the interoperability report upon execution of a mobility event. For example, such mobility event may be one or more of: cell selection, reselection to a different serving cell, handover, etc. In examples, the WTRU may transmit the interoperability report upon recovery from radio link failure event and / or beam failure event. For example, the WTRU may transmit the interoperability in RRC re-establishment, beam recovery, etc.

[0153] Embodiments described herein may include the determination of transformation function based on an interoperability report.

[0154] FIG.4 depicts an example setup for training network (NW) side decoder model to handle relative representations 400. In examples, the secondary model in a 2-sided model setup (e.g. the decoder), may be hosted and / or operated at the gNB and / or an additional transformation function may be utilized to makethe first model (e.g. encoder) and / or the second model interoperable. For example, in the CSI feedback, an encoder model at the WTRU and / or the decoder model at the NW / gNB may be made interoperable using a transformation function to map the latent vector (and / or the compressed vector and / or the encoded vector received as output of the encoder) through a function such that it is interoperable with NW side decoder model. In examples, the gNB may receive an interoperability report from the WTRU enabling the training of the secondary AI / ML model at the gNB and / or enabling the gNB to build a transformation function for interoperability. In examples, the gNB may perform the training of the secondary model (e.g. decoder). In a CSI setup, for example, the gNB may train the autoencoder model with encoder and decoder pairs given as E2 and D2. The training may be performed to reach a pre-defined performance bound and / or threshold. In examples, the gNB may receive the anchor vectors, A1, A2…AN, from the WTRU as part of the interoperability report and / or may calculate [E2(A1), E2(A2) …, E2(AN)] by passing the anchor vectors through the encoder model E2.Additionally or alternatively, as part of the interoperability report, the gNB may receive the encoded anchor vectors [E1(A1), E1(A2) …, E1(AN)] using the primary model at the WTRU.

[0155] In examples, the gNB may calculates a transformation function (e.g., linear, polynomial) transformation T(.) such that E2(Ai) = T(E1(Ai)) using one or more (e.g., all or a subset of) anchor vectors A1, A2…AN.. In examples, the function T(.) may be linear and may be obtained as a least squares fit between E1(Ai) and E2(Ai). In examples, the transformation function T(.) may be non-linear and / or may be modeled as a polynomial and / or a neural network and / or learnt using gradient descent.

[0156] In examples, during the operation phase, the gNB may receive the encoded data vector E1(Hi), from the WTRU and / or may determine equivalent representation compatible with NW decoder, by using the transformation T(.). For example, T(E1(Hi) may be passed to the decoder model D2(.) for decoding.

[0157] Methods described herein for the training and / or determination of a decoder model at a gNB and / or NW may be performed by a WTRU (e.g., a second WTRU) for the training and / or determination of a decoder model (e.g., for a decoder model located at the second WTRU).

[0158] Embodiments described herein may include interoperability through relative representations. Embodiments described herein relate to a method for training 2-sided model (e.g., for CSI compression) where the WTRU trains the UE-side model (NW trains the NW-side model) independently to satisfy a preconfigured performance condition and / or the WTRU may achieve interoperability for inference by mapping the WTRU-side model output (e.g., encoded channel state information (CSI)) to a relative representation based on anchor vector configuration.

[0159] A base station may send configuration information. The configuration information may indicate a set of anchor vectors. The base station may receive an indication of the at least one relative representation, for example, from a WTRU. The base station may receive an indication of a capability of the WTRU associated with one or more AI / ML models, for example, for interoperability of one or more models at the WTRU and one or more models at the base station.

[0160] A WTRU may receive configuration information. The configuration information may indicate one or more (e.g., a plurality of datasets) and / or a set of anchor vectors. For example, the WTRU may be configured with a plurality of datasets. Each dataset may be associated with one or more of deployments, vendors, WTRU / NW additional condition(s), etc. For example, each dataset may be associated with Dataset ID. For example, each dataset may be associated with one or more applicable conditions. For example, each dataset may be configured as distribution of parameters (e.g., expressed as fundamental channel parameters (e.g., decomposed channel), expressed in terms of (e.g., 3GPP) channel model parameters 3D-urban macro (Uma) / clustered delay line (CDL) / etc., doppler, delay, etc.). For example, each anchor vector of the set of anchor vectors may be associated with at least one dataset of a plurality of datasets. The configuration information may include an indication of a plurality of datasets. The plurality of datasets may be associated with the set of anchor vectors. The configuration information may include a plurality of datasets, training criteria, and / or training information. Each dataset of the plurality of datasets may be associated with a respective dataset ID. The set of anchor vectors may be associated with an anchor set ID. The configuration information may include interoperability criteria. The interoperability criteria may be indicative of interoperability between a (e.g., first) model (e.g., at the WTRU) and another (e.g., second) model at the base station.

