Federated learning in a wireless communication network

Abstract

Various aspects of the present disclosure relate to a first network entity for wireless communication, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the first network entity to: send, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receive, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

Description

FEDERATED LEARNING IN A WIRELESS COMMUNICATION NETWORKTECHNICAL FIELD

[0001] The subject matter disclosed herein relates generally to the field of implementing vertical federated learning and / or vertical federated transfer learning. In particular, this document defines a first network entity for wireless communication, a processor for wireless communication, a method performed by a first network entity, a second network entity for wireless communication, a processor for wireless communication, and a method performed by a second network entity.BACKGROUND

[0002] A wireless communications system may include one or multiple network communication devices, such as base stations, which 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, or the like). 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)).SUMMARY

[0003] An article “a” before an element is unrestricted and understood to refer to “at least one” of those elements or “one or more” of those elements. The terms “a,” “at least one,” “one or more,” and “at least one of one or more” may be interchangeable. 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 aninclusive 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 (i.e., 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.

[0004] There is provided a first network entity for wireless communication, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the first network entity to: send, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receive, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0005] There is provided processor for wireless communication, comprising: at least one controller coupled with at least one memory and configured to cause the processor to: send, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receive, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0006] There is provided a method performed by a first network entity, the method comprising: sending, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receiving, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0007] There is provided a second network entity for wireless communication, comprising: at least one memory; and at least one processor coupled with the at least onememory and configured to cause the second network entity to: receive, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and send, to the first network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0008] There is provided a processor for wireless communication, comprising: at least one controller coupled with at least one memory and configured to cause the processor to: receive, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and send, to the first network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0009] There is provided a method performed by a second network entity, the method comprising: receiving, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and sending, to the first network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 illustrates an example of a wireless communications system in accordance with aspects of the present disclosure.

[0011] Figure 2 illustrates an overview of NWDAF flavours including the potential input data sources and output consumers in accordance with aspects of the present disclosure.

[0012] Figure 3a is a chart illustrating horizontal federated learning in accordance with aspects of the present disclosure.

[0013] Figure 3b is a chart illustrating vertical federated learning in accordance with aspects of the present disclosure.

[0014] Figure 3c is a chart illustrating vertical federated transfer learning in accordance with aspects of the present disclosure.

[0015] Figures 4a and 4b illustrate a signalling diagram for a method of federated learning in accordance with aspects of the present disclosure.

[0016] Figure 5 illustrates an example of a user equipment (UE) 500 in accordance with aspects of the present disclosure.

[0017] Figure 6 illustrates an example of a processor 600 in accordance with aspects of the present disclosure.

[0018] Figure 7 illustrates an example of a network equipment (NE) 700 in accordance with aspects of the present disclosure.

[0019] Figure 8 illustrates a flowchart of a method 800 performed by a NE in accordance with aspects of the present disclosure.

[0020] Figure 9 illustrates a flowchart of a method 900 performed by a NE in accordance with aspects of the present disclosure.DETAILED DESCRIPTION

[0021] Network analytics and Artificial Intelligence (AI) / Machine Learning (ML) is typically deployed in the 5G core network via a Network Data Analytics Function (NWDAF). 3GPP TS 23.288, V18.5.0, is a discussion document titled “Network Data Analytics Services” that considers the support of various analytics types. Generally, each NWDAF may support one or more Analytics identifiers (IDs). An NWDAF may have the role of inference (called NWDAF Analytics Logical Function (AnLF)), the role of training (called NWDAF MTLF) or both. For example, an AnLF that supports a specific Analytics ID inference may subscribe to a corresponding MTLF that is responsible for training.

[0022] TS 23.288 considers several analytics services identified by an Analytics ID. These analytics services rely on certain input data that may be collected and processed by an ML model residing in an NWDAF AnLF to derive the analytics output. Typically, Analytics IDs rely on at least two or more input data that is collected from different data sources. NWDAF correlates different data inputs performing complex analytics that can beused as an insight by a consumer to assist in making decisions. Analytics ID may be related to user behaviour, communication patterns, mobility, service experience, NF load, slice load, etc. The ML model in NWDAF AnLF may be trained or re-trained when its accuracy degrades.

