RELATED APPLICATION

[0001] This application claims priority to U.S. Provisional Application Serial No. 63/387,833 filed 16 December 2022 entitled “ENCODING AND DECODING OF INPUT INFORMATION,” the disclosure of which is incorporated by reference herein in its entirety.

TECHNICAL FIELD

[0002] The present disclosure relates to wireless communications, and more specifically to input data reporting in wireless communications.

BACKGROUND

[0003] A wireless communications system may include one or multiple network communication devices, such as base stations, which may be otherwise known as an eNodeB (eNB), a nextgeneration NodeB (gNB), or other suitable terminology. Each network communication devices, such as a base station may support wireless communications for one or multiple user communication devices, which may be otherwise known as user equipment (UE), or other suitable terminology. The wireless communications system may support wireless communications with one or multiple user communication devices by utilizing resources of the wireless communication system (e.g., time resources (e.g., symbols, slots, subframes, frames, or the like) or frequency resources (e.g., subcarriers, carriers). Additionally, the wireless communications system may support wireless communications across various radio access technologies including third generation (3G) radio access technology, fourth generation (4G) radio access technology, fifth generation (5G) radio access technology, among other suitable radio access technologies beyond 5G (e.g., sixth generation (6G)).

[0004] Some wireless communications systems provide ways for gathering and reporting CSI, such as by UEs. Such implementations, however, may be inefficient and may fail to accurately communicate CSI. SUMMARY

[0005] The present disclosure relates to methods, apparatuses, and systems that support encoding and decoding of input information. For instance, implementations provide systems and signalling for compressing input feedback data (e.g., channel state information (CSI)), quantizing, and transferring the compressed information, as well as reconstructing the input data from the received compressed and quantized data. The described implementations can utilize multiple latent representations of the input feedback data which can be quantized using different quantization schemes. Further, the described implementations are adaptable to use different numbers of feedback bits. Implementations also include aspects for training neural network models for compressing and decompressing feedback data and explained how the neural network models can be used during the inference phase at a receiver side with assistance signals from a transmitter side.

[0006] By utilizing the described techniques, signaling overhead for communicating feedback data (e.g., CSI) can be decreased and accuracy of feedback data reporting can be increased, thus decreasing usage of system resources and increasing signal quality for wireless communication.

[0007] Some implementations of the methods and apparatuses described herein may further include generating, at a first apparatus, a subset of latent representations of an input data from a set of latent representations based at least in part on a first set of information of an encoder neural network model and a second set of information; generating at least a quantized representation of a latent representation of the subset of latent representations and based at least in part on at least one of a scalar quantization scheme or a vector quantization scheme, and a corresponding set of quantizer parameters; and transmitting a third set of information based at least in part on one or more of the subset of latent representations or the quantized representation of the latent representation.

[0008] Some implementations of the methods and apparatuses described herein may further include: where the first set of information includes at least one of a structure of the encoder neural network model or one or more weights of the encoder neural network model; where the first set of information includes an indication of the encoder neural network model from a plurality of encoder neural network models; determining the first set of information based at least in part on an indication from a second apparatus; the second apparatus includes an apparatus to which the first apparatus transmits the third set of information, or a different apparatus than the apparatus to which the first apparatus transmits the third set of information; determining the first set of information in conjunction with training of the encoder neural network model; where the input data is based at least in part on a channel data representation; where the set of quantizer parameters corresponds to at least one of a quantization codebook associated with the vector quantization scheme, a type of the scalar quantization scheme, or a number of quantization levels for the scalar quantization scheme; determining the set of quantizer parameters based at least in part on a predefined value or an indication from a second apparatus; where the second apparatus includes an apparatus to which the first apparatus transmits the third set of information, or a different apparatus than the apparatus to which the first apparatus transmits the third set of information; determining the set of quantizer parameters in conjunction with training of the encoder neural network model.

[0009] Some implementations of the methods and apparatuses described herein may further include: where the second set of information is based at least in part on one or more of a predefined order or an indication received from a second apparatus corresponding to the subset of latent representations; where the second apparatus includes an apparatus to which the first apparatus transmits the third set of information, or a different apparatus than the apparatus to which the first apparatus transmits the third set of information; where the second set of information is based at least in part on one or more of the input data, specifications of the input data, features of the input data, an expected output of a encoder neural network model, a state of a transmission medium between the first apparatus and a second apparatus to which the third set of information is to be transmitted, or one or more model design parameters of the encoder neural network model.

[0010] Some implementations of the methods and apparatuses described herein may further include: generating the subset of latent representations to include at least one of the latent representation quantized based on the vector quantization scheme or the latent representation quantized based on the scalar quantization scheme; determining a fourth set of information based at least in part on the subset of latent representations; where the third set of information is based on at least in part on the fourth set of information; where the fourth set of information includes an indication of the subset of latent representations selected from the set of latent representations; where the fourth set of information includes a number of latent representations selected from the set of latent representations; where the fourth set of information includes one or more of a size of latent representations or a size of the quantized representation generated from the set of latent representations; where the fourth set of information includes one or more of a number of latent representations associated with the vector quantization scheme selected in the subset of latent representations, or a number of latent representations associated with the scalar quantization scheme selected in the subset of latent representations; transmitting the third set of information to a second apparatus, and where one of: the first apparatus includes a user equipment and the second apparatus includes a network entity; or the first apparatus includes a network entity and the second apparatus includes a user equipment.

[0011] Some implementations of the methods and apparatuses described herein may further include: receiving, at a first apparatus, a first set of information from a second apparatus; generating decoder input of a decoder neural network model based on at least one of the first set of information or a first set of parameters; and generating output of the decoder neural network model using the decoder input and a second set of information used to determine the decoder neural network model for decoding the decoder input.

[0012] Some implementations of the methods and apparatuses described herein may further include: where the first set of information includes a set of latent representations of data; where the first set of information includes an indication of a relation between one or more elements of the set of latent representations of data and the decoder input; where the first set of parameters corresponds to at least one of a quantization codebook associated with at least one vector dequantization scheme, a type of at least one scalar dequantization scheme, or a number of quantization levels for the at least one scalar dequantization scheme; determining the first set of parameters based on one or more of a predefined value or an indication received from the second apparatus or a different apparatus than the second apparatus.

[0013] Some implementations of the methods and apparatuses described herein may further include: where determining the first set of parameters in conjunction with training of the decoder neural network model; generating the decoder input based on one or more of a preconfigured value, a predefined value, or an indication received from the second apparatus or a different apparatus than the second apparatus; where the second set of information includes at least one of a structure of the decoder neural network model or one or more weights of the decoder neural network model; determining the second set of information to include the decoder neural network model from a plurality of decoder neural network models; determining the second set of information to include the decoder neural network model based on an indication received from the second apparatus or a different apparatus than the second apparatus; determining the second set of information in conjunction with training of the decoder neural network model; where one of: the first apparatus includes a network entity and the second apparatus includes a user equipment; or the first apparatus includes a user equipment and the second apparatus includes a network entity.

BRIEF DESCRIPTION OF THE DRAWINGS

[0014] FIG. 1 illustrates an example of a wireless communications system that supports encoding and decoding of input information in accordance with aspects of the present disclosure.

[0015] FIG. 2 illustrates a wireless network including a base station (e.g., gNB) and multiple UEs.

[0016] FIG. 3 illustrates a high-level structure of a two-sided model.

[0017] FIGs. 4a and 4b illustrate an example two-sided model system.

[0018] Fig. 5 illustrates a scenario for aperiodic trigger state defining a list of CSI report settings.

[0019] FIG. 6 illustrates an information element for aperiodic trigger state indicating a resource set and QCL information.

[0020] FIG. 7 illustrates an information element for RRC configuration for NZP-CSI-RS/CSI- IM resources.

[0021] FIG. 8 illustrates an example ordering for CSI reporting.

[0022] FIGs. 9a and 9b illustrate an example structure of a two-sided model system that supports encoding and decoding of input information in accordance with aspects of the present disclosure.

[0023] FIG. 10 illustrates an example of a block diagram of devices (e.g., apparatus) that support encoding and decoding of input information in accordance with aspects of the present disclosure. [0024] FIGs. 11 and 12 illustrate flowcharts of methods that support encoding and decoding of input information in accordance with aspects of the present disclosure.

DETAILED DESCRIPTION

[0025] In wireless communications systems, CSI feedback can be reported by UEs to the network, where the CSI feedback is compressed via transformation of a channel over a combination of the spatial domain, frequency domain, or the time domain, with pre-determined sets of spatial, frequency, or time basis vectors, respectively. In addition to conventional CSI feedback mechanisms, artificial intelligence/machine learning (AIZML)-enabled CSI acquisition schemes have been proposed. While AI/ML-enabled CSI acquisition schemes may infer CSI, such schemes may fail to provide for feeding back a part of the CSI from the UE to the network. Such corresponding CSI components, for instance, may not be efficiently inferred or further compressed via existing AI/ML models.