[0161] The WTRU may be configured with a set of anchor vectors (Ax1, Ax2, … AxN) for each dataset. For example, the WTRU may be configured with [A11, A12…. A1N] for dataset 1, [A21, A22…A2N] for dataset 2, and so on. For example, the WTRU may be configured with a set of anchor vectors that may be identified by a logical ID. For example, anchor vectors may be determined based on: Clustering-based selection- K- Means, etc.; Convex hull selection- vertex component analysis (VCA); Random selection; and / or Eigenvectors of original and / or encoded space. The WTRU may choose anchor vectors based on preconfigured condition(s). Different subsets of anchor vector within an anchor vector set may be configured with different priority.

[0162] The WTRU may be configured with performance metric(s) and / or requirement(s), for example for each dataset. Metric(s) for performance may include normalized means square error (NMSE), squaredgeneralized cosine similarity (SGCS), etc. Metric(s) for performance may include key performance indicators (KPIs) (e.g., minimum performance thresholds).

[0163] The WTRU may be configured with rules and / or configuration and / or function(s) to map from latent representation to relative representation. For example, the WTRU may be configured with a similarity metric e.g., similarity metric S(.,.) – cosine similarity. For example, the WTRU may be configured with a distance metric - L1 / L2 / Lp norm. For example, the configuration information may include a similarity metric. The at least one (e.g., a plurality of) relative representation(s) may be determined based on the similarity metric. The similarity metric may be and / or include one or more of a cosine similarity and / or a distance metric.

[0164] The WTRU may be configured with dimensionality of relative representation and / or quantization of relative representation. The WTRU may be configured with latent dimension and / or quantization. The WTRU may be configured with training hyperparameters (e.g., random seed, optimizer, etc.). The WTRU may be configured with model hyperparameters, architecture / backbone, model complexity (e.g. number of parameters, FLOPS), number of layers etc. The WTRU may be configured with a training loss function.

[0165] The WTRU may train one or more AI / ML model(s) on one or more configured datasets using training configuration until the performance requirements are satisfied. It may be up to WTRU implementation to train one encoder for more than one dataset. For example, the WTRU may have one or more encoders E1, E2...EM. The WTRU may use each dataset of the plurality of datasets and / or the set of anchor vectors to train the encoder, for example, based on the training criteria and / or the training information such that the encoder (e.g., at the WTRU) is trained to be interoperable with at least one model at the base station. The WTRU may receive an indication that indicates one or more datasets. The encoder may be trained based on the indication of the one or more datasets.

[0166] The WTRU may report WTRU capability associated to one or more trained AI / ML model(s). For example, the WTRU may report support of Dataset ID(s) and / or Anchor vectors set. For example, the WTRU may send an indication that indicates a capability of the WTRU associated with one or more AI / ML models.