[0023] The NWDAF MTLF is responsible for carrying out the ML model training. The ML model training can be performed by collecting the appropriate input data at a certain NWDAF MTLF or alternatively if the input data cannot be shared it may be performed in a distributed way by adopting Federated Learning (FL). In FL, an NWDAF MTLF FL Server is typically responsible for carrying out the training process by selecting NWDAF MTLF FL Clients and distributing the ML model weights or parameters, which are used by NWDAF MTLF FL Clients to perform ML model training based on locally collected data (local data set). The NWDAF MTLF FL Server collects the ML model weights or parameters from each NWDAF MTLF FL Client and integrates them into a global ML model (or global model). The NWDAF MTLF FL Server can then share the weights or parameters of the ML model to the selected NWDAF MTLF FL Clients to repeat the process until the desired ML model performance is achieved or the target training time expires.

[0024] Generally, there are three types of FL:(i) horizontal FL: which uses different datasets with the same feature space across all NWDAF MTLF FL Clients;(ii) vertical FL: which uses the same datasets with different feature space to jointly train a global model; and(iii) vertical federated transfer learning: which is a type of vertical FL utilized with a pre-trained model that is trained on a similar dataset for solving a different problem.

[0025] Vertical Federated Learning relies on the capability and availability of the respective NWDAF MTLF FL Client to perform training of specified features with local data. However, in some circumstances an NWDAF MTLF FL Client may not have the capability or availability to perform training of specified features with local data. There areother circumstances that may introduce a latency into the vertical federated learning beyond a desired limit.

[0026] Vertical federated transfer learning tends to enable multiple features (for example, multiple features that are related to the same or even different Analytics IDs and ML models) to be transferred thereby speeding up the learning process. In previous wireless communication systems, there is no indication when, under which conditions and which knowledge may be transferred between a vertical federated server and a respective set of clients.

[0027] The examples described herein generally relate to an apparatus and method for transfer learning in vertical federated learning scenarios in a wireless communication system. Vertical federated Transfer Learning (TL) may allow feature transfer among different Analytics IDs and is based on the concept that related features may be adapted to serve the needs of another problem.

[0028] Aspects of the present disclosure are described in the context of a wireless communications system.

[0029] Figure 1 illustrates an example of a wireless communications system 100 in accordance with aspects of the present disclosure. The wireless communications system 100 may include one or more NE 102, one or more UE 104, and a core network (CN) 106. 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 NR network, such as a 5G network, a 5G-Advanced (5G-A) network, or a 5G ultrawideband (5G-UWB) 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, for example, 6G. Additionally, the wireless communications system 100 may support technologies, such as time division multipleaccess (TDMA), frequency division multiple access (FDMA), or code division multiple access (CDMA), etc.

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

[0031] An NE 102 may provide a geographic coverage area for which the NE 102 may support services for one or more UEs 104 within the geographic coverage area. For example, an NE 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, an NE 102 may be moveable, for example, a satellite associated with a non-terrestrial network (NTN). In some implementations, different geographic coverage areas associated with the same or different radio access technologies may overlap, but the different geographic coverage areas may be associated with different NE 102.

[0032] The one or more UE 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 remote unit, a mobile device, a wireless device, a remote device, a subscriber device, a transmitter device, a receiver 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.

[0033] A UE 104 may be able to support wireless communication directly with other UEs 104 over a communication link. 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 vehi cl e-to- vehicle (V2V) deployments, vehicle-to-everything (V2X) deployments, or cellular-V2X deployments, the communication link 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.

[0034] An NE 102 may support communications with the CN 106, or with another NE 102, or both. For example, an NE 102 may interface with other NE 102 or the CN 106 through one or more backhaul links (e.g., SI, N2, N2, or network interface). In some implementations, the NE 102 may communicate with each other directly. In some other implementations, the NE 102 may communicate with each other or indirectly (e.g., via the CN 106. In some implementations, one or more NE 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).