[0026] Accordingly, this disclosure provides for techniques that support encoding and decoding of input information. For instance, implementations provide systems and signalling for compressing input feedback data (e.g., CSI), quantizing, and transferring the compressed information, as well as reconstructing the input data from the received compressed and quantized data. The described implementations can utilize multiple latent representations of the input feedback data which can be quantized using different quantization schemes. Further, the described implementations are adaptable to use different numbers of feedback bits. Implementations also include aspects for training neural network models for compressing and decompressing feedback data and explained how the neural network models can be used during the inference phase at a receiver side with assistance signals from a transmitter side.

[0027] By utilizing the described techniques, signaling overhead for communicating feedback data (e.g., CSI) can be decreased and accuracy of feedback data reporting can be increased, thus decreasing usage of system resources and increasing signal quality for wireless communication.

[0028] Aspects of the present disclosure are described in the context of a wireless communications system. Aspects of the present disclosure are further illustrated and described with reference to device diagrams and flowcharts. [0029] FIG. 1 illustrates an example of a wireless communications system 100 that supports encoding and decoding of input information in accordance with aspects of the present disclosure. The wireless communications system 100 may include one or more network entities 102, one or more UEs 104, a core network 106, and a packet data network 108. The wireless communications system 100 may support various radio access technologies. In some implementations, the wireless communications system 100 may be a 4G network, such as an LTE network or an LTE- Advanced (LTE-A) network. In some other implementations, the wireless communications system 100 may be a 5G network, such as an NR network. In other implementations, the wireless communications system 100 may be a combination of a 4G network and a 5G network, or other suitable radio access technology including Institute of Electrical and Electronics Engineers (IEEE) 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), IEEE 802.20. The wireless communications system 100 may support radio access technologies beyond 5G. Additionally, the wireless communications system 100 may support technologies, such as time division multiple access (TDMA), frequency division multiple access (FDMA), or code division multiple access (CDMA), etc.

[0030] The one or more network entities 102 may be dispersed throughout a geographic region to form the wireless communications system 100. One or more of the network entities 102 described herein may be or include or may be referred to as a network node, a base station, a network element, a RAN, a base transceiver station, an access point, a NodeB, an eNodeB (eNB), a next-generation NodeB (gNB), or other suitable terminology. A network entity 102 and a UE 104 may communicate via a communication link 110, which may be a wireless or wired connection. For example, a network entity 102 and a UE 104 may perform wireless communication (e.g., receive signaling, transmit signaling) over a Uu interface.

[0031] A network entity 102 may provide a geographic coverage area 112 for which the network entity 102 may support services (e.g., voice, video, packet data, messaging, broadcast, etc.) for one or more UEs 104 within the geographic coverage area 112. For example, a network entity 102 and a UE 104 may support wireless communication of signals related to services (e.g., voice, video, packet data, messaging, broadcast, etc.) according to one or multiple radio access technologies. In some implementations, a network entity 102 may be moveable, for example, a satellite associated with a non-terrestrial network. In some implementations, different geographic coverage areas 112 associated with the same or different radio access technologies may overlap, but the different geographic coverage areas 112 may be associated with different network entities 102. Information and signals described herein may be represented using any of a variety of different technologies and techniques. For example, data, instructions, commands, information, signals, bits, symbols, and chips that may be referenced throughout the description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0032] The one or more UEs 104 may be dispersed throughout a geographic region of the wireless communications system 100. A UE 104 may include or may be referred to as a mobile device, a wireless device, a remote device, a remote unit, a handheld device, or a subscriber device, or some other suitable terminology. In some implementations, the UE 104 may be referred to as a unit, a station, a terminal, or a client, among other examples. Additionally, or alternatively, the UE 104 may be referred to as an Internet-of-Things (loT) device, an Internet-of-Everything (loE) device, or machine-type communication (MTC) device, among other examples. In some implementations, a UE 104 may be stationary in the wireless communications system 100. In some other implementations, a UE 104 may be mobile in the wireless communications system 100.

[0033] The one or more UEs 104 may be devices in different forms or having different capabilities. Some examples of UEs 104 are illustrated in FIG. 1. A UE 104 may be capable of communicating with various types of devices, such as the network entities 102, other UEs 104, or network equipment (e.g., the core network 106, the packet data network 108, a relay device, an integrated access and backhaul (IAB) node, or another network equipment), as shown in FIG. 1. Additionally, or alternatively, a UE 104 may support communication with other network entities 102 or UEs 104, which may act as relays in the wireless communications system 100.

[0034] A UE 104 may also be able to support wireless communication directly with other UEs 104 over a communication link 114. For example, a UE 104 may support wireless communication directly with another UE 104 over a device-to-device (D2D) communication link. In some implementations, such as vehicle-to-vehicle (V2V) deployments, V2X deployments, or cellular- V2X deployments, the communication link 114 may be referred to as a sidelink. For example, a UE 104 may support wireless communication directly with another UE 104 over a PC5 interface. [0035] A network entity 102 may support communications with the core network 106, or with another network entity 102, or both. For example, a network entity 102 may interface with the core network 106 through one or more backhaul links 116 (e.g., via an SI, N2, N2, or another network interface). The network entities 102 may communicate with each other over the backhaul links 116 (e.g., via an X2, Xn, or another network interface). In some implementations, the network entities 102 may communicate with each other directly (e.g., between the network entities 102). In some other implementations, the network entities 102 may communicate with each other or indirectly (e.g., via the core network 106). In some implementations, one or more network entities 102 may include subcomponents, such as an access network entity, which may be an example of an access node controller (ANC). An ANC may communicate with the one or more UEs 104 through one or more other access network transmission entities, which may be referred to as a radio heads, smart radio heads, or transmission-reception points (TRPs).

[0036] In some implementations, a network entity 102 may be configured in a disaggregated architecture, which may be configured to utilize a protocol stack physically or logically distributed among two or more network entities 102, such as an integrated access backhaul (IAB) network, an open RAN (O-RAN) (e.g., a network configuration sponsored by the O-RAN Alliance), or a virtualized RAN (vRAN) (e.g., a cloud RAN (C-RAN)). For example, a network entity 102 may include one or more of a central unit (CU), a distributed unit (DU), a radio unit (RU), a RAN Intelligent Controller (RIC) (e.g., a Near-Real Time RIC (Near-real time (RT) RIC), a Non-Real Time RIC (Non-RT RIC)), a Service Management and Orchestration (SMO) system, or any combination thereof.

[0037] An RU may also be referred to as a radio head, a smart radio head, a remote radio head (RRH), a remote radio unit (RRU), or a transmission reception point (TRP). One or more components of the network entities 102 in a disaggregated RAN architecture may be co-located, or one or more components of the network entities 102 may be located in distributed locations (e.g., separate physical locations). In some implementations, one or more network entities 102 of a disaggregated RAN architecture may be implemented as virtual units (e.g., a virtual CU (VCU), a virtual DU (VDU), a virtual RU (VRU)).

[0038] Split of functionality between a CU, a DU, and an RU may be flexible and may support different functionalities depending upon which functions (e.g., network layer functions, protocol layer functions, baseband functions, radio frequency functions, and any combinations thereof) are performed at a CU, a DU, or an RU. For example, a functional split of a protocol stack may be employed between a CU and a DU such that the CU may support one or more layers of the protocol stack and the DU may support one or more different layers of the protocol stack. In some implementations, the CU may host upper protocol layer (e.g., a layer 3 (L3), a layer 2 (L2)) functionality and signaling (e.g., radio resource control (RRC), service data adaption protocol (SDAP), Packet Data Convergence Protocol (PDCP)). The CU may be connected to one or more DUs or RUs, and the one or more DUs or RUs may host lower protocol layers, such as a layer 1 (LI) (e.g., physical (PHY) layer) or an L2 (e.g., radio link control (RLC) layer, Medium Access Control (MAC) layer) functionality and signaling, and may each be at least partially controlled by the CU.

[0039] Additionally, or alternatively, a functional split of the protocol stack may be employed between a DU and an RU such that the DU may support one or more layers of the protocol stack and the RU may support one or more different layers of the protocol stack. The DU may support one or multiple different cells (e.g., via one or more RUs). In some implementations, a functional split between a CU and a DU, or between a DU and an RU may be within a protocol layer (e.g., some functions for a protocol layer may be performed by one of a CU, a DU, or an RU, while other functions of the protocol layer are performed by a different one of the CU, the DU, or the RU).

[0040] A CU may be functionally split further into CU control plane (CU-CP) and CU user plane (CU-UP) functions. A CU may be connected to one or more DUs via a midhaul communication link (e.g., Fl, Fl-c, Fl-u), and a DU may be connected to one or more RUs via a fronthaul communication link (e.g., open fronthaul (FH) interface). In some implementations, a midhaul communication link or a fronthaul communication link may be implemented in accordance with an interface (e.g., a channel) between layers of a protocol stack supported by respective network entities 102 that are in communication via such communication links.