[0167] The WTRU may activate the model Eu. The WTRU may encode anchor vectors using the encoder Euand / or may generate the encoded anchor vectors in latent space. (e.g., [Eu(Ax1), Eu(Ax2) …, Eu(AxN)]. For example, the WTRU may encode the set of anchor vectors using an encoder to generate encoded anchor vectors. The WTRU may encode channel state information (CSI) using the encoder to generate encoded CSI. The WTRU being configured to encode the set of anchor vectors may include the WTRUbeing configured to encode the subset of anchor vectors. The encoded anchor vectors may be generated based on the command and / or the interoperability criteria. The encoded anchor vectors may be generated based on the one or more dataset IDs and / or the one or more anchor set IDs. The encoded CSI may include one or more data points in a latent space. The WTRU may determine a distance and / or similarity between each data point in the latent space to each encoded anchor vector, for example, to determine the similarity between the encoded CSI and the encoded anchor vectors to determine the at least one relative representation. For example, the WTRU may determine one or more (e.g., any) similarity metric (e.g., cosine similarity, a distance metric) between each data point in the latent space to each encoded anchor vector, for example, to determine the similarity between the encoded CSI and the encoded anchor vectors to determine the at least one relative representation. The relative representation may be one or more dimensions. The dimensionality of the relative representation(s) may be based on the number of anchor vectors. For example, if there are N anchor vectors, the relative representation may be N dimensional. There can be more than one relative representation. For example, if there are different sets of anchor vectors, there may be one relative representation corresponding to each set of anchor vectors. For example, a first relative representation may correspond to a first set of anchor vectors. For example, a second relative representation may correspond to a second set of anchor vectors. The WTRU may activate the model Euand / or may generate the encoded anchor vectors in latent space based on NW command (e.g., indicating dataset ID and / or anchor set ID broadcast in system information block (SIB) and / or in dedicated radio resource control (RRC) configuration). For example, the WTRU may receive a command. The command may include a radio resource configuration (RRC) message and / or a system information block (SIB) message that indicates one or more dataset IDs and / or one or more anchor set IDs. The WTRU may activate the model Eu and / or may generate the encoded anchor vectors in latent space based on NW configured condition (e.g., determine dataset ID based on measurement / applicable conditions, etc.).

[0168] For a reporting instance (e.g., CSI feedback), for example, the WTRU may derive encoded input Hi (e.g., Channel matrix) in latent space – Eu(Hi). For a reporting instance (e.g., CSI feedback), the WTRU may derives relative representation of input as a function of encoded latent representation of input, encoded latent representation of anchors and / or similarity metric (e.g., [S(Eu(Hi), Eu(Ax1)), S(Eu(Hi), Eu(Ax2)), …S(Eu(Hi), Eu(AxN))]). For example, the WTRU may determine at least one relative representation based on the encoded anchor vector(s) and / or the encoded CSI. The at least one relative representation may be indicative of similarity between the encoded anchor vector(s) and the encoded CSI. For example, one relative representation can mean that one representation was found using one or more (e.g., multiple)anchor vectors. A relative representation may be and / or include one or more dimensions. A relative representation may correspond to (e.g., only) one anchor vector. A relative representation can be multi- dimensional; each dimension may correspond to one of the anchor vectors (e.g., in the set of anchor vectors). For example, if there are one or more (e.g., multiple) sets of anchor vectors, there may be one relative representation for each set of anchor vectors.

[0169] The WTRU may apply preconfigured quantization to the relative representation. For example, the WTRU may apply prioritization rules to determine if one or more anchor vectors can be dropped for more compression. For example, the set of anchor vectors may be a subset of anchor vectors. The WTRU may determine the subset of anchor vectors, for example, by excluding at least one anchor vector included in the set of anchor vectors.

[0170] The WTRU may send an indication of the at least one relative representation, for example, to a base station. The WTRU may report the quantized relative representation (e.g., as compressed CSI feedback). For example, the indication of the at least one (e.g., a plurality of) relative representation(s) may be indicative of the quantized at least one (e.g., a plurality of) relative representation(s). The WTRU may report the anchor set ID and / or dataset ID. The WTRU may report the anchor subset ID. For example, the WTRU being configured to send the indication of the at least one (e.g., a plurality of) relative representation(s) may include the WTRU being configured to send an indication of each dataset ID and / or the anchor set ID and / or the anchor subset ID.

[0171] The gNB may perform gNB side training. The gNB may train autoencoder model E2 -> D2. Training may be completed to a pre-defined performance bound / threshold. The gNB may pass one or more sets of anchor vectors through the encoder E2 and / or may generate representation of anchor vectors in latent space (e.g., (E2(A1) …. E2(AN))). For each datapoint in the dataset,^^^^^^, may build relativerepresentation [^^൫^^2^^^1^,^^2^^^^^^൯, … , ^^^^^2^^^^^^, ^^2^^^^^^^]. The gNB may train another (e.g., new)decoder to go from modified representation space to the data space ^^^^ and / or to a space compatible with decoder D2.