[0035] The CN 106 may support user authentication, access authorization, tracking, connectivity, and other access, routing, or mobility functions. The CN 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 NE 102 associated with the CN 106.

[0036] The CN 106 may communicate with a packet data network over one or more backhaul links (e.g., via an SI, N2, N2, or another network interface). The packet data network may include an application server. In some implementations, one or more UEs 104 may communicate with the application server. A UE 104 may establish a session (e.g., a protocol data unit (PDU) session, or the like) with the CN 106 via an NE 102. The CN 106 may route traffic (e.g., control information, data, and the like) between the UE 104 and theapplication server 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 CN 106 (e.g., one or more network functions of the CN 106).

[0037] In the wireless communications system 100, the NEs 102 and the UEs 104 may use resources of the wireless communications 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 NEs 102 and the UEs 104 may support different resource structures. For example, the NEs 102 and the UEs 104 may support different frame structures. In some implementations, such as in 4G, the NEs 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 NEs 102 and the UEs 104 may support various frame structures (i.e., multiple frame structures). The NEs 102 and the UEs 104 may support various frame structures based on one or more numerologies.

[0038] 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., / t=0) may be associated with a first subcarrier spacing (e.g., 15 kHz) and a normal cyclic prefix. In some implementations, the first numerology (e.g., / t=0) associated with the first subcarrier spacing (e.g., 15 kHz) may utilize one slot per subframe. A second numerology (e.g., / / =1) may be associated with a second subcarrier spacing (e.g., 30 kHz) and a normal cyclic prefix. A third numerology (e.g., g=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., / t=3) may be associated with a fourth subcarrier spacing (e.g., 120 kHz) and a normal cyclic prefix. A fifth numerology (e.g., / t=4) may be associated with a fifth subcarrier spacing (e.g., 240 kHz) and a normal cyclic prefix.

[0039] 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.

[0040] 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. The number of slots in each subframe may also depend on the one or more numerologies supported in the wireless communications system 100. For instance, the first, second, third, fourth, and fifth numerologies (i.e., / t=0, / t=l, =2, jtz=3, =4) associated with respective subcarrier spacings of 15 kHz, 30 kHz, 60 kHz, 120 kHz, and 240 kHz may utilize a single slot per subframe, two slots per subframe, four slots per subframe, eight slots per subframe, and 16 slots per subframe, respectively. Each slot may include a number (e.g., quantity) of symbols (e.g., 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., / t=0) associated with a first subcarrier spacing (e.g., 15 kHz) may be used interchangeably between subframes and slots.

[0041] 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 NEs 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 NEs 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 NEs 102 and the UEs 104, among other equipment or devices for short-range, high data rate capabilities.

[0042] FR1 may be associated with one or multiple numerol ogies (e.g., at least three numerologies). For example, FR1 may be associated with a first numerology (e.g., / t=0), which includes 15 kHz subcarrier spacing; a second numerology (e.g., / / =1), which includes 30 kHz subcarrier spacing; and a third numerology (e.g., / / =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., / z=2), which includes 60 kHz subcarrier spacing; and a fourth numerology (e.g., / t=3), which includes 120 kHz subcarrier spacing.

[0043] Figure 2 illustrates an overview of NWDAF flavours, indicated generally by reference numeral 200, including the potential input data sources and output consumers in accordance with aspects of the present disclosure. The various NWDAF deployment flavours and their respective input data sources and the potential consumer for analytics output results are illustrated. Optionally, a first (Data Collection Coordination Function) DCCF (and / or Messaging Framework Adapter Function (MFAF)) 212 receives inputs from 5G Core Network Functions (NF) 202, Application Functions (AF) 204, untrusted AF 204 via a Network Exposure Function (NEF) 206, data repositories 208, and Operations, Administration and Maintenance (0AM) data 210.

[0044] The Data Repositories 208 may comprise a Network Repository Function (NRF), a Binding Support Function (BSF), an Analytics Data Repository Function (ADRF), a Unified Data Management (UDM), and / or a Unified Data Repository (UDR).