[0041] The core network 106 may support user authentication, access authorization, tracking, connectivity, and other access, routing, or mobility functions. The core network 106 may be an evolved packet core (EPC), or a 5G core (5GC), which may include a control plane entity that manages access and mobility (e.g., a mobility management entity (MME), an access and mobility management functions (AMF)) and a user plane entity that routes packets or interconnects to external networks (e.g., a serving gateway (S-GW), a Packet Data Network (PDN) gateway (P- GW), or a user plane function (UPF)). In some implementations, the control plane entity may manage non-access stratum (NAS) functions, such as mobility, authentication, and bearer management (e.g., data bearers, signal bearers, etc.) for the one or more UEs 104 served by the one or more network entities 102 associated with the core network 106.

[0042] The core network 106 may communicate with the packet data network 108 over one or more backhaul links 116 (e.g., via an SI, N2, N2, or another network interface). The packet data network 108 may include an application server 118. In some implementations, one or more UEs 104 may communicate with the application server 118. A UE 104 may establish a session (e.g., a PDU session, or the like) with the core network 106 via a network entity 102. The core network 106 may route traffic (e.g., control information, data, and the like) between the UE 104 and the application server 118 using the established session (e.g., the established PDU session). The PDU session may be an example of a logical connection between the UE 104 and the core network 106 (e.g., one or more network functions of the core network 106).

[0043] In the wireless communications system 100, the network entities 102 and the UEs 104 may use resources of the wireless communication system 100 (e.g., time resources (e.g., symbols, slots, subframes, frames, or the like) or frequency resources (e.g., subcarriers, carriers) to perform various operations (e.g., wireless communications). In some implementations, the network entities 102 and the UEs 104 may support different resource structures. For example, the network entities 102 and the UEs 104 may support different frame structures. In some implementations, such as in 4G, the network entities 102 and the UEs 104 may support a single frame structure. In some other implementations, such as in 5G and among other suitable radio access technologies, the network entities 102 and the UEs 104 may support various frame structures (e.g., multiple frame structures). The network entities 102 and the UEs 104 may support various frame structures based on one or more numerologies.

[0044] One or more numerologies may be supported in the wireless communications system 100, and a numerology may include a subcarrier spacing and a cyclic prefix. A first numerology (e.g., /r=0) may be associated with a first subcarrier spacing (e.g., 15 kHz) and a normal cyclic prefix. The first numerology (e.g., /r=0) associated with the first subcarrier spacing (e.g., 15 kHz) may utilize one slot per subframe. A second numerology (e.g., jU=l) may be associated with a second subcarrier spacing (e.g., 30 kHz) and a normal cyclic prefix. A third numerology (e.g., /r=2) may be associated with a third subcarrier spacing (e.g., 60 kHz) and a normal cyclic prefix or an extended cyclic prefix. A fourth numerology (e.g., jU=3) may be associated with a fourth subcarrier spacing (e.g., 120 kHz) and a normal cyclic prefix. A fifth numerology (e.g., /r=4) may be associated with a fifth subcarrier spacing (e.g., 240 kHz) and a normal cyclic prefix.

[0045] A time interval of a resource (e.g., a communication resource) may be organized according to frames (also referred to as radio frames). Each frame may have a duration, for example, a 10 millisecond (ms) duration. In some implementations, each frame may include multiple subframes. For example, each frame may include 10 subframes, and each subframe may have a duration, for example, a 1 ms duration. In some implementations, each frame may have the same duration. In some implementations, each subframe of a frame may have the same duration.

[0046] Additionally or alternatively, a time interval of a resource (e.g., a communication resource) may be organized according to slots. For example, a subframe may include a number (e.g., quantity) of slots. Each slot may include a number (e.g., quantity) of symbols (e.g., orthogonal frequency-division multiplexing (OFDM) symbols). In some implementations, the number (e.g., quantity) of slots for a subframe may depend on a numerology. For a normal cyclic prefix, a slot may include 14 symbols. For an extended cyclic prefix (e.g., applicable for 60 kHz subcarrier spacing), a slot may include 12 symbols. The relationship between the number of symbols per slot, the number of slots per subframe, and the number of slots per frame for a normal cyclic prefix and an extended cyclic prefix may depend on a numerology. It should be understood that reference to a first numerology (e.g., /r=0) associated with a first subcarrier spacing (e.g., 15 kHz) may be used interchangeably between subframes and slots.

[0047] In the wireless communications system 100, an electromagnetic (EM) spectrum may be split, based on frequency or wavelength, into various classes, frequency bands, frequency channels, etc. By way of example, the wireless communications system 100 may support one or multiple operating frequency bands, such as frequency range designations FR1 (410 MHz - 7.125 GHz), FR2 (24.25 GHz - 52.6 GHz), FR3 (7.125 GHz - 24.25 GHz), FR4 (52.6 GHz - 114.25 GHz), FR4a or FR4-1 (52.6 GHz - 71 GHz), and FR5 (114.25 GHz - 300 GHz). In some implementations, the network entities 102 and the UEs 104 may perform wireless communications over one or more of the operating frequency bands. In some implementations, FR1 may be used by the network entities 102 and the UEs 104, among other equipment or devices for cellular communications traffic (e.g., control information, data). In some implementations, FR2 may be used by the network entities 102 and the UEs 104, among other equipment or devices for short- range, high data rate capabilities.

[0048] FR1 may be associated with one or multiple numerologies (e.g., at least three numerologies). For example, FR1 may be associated with a first numerology (e.g., ^=0), which includes 15 kHz subcarrier spacing; a second numerology (e.g., /z=l ), which includes 30 kHz subcarrier spacing; and a third numerology (e.g., /r=2), which includes 60 kHz subcarrier spacing. FR2 may be associated with one or multiple numerologies (e.g., at least 2 numerologies). For example, FR2 may be associated with a third numerology (e.g., /r=2), which includes 60 kHz subcarrier spacing; and a fourth numerology (e.g., /r=3), which includes 120 kHz subcarrier spacing.

[0049] According to implementations for encoding and decoding of input information, a UE 104 receives wireless signal 120 from a network entity 102 (e.g., a base station), such as part of downlink communication from the network entity 102 to the UE 104. The UE 104 determines signal conditions (e.g., CSI) of the wireless signal 120 and performs feedback encoding 122 to generate encoded feedback 124 based on the signal conditions, such as described throughout this disclosure. The UE 104 transmits the encoded feedback 124 to the network entity 102 and the network entity 102 performs feedback decoding 126 to decode the encoded feedback 124 and identify signal condition information (e.g., CSI) included in the encoded feedback 124. The network entity 102 can then perform transmission adaption 128 based at least in part on the decoded signal condition information, such as to attempt to increase signal quality between the network entity 102 and the UE 104.

[0050] FIG. 2 illustrates a wireless network 200 including a base station (e.g., gNB) and multiple UEs. The UEs, for instance, include a UEi, UE2, and UEK. The base station can be represented as a node B equipped with M antennas and the K UEs denoted by U ₁ , U ₂ , ---, U _K with each having N antennas. H^(t) can denote a channel at time t over a frequency band I, I G {1,2, ... , L] , between B ₁ and U _k which is a matrix of size N x M with complex entries, i.e., H^(t) 6 ( NXM [0051] At time t and frequency band I, it can be assumed that the base station is to transmit a message x ^k (t) to U _K , where K = {1,2, ••• , K} while the base station uses w ^k (t) 6 C ^Mxl as the precoding vector. The received signal at U _k , y ^k (t), can be indicated as: y ^k (t) = H ^k (t)w ^k (t)x ^k (t) + n ^k (t) where n ^k (t) represents the noise vector at the receiver.

[0052] To attempt to improve the achievable rate of the link, the base station can select w ^k (t) that maximizes the received Signal-to-Noise Ratio (SNR). Several schemes have been proposed for selection of w ^k (t) where some rely on having some knowledge about H ^k (t).

[0053] The base station can obtain knowledge of H ^k (t) by direct measurement (e.g., in Time- Division Duplexing (TDD) mode and assuming reciprocity of the channel) or indirectly using information that the UE sends to the base station (e.g., in Frequency -Division Duplexing (FDD) mode). In the latter case, a large amount of feedback may be needed to send accurate information about H ^k (t).

[0054] In the description herein, implementations are discussed with reference to a single time slot, but implementations can be further extended to more than a single time slot. Thus, H ^k (t) can be denoted using H ^k .

[0055] H ^k (t) can be defined as matrix of size N x M X L which can be composed by stacking H ^k for multiple frequency bands, e.g., the entries at H ^k [n, m, can be equal to H ^k [n, m](t). Thus, each UE can be feeding back the information about the most recent N x M x L complex numbers to the base station.