[0172] Embodiments described herein may avoid drawbacks of the (e.g., current) training methods, where the encoder-decoder models may be trained together, and / or large amounts training data and / or gradient data transfer has to be undertaken. The WTRU side and / or NW side models can be independently trained and / or interoperability between the 2 models can be ensured with (e.g., only) minimal amount of coordination / data exchange and / or data processing. There may be no constraints on the model architecture, loss functions and / or one or more (e.g., any) training parameters. For a 2-sided model (like anencoder-decoder), no prior knowledge (and / or coordination) about the model architecture, and / or other training parameters may be required. Minimal data exchange to ensure operability may (e.g., only) be required. Different WTRU vendors and / or NW vendors could have their own trained models and / or interoperability can be achieved (e.g., easily) on the fly by transmitting a small amount of data synchronizing data.

[0173] Embodiments described herein may include interoperability through transformation. The WTRU- side model (NW may train the NW-side model) independently to satisfy a preconfigured performance condition and / or the WTRU may achieve interoperability for inference by transmission of interoperability report characterizing the WTRU-side latent space – where the characterization may enable a transformation to NW-side latent space.

[0174] The WTRU may be configured with a plurality of datasets. Each dataset may be associated with one or more of deployments, vendors, WTRU / NW additional condition(s), etc. For example, each dataset may be associated with Dataset ID. For example, each dataset may be associated with one or more applicable conditions. For example, each dataset may be configured as distribution of parameters (e.g., expressed as fundamental channel parameters (e.g., decomposed channel), expressed in terms of (e.g., 3GPP) channel model parameters 3D-urban macro (UMa) / clustered delay line (CDL) / etc., doppler, delay, etc.)

[0175] The WTRU may be configured with a set of anchor vectors (Ax1, Ax2, … AxN) for each dataset. For example, the WTRU may be configured with [A11, A12…. A1N] for dataset 1, [A21, A22…A2N] for dataset 2, and so on. For example, the WTRU may be configured with a set of anchor vectors that may be identified by a logical ID. For example, anchor vectors may be determined based on: Clustering-based selection- K- Means, etc; Convex hull selection- vertex component analysis (VCA); Random selection; and / or Eigenvectors of original and / or encoded space. The WTRU may choose anchor vectors based on preconfigured condition(s). Different subsets of anchor vector within an anchor vector set may be configured with different priority.

[0176] The WTRU may be configured with performance metric(s) and / or requirement(s), for example for each dataset. Metric(s) for performance may include normalized means square error (NMSE), squared generalized cosine similarity (SGCS), etc. Metric(s) for performance may include key performance indicators (KPIs) (e.g., minimum performance thresholds).

[0177] The WTRU may be configured with a size of anchor vector set (e.g., N).

[0178] The WTRU may be configured with dimensionality of relative representation and / or quantization of relative representation. The WTRU may be configured with latent dimension and / or quantization. The WTRU may be configured with training hyperparameters (e.g., random seed, optimizer, etc.). The WTRU may be configured with model hyperparameters, architecture / backbone, model complexity (e.g. number of parameters, FLOPS), number of layers etc. The WTRU may be configured with a training loss function.

[0179] The WTRU may be configured with interoperability reporting configuration. Interoperability reporting configuration may include reporting resources, triggering, etc. Triggers may include model activation, model identification, model switching, WTRU capability report, model fine-tuning, model update, explicit NW trigger, etc.

[0180] The WTRU may train one or more AI / ML model(s) on one or more configured datasets using training configuration until the performance requirements are satisfied. It may be up to WTRU implementation to train one encoder for more than one dataset. For example, the WTRU may have one or more encoders E1, E2...EM.

[0181] The WTRU may report WTRU capability associated to one or more trained AI / ML model(s). For example, the WTRU may report support of Dataset ID(s) and / or Anchor vectors set.