[0045] The 0AM 210 may comprise a Management Services (MnS) Producer or a Management Function (MF) that provides Performance Measurements (PMs), Key Performance indicators (KPIs), Configuration Management, and Alarm information. Optionally, the first DCCF 212 may provide data to an NWDAF AnLF / MTLF 214, an NWDAF AnLF 216, and an NWDAF MTLF 218, an NWDAF MTLF FL Server 219 or an NWDAF MTLF FL Client 220.

[0046] The NWD AF AnLF / MTLF 214, the NWD AF AnLF 216, the NWD AF MTLF 218, the NWDAF MTLF FL Server 219 or the NWDAF MTLF FL Client 220 may pass on data, i.e., analytics results, to the second DCCF (and / or MFAF) 222. The second DCCF 222 may further provide, e.g., distribute, such data to the 5G Core Network Functions 224, Application Functions (AFs) 228, untrusted AF 228 via a Network Exposure Function (NEF) 226, Data repositories 230, and 0AM data 232. The Data Repositories 230 may comprise an Analytics Data Repository Function (ADRF), a Unified Data Management (UDM), and / or a Unified Data Repository (UDR). The 0AM 232 may comprise a Management Services (MnS) Consumer or a Management Function (MF).

[0047] In operation, an NWDAF relies on various sources of data input including data from 5G core NFs 202, AFs 204, Data repositories 208, and 0AM data 210. An NWDAF may provide in turn analytics output results to 5G core NF 224, AFs 228, and 0AM 232. Optionally, the first DCCF 212 or second DCCF 222 may be involved to distribute and collect repeated data towards or from various data sources.

[0048] Figure 3a is a chart 300 illustrating horizontal federated learning in accordance with aspects of the present disclosure.

[0049] The chart 300 illustrates an example of horizontal FL that uses datasets (e.g., DI, D2, D3, D4, D5, D6) with the same feature space (e.g., Fl, F2, F3) across all NWDAF MTLF FL Clients (the chart 300 shows FL Client A and FL Client B as an example).

[0050] Figure 3b is a chart 310 illustrating vertical federated learning in accordance with aspects of the present disclosure.

[0051] The chart 310 illustrates an example of vertical FL (VFL) that uses the same datasets (e.g., DI, D2, D3) of different feature space (e.g., Fl, F2, F3, F4, F5, F6) across all NWDAF MTLF FL Clients (e.g., FL Client A and FL Client B) to jointly train a global model.

[0052] Figure 3c is a chart 320 illustrating vertical federated Transfer Learning (TL) in accordance with aspects of the present disclosure.

[0053] The chart 320 illustrates an example of vertical federated transfer learning (TL), which is a type of vertical FL, utilized with a pre-trained model (e.g., trained for Fl) that is trained on a similar dataset for solving a different problem.

[0054] The definition of a feature can play a critical role on how vertical FL (VFL) can be applied in the 5G core network for example considering among which NWDAF types and under which scenarios, e.g., across domains, e.g., 5G core network and application. Furthermore, in the past, there is no means for applying transfer learning in vertical federated learning scenarios in wireless communication systems, e.g., among which features and situations can transfer learning be applied and benefit vertical federated learning.

[0055] Examples described herein generally relate to an apparatus and method that applies TL in vertical FL scenarios among NWDAFs in the same 5G core domain that share the same Features ID. A Feature ID may be at least one of: an Events ID, Performance Measurements (PMs) or Key Performance Indicators (KPIs)). For example, a feature may correspond to “Network Function (NF) load”. The “NF load” feature may be used in both Analytics ID = “NF Load” and Analytics ID = “Network Performance”. In another example, a feature may correspond to “User Equipment (UE) location”. The “UE location” feature may be used in both Analytics ID = “Dispersion Analytics” and Analytics ID = “UE Mobility”.