[0056] Several methods have been proposed to attempt to reduce the rate of required feedback. For instance, a group of these methods include two parts where a first part is deployed at the UE side and the second part is deployed at the base station side. The UE and base station sides include one or more neural network blocks which are trained using data driven approaches. The UE side can compute a latent representation of input data (e.g., to be transferred to the base station), such as with as low number of bits as possible. The base station can receive data transmitted by the UE side, and the base station can attempt to reconstruct the information intended by the UE to be transmitted to the base station.

[0057] FIG. 3 illustrates a high-level structure of a two-sided model 300. The two-sided model 300 includes a with neural network-based UE and gNB sides referred to here as M _e (encoding model) and M _d (decoding model), respectively. The input of the model is based on the channel measurement, can be for example be raw channel measurement, or eigenvectors associated to the measured channel.

[0058] FIGs. 4a and 4b illustrate an example two-sided model system 400. The two-sided model system 400, for instance, can be utilized for communicating signal information, such as CSI feedback between a UE and a network entity. FIG. 4a, for instance, illustrates a model subsystem 400a that represents an encoder portion of the two-sided model system 400, such as implemented at a UE. Further, FIG. 4b illustrates a model subsystem 400b that represents a decoder portion of the two-sided model system 400, such as implemented at a network entity such as a base station, e.g., gNB.

[0059] In implementations information about H ^k [n, m, Z](t) (e.g., CSI) is to be sent from the subsystem 400a (e.g., UE) to the subsystem 400b, e.g., gNB. Further, transmission of one or more eigenvectors associated with channel state can be performed. In the subsystems 400a, 400b, blocks Bl to B6 can be multilayer neural networks. The subsystem 400a (encoder) can generate two latent representations of the input data, e.g., Int t l and Int_t_2. The subsystem 400a quantizes the latent representations using vector and scalar quantization methods, respectively. The result of the quantization is transmitted to the subsystem 400b (e.g., gNB) where the decoder attempts to generate an accurate output, such as a reconstruction of the input data. In the system, the subsystem 400a and subsystem 400b Quantization Codebook can be assumed to be the same and their value along the weights of the neural networks (e.g., for blocks B) can be learned during a training phase. Other hyperparameters like value of J and the number of quantization levels (Q) for the “Quantizer Module” also can be set during the training phase.

[0060] In implementations, the subsystem 400a (encoder) generates at least one latent representation of the model and quantizes the latent representation(s) using at least one of scalar quantization and/or one vector quantization scheme. The subsystem 400b (decoder, e.g., gNB) receives these quantized data, dequantizes the quantized data, and then use them as the input of the decoder. With the described two-sided model system 400, the total bits for transmission of quantized information (for both scalar and vector quantization) can be equal to the number of feedback bits.

[0061] In implementations, a number of feedback bits can be different based on different factors including number of layers to feedback, bandwidth, and the load of the gNB. Thus, implementations described herein can address the issue of how a neural network model (e.g., the two-sided model system 400) can support different numbers of feedback bits.

[0062] In implementations, CSI codebooks can be defined as well as feedback for CSI-related bits. For instance, assume a gNB is equipped with a two-dimensional (2D) antenna array with Ni, N2 antenna ports per polarization placed horizontally and vertically and communication occurs over N3 PMI sub-bands. A PMI subband consists of a set of resource blocks, each resource block consisting of a set of subcarriers. In such cases, 2N1N2 CSI-Reference Signal (RS) ports can be utilized to enable downlink (DL) channel estimation with high resolution for NR Rel. 15 Type-II codebook. In order to reduce uplink feedback overhead, a Discrete Fourier transform (DFT)-based CSI compression of the spatial domain can be applied to L dimensions per polarization, where L<NIN2. In the sequel the indices of the 2L dimensions can be referred as the Spatial Domain (SD) basis indices. The amplitude and phase values of the linear combination coefficients for each subband can be fed back to the gNB as part of the CSI report. The 2N1N2 N3 codebook per layer / can take on the form

W _L = W ₁ W _2fl , where Wi is a 2NIN2 2L block-diagonal matrix (L<NIN2) with two identical diagonal blocks, e.g., and B is an N1N2XL matrix with columns drawn from a 2D oversampled DFT matrix, as follows. ₂ , where the superscript ^T denotes a matrix transposition operation.

[0063] Note that Oi, C oversampling factors can be assumed for the 2D DFT matrix from which matrix B is drawn. Note that Wi can be common across all layers. W2,i is a 2Lx Ns matrix, where the i ^th column corresponds to the linear combination coefficients of the 2L beams in the i ^th sub-band. The indices of the L selected columns of B can be reported, along with the oversampling index taking on O1O2 values. Note that W2,i are independent for different layers.

[0064] For Type-II Port Selection codebook, K (where K < 2N1N2) beamformed CSI-RS ports can be utilized in DL transmission, in order to reduce complexity. The KxNs codebook matrix per layer can take the form

W _L = W^ ^s W _{2 l} .

[0065] Here, W2 can follow the same structure as the conventional NR Rel. 15 Type-II Codebook and can be layer specific. W ^s is a KN2L block-diagonal matrix with two identical diagonal blocks, e.g., and E is an - X L matrix which columns can be standard unit vectors, such as follows.

2 where e ₍ ^! is a standard unit vector with a 1 at the i ^th location. Here dps is an RRC parameter which takes on the values {1,2, 3, 4} under the condition dps < min(K/2, L whereas mps takes on the values

] 0, , —7— — 11 and is reported as part of the uplink CSI feedback overhead. Wi can be common I I 2dps I J across all layers. [0066] For '= l 6, /.=4 and dps =1, the 8 possible realizations of E corresponding to mps = {0,1,..., 7} are as follows

1 00 0- 0 00 0 0 00 0- 0 00 0- 0 00 0 0 00 1 0 01 0 0 10 0 0 10 0 1 00 0 0 00 0 0 00 0 0 00 0 0 00 0 0 00 1 0 01 0

0 01 0 0 10 0 1 00 0 0 00 0 0 00 0 0 00 0 0 00 0 0 00 1

0 00 1 0 01 0 0 10 0 1 00 0 0 00 0 0 00 0 0 00 0 0 00 0

0 00 0 0 00 1 0 01 0 0 10 0 1 00 0 0 00 0 0 00 0 0 00 0

0 00 0 0 00 0 0 00 1 0 01 0 0 10 0 1 00 0 0 00 0 0 00 0

0 00 0 0 00 0 0 00 0 0 00 1 0 01 0 0 10 0 1 00 0 0 00 0

-0 00 0- -0 00 0 -0 00 0- -0 00 0- -0 00 1 -0 01 0 -0 10 0 -1 00 0

[0067] When dps =2, the 4 possible realizations of E corresponding to mps ={0,1, 2, 3} are as follows

1 00 0- 0 00 0 0 00 0 0 01 0 0 10 0 0 00 0 0 00 0 0 00 1

0 01 0 1 00 0 0 00 0 0 00 0

0 00 1 0 10 0 0 00 0 0 00 0

0 00 0 0 01 0 1 00 0 0 00 0

0 00 0 0 00 1 0 10 0 0 00 0

0 00 0 0 00 0 0 01 0 1 00 0

-0 00 0- -0 00 0 -0 00 1 -0 10 0

[0068] When dps =3, the 3 possible realizations of E corresponding of mps = {0,1,2} are as follows

1 00 0- 0 00 0 0 01 0 0 10 0 0 00 0 0 00 1

0 01 0 0 00 0 0 00 0

0 00 1 1 00 0 0 00 0

0 00 0 0 10 0 0 00 0

0 00 0 0 01 0 0 00 0

0 00 0 0 00 1 1 00 0

-0 00 0- -0 00 0 -0 10 0

[0069] When dps =4, the 2 possible realizations of E corresponding of mps = {0,1 } are as follows

[0070] To summarize, mps parametrizes the location of the first 1 in the first column of E, whereas dps represents the row shift corresponding to different values of mps.

[0071] NR Rel. 15 Type-I codebook can be considered a baseline codebook for NR, with a variety of configurations. The most common utility of Rel. 15 Type-I codebook is a special case of NRRel. 15 Type-II codebook with L=1 for Rank Indicator (RI)=1,2, wherein a phase coupling value is reported for each sub-band, e.g., W2,i is 2x2V?, with the first row equal to [1, 1, ... , 1] and the second row equal to ]. Under specific configurations, < > ₀ ⁼ i ^{= =}

4>N ₃ -I, e.g., wideband reporting. For RI>2 different beams are used for each pair of layers.