[0182] The WTRU may activate the model Eu. The WTRU may encode anchor vectors using the encoder Euand / or may generate the encoded anchor vectors in latent space. (e.g., [Eu(Ax1), Eu(Ax2) …, Eu(AxN)]. The WTRU may activate the model Eu and / or may generate the encoded anchor vectors in latent space based on NW command (e.g., indicating dataset ID and / or anchor set ID broadcast in system information block (SIB) and / or in dedicated radio resource control (RRC) configuration). The WTRU may activate the model Eu and / or may generate the encoded anchor vectors in latent space based on NW configured condition (e.g., determine dataset ID based on measurement / applicable conditions, etc.). The WTRU may activate the model Eubased on NW command (e.g., indicating dataset ID and / or anchor set ID broadcast in system information block (SIB) and / or in dedicated radio resource control (RRC) configuration). The WTRU may activate the model Eu based on NW configured condition (e.g., determine dataset ID based on measurement / applicable conditions, etc.). If not (e.g., already) configured, the WTRU may select N anchor vectors (e.g., based on implementation and / or based on one or more rules). The NW can configure the WTRU to trigger interoperability report. The WTRU may pass a set of anchor vectors through the encoded Euand / or may generate the encoded anchor vectors in latent space (e.g. [Eu(Ax1), Eu(Ax2) …, Eu(AxN)).

[0183] Upon preconfigured trigger conditions, for example, the WTRU may transmit one or more of the following on the configured interoperability reporting resources: Dataset ID (e.g., if preconfigured) and / or anchor set ID; and / or the encoded anchor vector(s).

[0184] For a reporting instance (e.g., CSI feedback), the WTRU may derive encoded input Hi (e.g., Channel matrix) in latent space – Eu(Hi).

[0185] The WTRU may report the quantized latent representation (e.g., as compressed CSI feedback).

[0186] The gNB performs gNB side training. The gNB may train autoencoder model E2 -> D2. Training may be completed to a pre-defined performance bound and / or threshold. The gNB may receive and / or may determine A1, A2...ANand / or may calculate [E2(Ax1), E2(Ax2) …, E2(AxN)] . The gNB may receive [Eu(Ax1), Eu(Ax2) …, Eu(AxN)] from a WTRU. The gNB may calculate (e.g., linear) transformation function T(.) such that E2(Ai) = T( E1(Ai) ). The gNB may receive Eu(Hi), and / or may determine equivalent representation for NW decoder, E2(Hi), using T(.). The may gNB input E2(Hi) to D2.

[0187] Embodiments described herein may avoid drawbacks of the (e.g., current) training methods, where the encoder-decoder models may trained together, and / or large amounts training data and / or gradient data transfer has to be undertaken. The WTRU side and / or NW side models can be independently trained and / or interoperability between the 2 models can be ensured with (e.g., only) minimal amount of coordination / data exchange and / or data processing. There may be no constraints on the model architecture, loss functions and / or one or more (e.g., any) training parameters. For a 2-sided model (like an encoder-decoder), no prior knowledge (and / or coordination) about the model architecture, and / or other training parameters may be required. Minimal data exchange to ensure operability may (e.g., only) be required. Different WTRU vendors and / or NW vendors could have their own trained models and / or interoperability can be achieved (e.g., easily) on the fly by transmitting a small amount of data synchronizing data.

Claims

CLAIMS:

1. A wireless transmit / receive unit (WTRU) comprising: a processor configured to: receive configuration information indicating a set of anchor vectors; encode the set of anchor vectors using an encoder to generate encoded anchor vectors; encode channel state information (CSI) using the encoder to generate encoded CSI; determine at least one relative representation based on the encoded anchor vectors and the encoded CSI, wherein the at least one relative representation is indicative of similarity between the encoded anchor vectors and the encoded CSI; and send an indication of the at least one relative representation to a base station.

2. The WTRU of claim 1, wherein each anchor vector of the set of anchor vectors is associated with at least one dataset of a plurality of datasets.

3. The WTRU of claim 1, wherein each relative representation of the at least one relative representation is a scalar value and is defined by the dimensionality of each anchor vector in the set of anchor vectors, and wherein the at least one relative representation is indicative of compressed channel state information (CSI) feedback.

4. The WTRU of claim 1, wherein the processor is configured to send an indication that indicates a capability of the WTRU associated with one or more artificial intelligence or machine learning (AI / ML) models.

5. The WTRU of claim 1, wherein the configuration information comprises a plurality of datasets, training criteria, and training information, and wherein the processor is configured to use each dataset of the plurality of data sets and the set of anchor vectors to train the encoder based on the training criteria and the training information such that the encoder is trained to be interoperable with at least one model at the base station.