[0056] An NWDAF MTLF FL Server may select an NWDAF MTLF FL Client among NWDAF MTLFs that are capable of transferring knowledge among ML models; for example, from a first local model to a global model. The ML models (e.g., the first local model and global model) may be related to different Analytics IDs but may share partially the same feature space for the same input data set space.

[0057] A feature may be a characteristic of collected raw data. A feature may be a property related to collected raw data. A feature may be an attribute related to collected raw data.

[0058] A feature may be a numerical feature. A numerical feature may hold numerical values that can be measured on a specific scale. For example, a feature may be a Network Function (NF) load or number of handovers.

[0059] A feature may be a categorical feature. A categorical feature may comprise discrete values that can be grouped in categories. For example, a categorical feature may be Quality of Experience (QoE) levels.

[0060] A feature may be time series features. Time series features may comprise measurements taken over different times. For example, time series features may be peak and off-peak times.

[0061] A feature may be text or evaluation features. Text or evaluation features may represent words. For example, text or evaluation features may be reviews or feedback related to a service.

[0062] A feature may be identified by an identifier of the feature. An identifier of the feature may be a feature ID.

[0063] In the context of VFL and / or Vertical Federated Transfer Learning (TL) a Feature ID may represent one or more of the following:• An event ID. An event ID may represent an event. For example, an event may be when a UE enters an area of interest. Another example may be when a thresholdis triggered.• A Performance Measurement (PM) or Key Performance Indicator (KPI). A PM or KPI may be related to specific types of measurements. For example, a PM or KPI may relate to NF load. Another example of a PM or KPI may relate to end-to-end latency. PMs or KPIs provided by different domains may correspond to different Feature IDs. For example, a UE throughput PM or KPI in the 5G core domain may be provided per Subscription Permanent Identifier (SUPI) (corresponding to a UE ID), while the UE throughput PM or KPI in the Operations, Administration and Maintenance (0AM) domain may be provided on UE average at a particular cell.• An Analytics ID. A first Analytics ID may be an input for a second Analytics ID. For example, a first Analytics ID may be “UE Mobility” which may be used as an input for a second Analytics ID which may be “Network Performance”.• Time range for collecting data. For example, the time range for data collected at daytime or night.• UE types. For example, a UE type may be a mobile user, a car, or a drone.• User mobility profile. For example, a mobility profile may be classified as high, medium, or low speed or static. A mobility profile may comprise direction and / or speed.• User contract profiles. User contract profiles may consider context information. For example, user contract profiles may include gold, silver, etc.• Mean Opinion Score (MOS) related to application performance. For example, MOS may be from 1 to 5.• Load or congestion level. For example, the load or congestion level may be classified as low, medium or high.• Quality of Service (QoS) / Quality of Experience (QoE) levels. For example, QoS / QoE levels may be classified as low, medium, high.• User experience feedback in relation to an application. For example, user experience feedback may comprise text or a rating.

[0064] Features can be adapted from a trained ML model based on one set of labels to a new / modified set of labels. An NWDAF MTLF FL Client may perform transfer learning per feature by directly exchanging information (or knowledge) from the ML model (e.g., from a first local model) to the NWDAF MTLF FL Server. The information (or knowledge transfer) from the ML model may comprise one or more of weights, parameters, or ML model transfer.

[0065] The NWDAF MTLF FL Client may perform transfer learning by providing a set of rules for providing guidance on how the information (or knowledge) may be transferredfrom the ML model (or the first local model) to the global model. The set of rules may comprise at least one of the following:• a weight that can assist to adapt the information (or knowledge) from a first Analytics ID to a second Analytics ID, wherein the first Analytics ID is different to the second Analytics ID;• an indication of parts of the information (or knowledge) that may be transferred between the first Analytics ID to the second Analytics ID; or• an indication of rules for relating a first feature to a second feature. For example, a rule may relate mobility information to an increase of load in an area of interest; for example, for Analytics ID = “Network Performance”.