[0072] For NRRel. 16 Type-II codebook, it can be assumed a gNB is equipped with a two- dimensional (2D) antenna array with Ni, N2 antenna ports per polarization placed horizontally and vertically and communication occurs over N3 PMI subbands. A PMI subband consists of a set of resource blocks, each resource block consisting of a set of subcarriers. In such cases, 2N1N2N3 CSI- RS ports are utilized to enable DL channel estimation with high resolution for NRRel. 16 Type-II codebook. In order to reduce the uplink feedback overhead, a Discrete Fourier transform (DFT)- based CSI compression of the spatial domain can be applied to L dimensions per polarization, where L<NIN2. Similarly, additional compression in the frequency domain can be applied, where each beam of the frequency-domain precoding vectors is transformed using an inverse DFT matrix to the delay domain, and the amplitude and phase values of a subset of the delay-domain coefficients are selected and fed back to the gNB as part of the CSI report. The 2N1N2XN3 codebook per layer takes on the form

W _L = w ₁ w _2:l w^ _l , where Wi is a 2NiN 2 2L block-diagonal matrix (L<NIN ₂ ') with two identical diagonal blocks, e.g., ?]■ and B is an N1N2XL matrix with columns drawn from a 2D oversampled DFT matrix, as follows. where the superscript ^T denotes a matrix transposition operation.

[0073] Note that Oi, C oversampling factors can be assumed for the 2D DFT matrix from which matrix B is drawn. Note that Wi can be common across all layers. W/,i is an jxAF matrix (M<Ns) with columns selected from a critically-sampled size-M? DFT matrix, as follows

[0074] Further, indices of the L selected columns of B can be reported along with the oversampling index taking on Oi values. Similarly, for W/i, only the indices of the M selected columns out of the predefined size-N? DFT matrix can be reported. In the sequel the indices of the AT dimensions are referred as the selected Frequency Domain (FD) basis indices. Hence, L, M represent the equivalent spatial and frequency dimensions after compression, respectively.

[0075] Further, the 2/.xA7 matrix VF ₂ represents the linear combination coefficients (LCCs) of the spatial and frequency DFT-basis vectors. Both IV ₂ , and Wf are selected independently for different layers. Amplitude and phase values of an approximately fraction of the 2LM available coefficients are reported to the gNB (/<! ) as part of the CSI report. Note that coefficients with zero amplitude values are indicated via a layer-specific bitmap matrix Si of size 2LxM, wherein each bit of the bitmap matrix Si indicates whether a coefficient has a zero-amplitude value, wherein for these coefficients no quantized amplitude and phase values need to be reported. Since all non-zero coefficients reported within a layer are normalized with respect to the coefficient with the largest amplitude value (e.g., strongest coefficient), where the amplitude and phase values corresponding to the strongest coefficient are set to one and zero, respectively, and hence no further amplitude and phase information is explicitly reported for this coefficient, and only an indication of the index of the strongest coefficient per layer is reported. Hence, for a single-layer transmission, amplitude, and phase values of a maximum of [2/i/.A/|- l coefficients (along with the indices of selected L, M DFT vectors) are reported per layer, leading to significant reduction in CSI report size, compared with reporting 2N/N?xNj - I coefficients’ information. [0076] For Type-II Port Selection codebook for NR Rel. 16, K (where K < 2N1N2) beamformed CSI-RS ports are utilized in DL transmission, in order to reduce complexity. The. The KxNs codebook matrix per layer takes on the form

[0077] Here, IV _{2 ;} and W/i follow the same structure as the conventional NR Rel. 16 Type-II Codebook, where both are layer specific. The matrix IV ⁵ is a Kx2L block-diagonal matrix with the same structure as that in the NR Rel. 15 Type-II Port Selection Codebook.

[0078] Rel. 17 Type-II Port Selection codebook follows a similar structure as that of Rel. 15 and Rel. 16 port-selection codebooks, as follows

[0079] However, unlike Rel. 15 and Rel. 16 Type-II port-selection codebooks, the port-selection matrix W ₄ ^S supports free selection of the K ports. For instance, the 72 ports per polarization out of the CSI-RS ports per polarization, e.g., bits are used to identify the K/2 selected ports per polarization, wherein this selection is common across all layers. Here, 14^2, ( and

W/i follow the same structure as the conventional NR Rel. 16 Type-II Codebook, however AT is limited to 1,2 only, with the network configuring a window of size /V = { 2,4 } for M=2. Moreover, the bitmap is reported unless ?=! and the UE reports all the coefficients for a rank up to a value of two.

[0080] For Rel-18 potential Type-II codebook, the time-domain corresponding to slots is further compressed via DFT-based transformation, wherein the codebook is in the following form where Wi, W/i follow the same structure as Rel- 16 Type-II codebook, Wd.i is an N^Q matrix (O < N4) with columns selected from a critically-sampled size-M DFT matrix, as follows [0081] In implementations the indices of the Q selected columns of Wd,i are reported. Note that Wd,i may be layer specific, e.g., W _{d l} #= _d , or layer common, e.g., Wa,i ^{= =} _d ,Ri, where RI corresponds to the total number of layers, and the operator ® corresponds to a Kronecker matrix product. Here, IV _{2 ;} is a 2 MQ sized matrix with layer-specific entries representing the LCCs corresponding to the spatial-domain, frequency-domain and time-domain DFT-basis vectors. Thereby, a size 2 MQ bitmap may need to be reported associated with Rel-18 Type-II codebook.

[0082] In scenarios a codebook report can be partitioned into two parts based on the priority of information reported. Further, each part can be encoded separately, such as described below. Part 1 of a CSI report can include RI + Channel Quality Indicator (CQI) + Total number of coefficients. Part 2 of a CSI report can include SD basis indicator + FD basis indicator/layer + Bitmap/layer + Coefficient Amplitude info/layer + Coefficient Phase info/layer + Strongest coefficient indicator/layer.

[0083] Further, Part 2 CSI can be decomposed into sub-parts each with different priority, e.g., higher priority information listed first. Such partitioning can allow dynamic reporting size for codebook based on available resources in the uplink phase. Type-II codebook can be based on aperiodic CSI reporting, and reported in Physical Uplink Shared Channel (PUSCH)) via downlink control information (DCI) triggering. Type-I codebook can be based on periodic CSI reporting (e.g., Physical Uplink Control Channel (PUCCH)) or semi-persistent CSI reporting (PUSCH or PUCCH) or aperiodic reporting (PUSCH).

[0084] For priority reporting for Part 2 CSI, note that multiple CSI reports may be transmitted with different priorities, as shown in Table 1 below. Additionally, the priority of the NR _SP CSI reports can be based on the following:

1. A CSI report corresponding to one CSI reporting configuration for one cell may have higher priority compared with another CSI report corresponding to one other CSI reporting configuration for the same cell

2. CSI reports intended to one cell may have higher priority compared with other CSI reports intended to another cell 3. CSI reports may have higher priority based on the CSI report content, e.g., CSI reports carrying LI -Reference Signal Received Power (RSRP) information have higher priority

4. CSI reports may have higher priority based on their type, e.g., whether the CSI report is aperiodic, semi-persistent or periodic, and whether the report is sent via PUSCH or PUCCH, may impact the priority of the CSI report

[0085] Accordingly, CSI reports may be prioritized as follows, where CSI reports with lower identifiers (IDs) have higher priority

Priics/(y, k, c, s) = 2 ■ N _cells ■ M _s ■ y + N _cells ■ M _s ■ k + M _s ■ c + s s: CSI reporting configuration index, and AG: Maximum number of CSI reporting configurations c: Cell index, and N _ce iis'. Number of serving cells k. 0 for CSI reports carrying LI -RSRP or LI - Signal-to-Interference-and-Noise Ratio (SINR), 1 otherwise y: 0 for aperiodic reports, 1 for semi-persistent reports on PUSCH, 2 for semi-persistent reports on PUCCH, 3 for periodic reports.

Table 1 : Priority Reporting Levels for Part 2 CSI

[0086] For triggering aperiodic CSI reporting on PUSCH, a UE is to report CSI information for the network using the CSI framework as in NR Release 15. A triggering mechanism between a report setting and a resource setting can be summarized in Table 2 below.

Table 2: Triggering mechanism between a report setting and a resource setting

[0087] Moreover, note the following:

• Associated Resource Settings for a CSI Report Setting can have same time domain behaviour. • Periodic CSI-RS/ Interference Management (IM) resource and CSI reports can be assumed to be present and active once configured by RRC

• Aperiodic and semi-persistent CSI-RS/ IM resources and CSI reports can be explicitly triggered or activated.

• For aperiodic CSI-RS/ IM resources and aperiodic CSI reports, the triggering is done jointly by transmitting a DCI Format 0-1.

• Semi-persistent CSI-RS/ IM resources and semi-persistent CSI reports are independently activated.

[0088] For aperiodic CSI-RS/ IM resources and aperiodic CSI reports, the triggering can be done jointly by transmitting a DCI Format 0-1. The DCI Format 0 1 contains a CSI request field (0 to 6 bits). A non-zero request field points to a so-called aperiodic trigger state configured by RRC (see, e.g., FIG. 5). An aperiodic trigger state in turn is defined as a list of up to 16 aperiodic CSI Report Settings, identified by a CSI Report Setting ID for which the UE calculates simultaneously CSI and transmits it on the scheduled PUSCH transmission.