6. The WTRU of claim 1, wherein the processor is further configured to:quantize the relative representations, wherein the indication of the relative representations is an indication of the quantized relative representations.

7. The WTRU of claim 1, wherein the set of anchor vectors is a subset of anchor vectors, and wherein the processor is configured to determine the subset of anchor vectors by excluding at least one anchor vector comprised in the set of anchor vectors, and wherein the processor being configured to encode the subset of anchor vectors comprises the processor being configured to encode the subset of anchor vectors.

8. The WTRU of claim 1, wherein the configuring information comprises a similarity metric, and wherein the at least one relative representation is further determined based on the similarity metric, wherein the similarity metric comprises one or more of a cosine similarity or a distance metric.

9. The WTRU of claim 1, wherein the configuration information comprises a plurality of datasets, and wherein each dataset of the plurality of datasets is associated with a respective dataset identification (ID), wherein the set of anchor vectors is associated with an anchor set ID, and wherein the processor being configured to send the indication of relative representations comprises the processor being configured to send an indication of each dataset ID or the anchor set ID.

10. The WTRU of claim 9, wherein the processor is configured to receive a command, wherein the configuration information comprises interoperability criteria, wherein the encoded anchor vectors are generated based on the command or the interoperability criteria, wherein the command comprises a radio resource configuration (RRC) message or a system information block (SIB) message that indicates one or more dataset IDs or one or more anchor set IDs, wherein the encoded anchor vectors are generated based on the one or more dataset IDs or the one or more anchor set IDs, wherein the interoperability criteria is indicative of interoperability between a first model at the WTRU and a second model at the base station, and wherein the interoperability criteria comprises one or more of: a bottle-neck dimension size, a quantization type, a number of quantization bits, or an input data type.

11. The WTRU of claim 1, wherein the configuration information comprises an indication of a plurality of datasets, and wherein the plurality of datasets is associated with the set of anchor vectors.

12. The WTRU of claim 1, wherein the encoded CSI comprises one or more data points in a latent space, and wherein the processor is configured to determine a distance between each data point in the latent space to each encoded anchor vector to determine the similarity between the encoded CSI and the encoded anchor vectors to determine the at least one relative representation.

13. The WTRU of claim 1, wherein the processor is configured to: receive an indication that indicates one or more datasets, wherein the encoder is trained based on the indication of the one or more datasets.

14. A method performed by a wireless transmit / receive unit (WTRU), the method comprising: receiving configuration information indicating a set of anchor vectors; encoding the set of anchor vectors using an encoder to generate encoded anchor vectors; encoding channel state information (CSI) using the encoder to generate encoded CSI; determining at least one relative representation based on the encoded anchor vectors and the encoded CSI, wherein the at least one relative representation is indicative of similarity between the encoded anchor vectors and the encoded CSI; and sending an indication of the at least one relative representation to a base station.

15. The method of claim 14, wherein each anchor vector of the set of anchor vectors is associated with at least one dataset of a plurality of datasets.

16. The method of claim 14, wherein each relative representation of the at least one relative representation is a scalar value and is defined by the dimensionality of each anchor vector in the set of anchor vectors, and wherein the at least one relative representation is indicative of compressed channel state information (CSI) feedback.

17. The method of claim 14, further comprising sending an indication that indicates a capability of the WTRU associated with one or more artificial intelligence or machine learning (AI / ML) models.

18. The method of claim 14, wherein the configuration information comprises a plurality of datasets, training criteria, and training information, and wherein the method further comprising using each dataset of the plurality of data sets and the set of anchor vectors to train the encoder based on the training criteria andthe training information such that the encoder is trained to be interoperable with at least one model at the base station.

19. The method of claim 14, further comprising quantizing the relative representations, wherein the indication of the relative representations is an indication of the quantized relative representations.

20. The method of claim 14, wherein the set of anchor vectors is a subset of anchor vectors, and wherein the method further comprising determining the subset of anchor vectors by excluding at least one anchor vector comprised in the set of anchor vectors, and wherein encoding the subset of anchor vectors comprises encoding the subset of anchor vectors.

Images