[0089] Fig. 5 illustrates a scenario 500 for aperiodic trigger state defining a list of CSI report settings. For instance, when a CSI Report Setting is linked with aperiodic Resource Setting (e.g., including multiple Resource Sets), the aperiodic NZP CSI-RS Resource Set for channel measurement, the aperiodic CSI-IM Resource Set, and/or the aperiodic NZP CSI-RS Resource Set for IM to use for a given CSI Report Setting can be included in the aperiodic trigger state definition (see, e.g., the information element 600 of FIG. 6). For aperiodic Non-Zero Power (NZP) CSI-RS, the Quasi Co-Location (QCL) source to use can also be configured in the aperiodic trigger state. A UE can assume that the resources used for the computation of the channel and interference can be processed with the same spatial filter, e.g. quasi-co-located with respect to “QCL-TypeD.”

[0090] FIG. 6 illustrates an information element 600 for aperiodic trigger state indicating a resource set and QCL information.

[0091] FIG. 7 illustrates an information element 700 for RRC configuration for NZP-CSI- RS/CSI-IM resources. [0092] Table 3 below summarizes a type of uplink channels used for CSI reporting as a function of the CSI codebook type.

Table 3: Uplink channels used for CSI reporting as a function of the CSI codebook type

[0093] For aperiodic CSI reporting, PUSCH-based reports can be divided into two CSI parts: CSI Parti and CSI Part 2. CSI Part 1 can have a fixed payload size (e.g., and can be decoded by the gNB without prior information) and can contain the following:

• RI (if reported), CSI-RS Resource Index (CRI) (if reported) and CQI for the first codeword;

• number of non- zero wideband amplitude coefficients per layer for Type II CSI feedback on PUSCH.

[0094] CSI Part 2 can have a variable payload size that can be derived from the CSI parameters in CSI Part 1 and contains Precoding Matrix Indicator (PMI) and the CQI for the second codeword when RI > 4. For example, if the aperiodic trigger state indicated by DCI format 0 1 defines 3 report settings x, y, and z, then the aperiodic CSI reporting for CSI part 2 will be ordered as indicated in FIG. 8.

[0095] In scenarios, CSI reports can be prioritized according to:

1. Time-domain behavior and physical channel, where more dynamic reports are given precedence over less dynamic reports and PUSCH has precedence over PUCCH.

2. CSI content, where beam reports (e.g., Ll-RSRP reporting) has priority over regular CSI reports.

3. The serving cell to which the CSI corresponds (in case of carrier aggregation (CA) operation). CSI corresponding to the PCell has priority over CSI corresponding to Scells.

4. The reportConfigID.

[0096] In scenarios, a number of feedback bits for deep learning based methods can be adjusted. For instance, a number of feedback bits for deep learning methods can be based on: a) A dimension of the first set of latent representations (which are quantized using scalar quantizers) and the number of quantization levels used for scalar quantizing each entry. b) A dimension of the second set of latent representations (which are quantized using vector quantizers) and the number of codewords in the vector quantization codebook.

[0097] For instance, by adjusting these parameters, several two-sided models can be designed for a different number of different feedback bits. In some wireless communications systems an issue presented here may be a limitation of the generalization ability of each model (e.g., each limited to a specific number of feedback bits) and the existence of many models may lead to complexity in model storage and model switching.

[0098] Accordingly, the present disclosure presents ways for construction, training, and sending feedback data for a two-sided model which enables a deep learning-based structure to accommodate different numbers of feedback bits.

[0099] FIGs. 9a, 9b illustrate an example structure of a two-sided model system 900 that supports encoding and decoding of input information in accordance with aspects of the present disclosure. The two-sided model system 900, for instance, can be utilized for communicating signal information, such as CSI feedback between a UE and a network entity. FIG. 9a, for instance, illustrates an encoder model subsystem 900a that represents an encoder portion of the two-sided model system 900, such as implemented at a UE. Further, FIG. 9b illustrates a decoder model subsystem 900b that represents a decoder portion of the two-sided model system 900, such as implemented at a network entity such as a base station, e.g., gNB. While the two-sided model system 900 is discussed with reference to an encoder module 902 using both a vector quantizer and a scalar quantizer, it is to be appreciated that implementations disclosed herein can also be applied with one of the vector quantizer or the scalar quantizer.

[0100] In the two-sided model system 900, there is no limitation on how the Latent 1 to Latent k+m representations are determined (e.g., based on the input data) and any neural network-based techniques can be utilized. In implementations, the encoder module 902 can generate up to k+m latent representations. The first ‘k’ representations can be of size c ₍ x f meaning that c ₍ vectors of size fi for i = {1,2, •••, k}. The next ‘m’ representations can be of size It X 1 meaning that the i ^th representation has the size of It for i = (k + 1, k + 2, ••• , k + m}. Note that it is possible to have either of k or I be equal to zero. The exact values of fi, Ct, f, m and k are design parameters and can be set based on the specification of a particular implementation.

[0101] In the two-sided model system 900 each of the first k latent representations can be quantized using a respective corresponding vector quantizer where the codebook associated with the I ^th i = {1,2, ••• , k} vector quantizer has codewords and the codewords can be set through training and/or can be fixed. Each of the next ‘m’ latent representations can be quantized using a respective corresponding scalar quantizer, where the i ^th i = (k + 1, k + 2, ••• , k + m} scalar quantizer quantizes each value of its input latent representation using Pt bits. The method that each scalar quantizer uses, the codewords of each vector quantizer, the values of and Pt can be design parameters and can be set based on the specification of a particular implementation. Note that the ordering of the latent representation in the two-sided model system 900 is presented for purposes of illustration, e.g., implementations can utilize scaler quantizers first and then vector quantizers, or any combination thereof.

[0102] In implementations of the two-sided model system 900, for transmission of input data, the encoder model subsystem 900a can decide to use a subset of latent representations from the set of latent representations that can be generated by the encoder module 902 associated to the vector quantization scheme, e.g. , a subset v _vec of latent representations where v _vec c {1,2, • • • , k} and/or a subset of latent representations of the latent representations from the set of latent representations associated to the scalar quantization scheme, e.g., set v _sq of latent representations where v _sq c {k + 1, k + 2, ••• , k + m}.

[0103] In implementations, the v _vec can be an incremental subset of {1,2, ••• , k}, e.g., v _vec could be {1}, {1,2}, {1,2,3} but not for example {1,3}. The same restriction may apply for v _seq , requiring v _seq to be only incremental subset of (k + 1, k + 2, ••• , k + m}. The parameter y^ =

0 or 1 represent if latent representation i is active for transmission of an input data, say for example H. Other methods for selection of v _vec (e.g. other restrictions on how v _vec can be selected) can be implemented.

[0104] In implementations and based at least in part on these notations, the number of bits transmitted from the encoder model subsystem 900a to the decoder model subsystem 900b can be equal to

[0105] For instance, each particular selection of y ₍ can result in a particular number of feedback bits. This number of feedback bits (e.g., active sets) can be set in several ways including by another entity in the network, selected to minimize an amount of feedback, selected based on other setting of the network, e.g., how many layers are to be transmitted, a bandwidth of a signal for transmitting feedback, based on a projected distortion at the decoder module 904 and certain threshold, etc.

[0106] In implementations, y _L can follow a certain property meaning that if y ₇ =0, j G {1,2, ••• , k} then y^=0, * G {/ + 1, •••, k}, Similarly, if y ₇ =0, j G {k + 1, k + 2, ••• , k + m} then

[0107] In implementations, instead of having different latent representations, a same latent representation can be utilized for all or a group of quantizers, where the latent representation can be quantized by different vector and/or scalar quantizer schemes. In such implementations, an activation parameter can represent whether a particular quantizer is used for quantization of a latent representation or not.

[0108] In implementations the structure of the decoder module 904 can be static. For instance, the decoder module 904 can assume that there are k + m latent representations as the input of the decode part. However, if the corresponding latent representation from the encoder model subsystem 900a is not active (e.g., /i=0), the decoder module 904 can set the received latent representation to a predetermined value, e.g., all zero vector or matrix.

[0109] In implementations priority can be specified between the latent representations associated with scalar and the vector quantizers. For example, as a first activated latent representations, a first latent representation can be selected or the m ^th latent representation.

[0110] In implementations, a set of active latent representations can be communicated to the decoder model subsystem 900b in a separate message in several ways, for example using a bitmaplike code. In an incremental activation implementation, a set of active latent representations can be sent by transmission of a number of active latent representations for the vector and scalar quantization. For instance, activation can be incremental, e.g., where 3 latent representations are used from the scalar representations, (k + 1, k + 2, k + 3} can be the set of active latent representations. In at least some implementations, a total length of a feedback message can be transmitted by the encoder model subsystem 900a, and the decoder model subsystem 900b can determine which latent representations are active based on the total length and the size of each quantized representation, and optionally a restriction that is specified on selection of an active subset. In at least some implementations, where a receiving node associated with the decoder model subsystem 900b can determine the length of a feedback message (for example if the receiving node sets the feedback size itself), a length of the feedback message cannot be transmitted, and the receiving node can determine which latent representations are active based on a set of bits received.

[OHl] In implementations, training of the two-sided model system 900 can use common stochastic gradient approaches computed on different min-batch of samples and repeated for several epochs. In a forward path of this method, at each step, all the latent representations of the encoder module 902 can be computed based on most recent weights of their corresponding neural network blocks. Then, for each sample, a random subset of {1,2, •••, k} and a random subset of (k + 1, k + 2, ••• , k + m} can be selected, e.g., if incremental activation is specified the subsets are selected accordingly. Based on the selected subsets, the active latent representations can be transmitted to the decoder module 904. In at least some implementations channel noise and/or other sources of error can be modelled in the received versions of latent representations.

[0112] At the decoder model subsystem 900b node, the decoder model subsystem 900b can use the received latent representations for the active subsets and use a predetermined value (e.g., zero) for non-active latent representations. The output of the decoder module 904 can be computed using the described input data and the current values of the neural network blocks at the decoder module 904. A loss can be calculated based on the decoder module 904 output and the desired output, which can be used for the calculation of the gradients to be used for a backward path. In implementations, the codewords of different codebooks and the parameters of the scalar quantizer can be trained during the training procedure and/or can be fixed and preselected.

[0113] In implementations, training schemes for incremental activation scenarios of the two- sided model system 900 can give more weights to the first latent representations of the vector and scalar quantization schemes and can attempt to incrementally add information about the input data into the next representations. For instance, the bits which are more important for construction of the output can be initially transmitted, and gradually as the number of allowed feedback bits increases, more detailed information can be sent by the encoder model subsystem 900a to the decoder model subsystem 900b, such as by sending subsequent latent representations.

[0114] FIG. 10 illustrates an example of a block diagram 1000 of a device 1002 (e.g., an apparatus) that supports encoding and decoding of input information in accordance with aspects of the present disclosure. The device 1002 may be an example of network entity 102 and/or UE 104 as described herein. The device 1002 may support wireless communication with one or more network entities 102, UEs 104, or any combination thereof. The device 1002 may include components for bidirectional communications including components for transmitting and receiving communications, such as a processor 1004, a memory 1006, a transceiver 1008, and an VO controller 1010. These components may be in electronic communication or otherwise coupled (e.g., operatively, communicatively, functionally, electronically, electrically) via one or more interfaces (e.g., buses). [0115] The processor 1004, the memory 1006, the transceiver 1008, or various combinations thereof or various components thereof may be examples of means for performing various aspects of the present disclosure as described herein. For example, the processor 1004, the memory 1006, the transceiver 1008, or various combinations or components thereof may support a method for performing one or more of the operations described herein.

[0116] In some implementations, the processor 1004, the memory 1006, the transceiver 1008, or various combinations or components thereof may be implemented in hardware (e.g., in communications management circuitry). The hardware may include a processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic, discrete hardware components, or any combination thereof configured as or otherwise supporting a means for performing the functions described in the present disclosure. In some implementations, the processor 1004 and the memory 1006 coupled with the processor 1004 may be configured to perform one or more of the functions described herein (e.g., executing, by the processor 1004, instructions stored in the memory 1006). In the context of UE 104, for example, the transceiver 1008 and the processor coupled 1004 coupled to the transceiver 1008 are configured to cause the UE 104 to perform the various described operations and/or combinations thereof.

[0117] For example, the processor 1004 and/or the transceiver 1008 may support wireless communication at the device 1002 in accordance with examples as disclosed herein. For instance, the processor 1004 and/or the transceiver 1008 may be configured as and/or otherwise support a means to generate, at a first apparatus, a subset of latent representations of an input data from a set of latent representations based at least in part on a first set of information of an encoder neural network model and a second set of information; generate at least a quantized representation of a latent representation of the subset of latent representations and based at least in part on at least one of a scalar quantization scheme or a vector quantization scheme, and a corresponding set of quantizer parameters; and transmit a third set of information based at least in part on one or more of the subset of latent representations or the quantized representation of the latent representation.

[0118] Further, in some implementations, the first set of information includes at least one of a structure of the encoder neural network model or one or more weights of the encoder neural network model; the first set of information includes an indication of the encoder neural network model from a plurality of encoder neural network models; the processor is configured to cause the first apparatus to determine the first set of information based at least in part on an indication from a second apparatus; the second apparatus comprises an apparatus to which the first apparatus transmits the third set of information, or a different apparatus than the apparatus to which the first apparatus transmits the third set of information; the processor is configured to cause the first apparatus to determine the first set of information in conjunction with training of the encoder neural network model; the input data is based at least in part on a channel data representation; the set of quantizer parameters corresponds to at least one of a quantization codebook associated with the vector quantization scheme, a type of the scalar quantization scheme, or a number of quantization levels for the scalar quantization scheme; the processor is configured to cause the first apparatus to determine the set of quantizer parameters based at least in part on a predefined value or an indication from a second apparatus.

[0119] Further, in some implementations, the second apparatus includes an apparatus to which the first apparatus transmits the third set of information, or a different apparatus than the apparatus to which the first apparatus transmits the third set of information; the processor is configured to cause the first apparatus to determine the set of quantizer parameters in conjunction with training of the encoder neural network model; the second set of information is based at least in part on one or more of a predefined order or an indication received from a second apparatus corresponding to the subset of latent representations; the second apparatus includes an apparatus to which the first apparatus transmits the third set of information, or a different apparatus than the apparatus to which the first apparatus transmits the third set of information; the second set of information is based at least in part on one or more of the input data, specifications of the input data, features of the input data, an expected output of a encoder neural network model, a state of a transmission medium between the first apparatus and a second apparatus to which the third set of information is to be transmitted, or one or more model design parameters of the encoder neural network model.

[0120] Further, in some implementations, the processor is configured to cause the first apparatus to generate the subset of latent representations to include at least one of the latent representation quantized based on the vector quantization scheme or the latent representation quantized based on the scalar quantization scheme; the processor is configured to cause the first apparatus to: determine a fourth set of information based at least in part on the subset of latent representations; the third set of information is based on at least in part on the fourth set of information; the fourth set of information includes an indication of the subset of latent representations selected from the set of latent representations; where the fourth set of information includes a number of latent representations selected from the set of latent representations; the fourth set of information includes one or more of a size of latent representations or a size of the quantized representation generated from the set of latent representations; where the fourth set of information includes one or more of a number of latent representations associated with the vector quantization scheme selected in the subset of latent representations, or a number of latent representations associated with the scalar quantization scheme selected in the subset of latent representations; the processor is configured to cause the first apparatus to transmit the third set of information to a second apparatus, and one of: the first apparatus includes a user equipment and the second apparatus includes a network entity; or the first apparatus includes a network entity and the second apparatus includes a user equipment.

[0121] In a further example, the processor 1004 and/or the transceiver 1008 may support wireless communication at the device 1002 in accordance with examples as disclosed herein. The processor 1004 and/or the transceiver 1008, for instance, may be configured as or otherwise support a means to receive, at a first apparatus, a first set of information from a second apparatus; generate decoder input of a decoder neural network model based on at least one of the first set of information or a first set of parameters; and generate output of the decoder neural network model using the decoder input and a second set of information used to determine the decoder neural network model for decoding the decoder input.

[0122] Further, in some implementations, the first set of information includes a set of latent representations of data; the first set of information includes an indication of a relation between one or more elements of the set of latent representations of data and the decoder input; the first set of parameters corresponds to at least one of a quantization codebook associated with at least one vector dequantization scheme, a type of at least one scalar dequantization scheme, or a number of quantization levels for the at least one scalar dequantization scheme; the processor is configured to cause the first apparatus to determine the first set of parameters based on one or more of a predefined value or an indication received from the second apparatus or a different apparatus than the second apparatus. [0123] Further, in some implementations, the processor is configured to cause the first apparatus to determine the first set of parameters in conjunction with training of the decoder neural network model; the processor is configured to cause the first apparatus to generate the decoder input based on one or more of a preconfigured value, a predefined value, or an indication received from the second apparatus or a different apparatus than the second apparatus; the second set of information includes at least one of a structure of the decoder neural network model or one or more weights of the decoder neural network model; the processor is configured to cause the first apparatus to determine the second set of information to include the decoder neural network model from a plurality of decoder neural network models; the processor is configured to cause the first apparatus to determine the second set of information to include the decoder neural network model based on an indication received from the second apparatus or a different apparatus than the second apparatus; the processor is configured to cause the first apparatus to determine the second set of information in conjunction with training of the decoder neural network model; one of: the first apparatus includes a network entity and the second apparatus includes a user equipment; or the first apparatus includes a user equipment and the second apparatus includes a network entity.

[0124] The processor 1004 of the device 1002, such as a UE 104, may support wireless communication in accordance with examples as disclosed herein. The processor 1004 includes at least one controller coupled with at least one memory, and the at least one controller is configured to and/or operable to cause the processor to generate a subset of latent representations of an input data from a set of latent representations based at least in part on a first set of information of an encoder neural network model and a second set of information; generate at least a quantized representation of a latent representation of the subset of latent representations and based at least in part on at least one of a scalar quantization scheme or a vector quantization scheme, and a corresponding set of quantizer parameters; and transmit a third set of information based at least in part on one or more of the subset of latent representations or the quantized representation of the latent representation. The at least one controller may be configured to cause the processor 1004 to perform any of the various operations described herein, such as with reference to a UE 104 and/or the device 1002.

[0125] The processor 1004 may include an intelligent hardware device (e.g., a general-purpose processor, a DSP, a CPU, a microcontroller, an ASIC, an FPGA, a programmable logic device, a discrete gate or transistor logic component, a discrete hardware component, or any combination thereof). In some implementations, the processor 1004 may be configured to operate a memory array using a memory controller. In some other implementations, a memory controller may be integrated into the processor 1004. The processor 1004 may be configured to execute computer- readable instructions stored in a memory (e.g., the memory 1006) to cause the device 1002 to perform various functions of the present disclosure.

[0126] The memory 1006 may include random access memory (RAM) and read-only memory (ROM). The memory 1006 may store computer-readable, computer-executable code including instructions that, when executed by the processor 1004 cause the device 1002 to perform various functions described herein. The code may be stored in a non-transitory computer-readable medium such as system memory or another type of memory. In some implementations, the code may not be directly executable by the processor 1004 but may cause a computer (e.g., when compiled and executed) to perform functions described herein. In some implementations, the memory 1006 may include, among other things, a basic I/O system (BIOS) which may control basic hardware or software operation such as the interaction with peripheral components or devices.

[0127] The I/O controller 1010 may manage input and output signals for the device 1002. The I/O controller 1010 may also manage peripherals not integrated into the device M02. In some implementations, the I/O controller 1010 may represent a physical connection or port to an external peripheral. In some implementations, the I/O controller 1010 may utilize an operating system such as iOS®, ANDROID®, MS-DOS®, MS-WINDOWS®, OS/2®, UNIX®, LINUX®, or another known operating system. In some implementations, the I/O controller 1010 may be implemented as part of a processor, such as the processor M08. In some implementations, a user may interact with the device 1002 via the I/O controller 1010 or via hardware components controlled by the I/O controller 1010.

[0128] In some implementations, the device 1002 may include a single antenna 1012. However, in some other implementations, the device 1002 may have more than one antenna 1012 (e.g., multiple antennas), including multiple antenna panels or antenna arrays, which may be capable of concurrently transmitting or receiving multiple wireless transmissions. The transceiver 1008 may communicate bi-directionally, via the one or more antennas 1012, wired, or wireless links as described herein. For example, the transceiver 1008 may represent a wireless transceiver and may communicate bi-directionally with another wireless transceiver. The transceiver 1008 may also include a modem to modulate the packets, to provide the modulated packets to one or more antennas 1012 for transmission, and to demodulate packets received from the one or more antennas 1012.

[0129] FIG. 11 illustrates a flowchart of a method 1100 that supports encoding and decoding of input information in accordance with aspects of the present disclosure. The operations of the method 1100 may be implemented by a device or its components as described herein. For example, the operations of the method 1100 may be performed by a network entity 102 and/or a UE 104 as described with reference to FIGs. 1 through 10. In some implementations, the device may execute a set of instructions to control the function elements of the device to perform the described functions. Additionally, or alternatively, the device may perform aspects of the described functions using special-purpose hardware.

[0130] At 1102, the method may include generating, at a first apparatus, a subset of latent representations of an input data from a set of latent representations based at least in part on a first set of information of an encoder neural network model and a second set of information. The operations of 1102 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1102 may be performed by a device as described with reference to FIG. 1.

[0131] At 1104, the method may include generating at least a quantized representation of a latent representation of the subset of latent representations and based at least in part on at least one of a scalar quantization scheme or a vector quantization scheme, and a corresponding set of quantizer parameters. The operations of 1104 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1104 may be performed by a device as described with reference to FIG. 1.

[0132] At 1106, the method may include transmitting a third set of information based at least in part on one or more of the subset of latent representations or the quantized representation of the latent representation. The operations of 1106 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1106 may be performed by a device as described with reference to FIG. 1. [0133] FIG. 12 illustrates a flowchart of a method 1200 that supports encoding and decoding of input information in accordance with aspects of the present disclosure. The operations of the method 1200 may be implemented by a device or its components as described herein. For example, the operations of the method 1200 may be performed by a network entity 102 and/or a UE 104 as described with reference to FIGs. 1 through 10. In some implementations, the device may execute a set of instructions to control the function elements of the device to perform the described functions. Additionally, or alternatively, the device may perform aspects of the described functions using special-purpose hardware.

[0134] At 1202, the method may include receiving, at a first apparatus, a first set of information from a second apparatus. The operations of 1202 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1202 may be performed by a device as described with reference to FIG. 1.

[0135] At 1204, the method may include generating decoder input of a decoder neural network model based on at least one of the first set of information or a first set of parameters. The operations of 1204 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1204 may be performed by a device as described with reference to FIG. 1.

[0136] At 1206, the method may include generating output of the decoder neural network model using the decoder input and a second set of information used to determine the decoder neural network model for decoding the decoder input. The operations of 1206 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 1206 may be performed by a device as described with reference to FIG. 1.

[0137] It should be noted that the methods described herein describes possible implementations, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible. Further, aspects from two or more of the methods may be combined.

[0138] The various illustrative blocks and components described in connection with the disclosure herein may be implemented or performed with a general-purpose processor, a DSP, an ASIC, a CPU, an FPGA or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices (e.g., a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.

[0139] The functions described herein may be implemented in hardware, software executed by a processor, firmware, or any combination thereof. If implemented in software executed by a processor, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Other examples and implementations are within the scope of the disclosure and appended claims. For example, due to the nature of software, functions described herein may be implemented using software executed by a processor, hardware, firmware, hardwiring, or combinations of any of these. Features implementing functions may also be physically located at various positions, including being distributed such that portions of functions are implemented at different physical locations.

[0140] Computer-readable media includes both non-transitory computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A non-transitory storage medium may be any available medium that may be accessed by a general-purpose or special-purpose computer. By way of example, and not limitation, non-transitory computer-readable media may include RAM, ROM, electrically erasable programmable ROM (EEPROM), flash memory, compact disk (CD) ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other non-transitory medium that may be used to carry or store desired program code means in the form of instructions or data structures and that may be accessed by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor.

[0141] Any connection may be properly termed a computer-readable medium. For example, if the software is transmitted from a website, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of computer-readable medium. Disk and disc, as used herein, include CD, laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above are also included within the scope of computer-readable media.

[0142] As used herein, including in the claims, “or” as used in a list of items (e.g., a list of items prefaced by a phrase such as “at least one of’ or “one or more of’ or “one or both of’) indicates an inclusive list such that, for example, a list of at least one of A, B, or C means A or B or C or AB or AC or BC or ABC (e.g., A and B and C). Also, as used herein, the phrase “based on” shall not be construed as a reference to a closed set of conditions. For example, an example step that is described as “based on condition A” may be based on both a condition A and a condition B without departing from the scope of the present disclosure. In other words, as used herein, the phrase “based on” shall be construed in the same manner as the phrase “based at least in part on. Further, as used herein, including in the claims, a “set” may include one or more elements.

[0143] The terms “transmitting,” “receiving,” or “communicating,” when referring to a network entity, may refer to any portion of a network entity (e.g., a base station, a CU, a DU, a RU) of a RAN communicating with another device (e.g., directly or via one or more other network entities).

[0144] The description set forth herein, in connection with the appended drawings, describes example configurations and does not represent all the examples that may be implemented or that are within the scope of the claims. The term “example” used herein means “serving as an example, instance, or illustration,” and not “preferred” or “advantageous over other examples.” The detailed description includes specific details for the purpose of providing an understanding of the described techniques. These techniques, however, may be practiced without these specific details. In some instances, known structures and devices are shown in block diagram form to avoid obscuring the concepts of the described example.

[0145] The description herein is provided to enable a person having ordinary skill in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to a person having ordinary skill in the art, and the generic principles defined herein may be applied to other variations without departing from the scope of the disclosure. Thus, the disclosure is not limited to the examples and designs described herein but is to be accorded the broadest scope consistent with the principles and novel features disclosed herein.