[0066] The NWDAF MTLF FL Server may select TL for specified Feature ID(s) for each NWDAF MTLF FL Client by considering at least one of the following criteria:• a timestamp of the last training process. The timestamp may indicate the “freshness” of the weights of a particular feature;• a feature inter-related factor, that may include at least one of the following cases: for non-identical features, a first feature may be used to create a second feature. The first feature may be different to the second feature. For example, a UE mobility information feature may be related to a network load feature. In another example, the UE mobility information feature may be related to Network Function (NF) load. Therefore, the UE mobility information feature may be used to create the network load feature or the NF load feature; for non-identical features, the second feature or set of second features may be created based on a logical relation with the first feature. Alternatively, the second feature or set of second features may be created based on rules among different domains. A domain may be a geographical area. A domain may be a 5G core. A domain may be a network management domain. Adomain may be an application domain. For example, a logical relation may be a relation between network congestion and UE service experience; or for identical features, the first feature may be the same as the second feature. However, the first feature may be trained using a first input data set and the second feature may be trained using a second input data set. The first input data set may have similar data statistics to the second input data set. The first feature set may comprise a similarity element with the second feature. The similarity element may be one or more of a weight, an adaptation factor or a loss distribution. For example, the first feature and second feature may correspond to NF load; however, the first feature and second feature may be trained on different datasets corresponding to different areas of interest.• vendor interoperability considering specific features and / or ML model information;• an indication of input data availability and / or time availability. A negative indication may favour the TL without necessarily meaning that TL would be the choice since other parameters needs to be considered.

[0067] One or more NWDAF MTLF FL Clients may register, with a Network Repository Function (NRF), a capability to support VFL and / or Vertical Federated transfer learning (TL). The one or more NWDAF MTLF FL Clients may send an indication of features that are available for TL to the NRF. The one or more NWDAF MTLF FL Clients may send an indication, e.g., of Event IDs, PMs or KPIs available for transfer learning to the NRF. The registration of one or more NWDAF MTLF FL Clients with the NRF may be configured by an 0AM. The registration of one or more NWDAF MTLF FL Clients with the NRF may be configured by the respective NWDAF MTLF FL Client.

[0068] TL may include an indication of how the knowledge may be transferred including: (i) a weight (for example, a weight having values between the set [0,1]) for adapting a trained feature from one Analytics ID to another; or (ii) an indication of specific objects of interest related to a feature (for example, a UE ID, a group UE ID, a Packet Data Unit (PDU) session ID or Quality of Service (QoS) Flow ID; or (iii) an indication of rules that relate to specific features and / or knowledge form one domain to another.

[0069] Once discovered, an NWDAF MTLF FL Server may check a time availability for a Feature ID. The NWDAF MTLF FL Server may indicate a desired usage of the Feature ID for TL. For example, the NWDAF MTLF FL Server may indicate the desired usage of the Feature ID for TL using an indicator. In another example, the NWDAF MTLF FL Server may indicate the desired usage of the Feature ID for TL using a flag.

[0070] For TL in VFL, a FL preparation request message from the NWDAF MTLF FL Server to one or more NWDAF(s) MTLF FL Clients may be aNnwdaf MLModelTraining Subscribe message. For TL in VFL, the FL preparation request message from the NWDAF MTLF FL Server to one or more NWDAF(s) MTLF FL Clients may be a Nnwdaf MLModelTraininglnfo Request message. The FL preparation request message may include information related to the selected Feature ID(s) for transferring knowledge. The FL preparation request message may include an indication of the adaptation weights. The FL preparation request message may include an indication of target objects of interest. The FL preparation request message may include an indication of certain rules related to certain features.

[0071] The one or more NWDAF(s) MTLF FL Clients may be a labels owner that supports VFL. The labels of the NWDAF MTLF FL Server and NWDAF(s) MTLF FL Clients may be the same. For example, the NWDAF MTLF Server may request a pretrained Feature ID from a ML model associated with the same Analytics ID. The labels of the NWDAF MTLF FL Server and NWDAF(s) MTLF FL Clients may be different. For example, the NWDAF MTLF Server may request a pre-trained Feature ID from the ML model associated with a different Analytics ID.

[0072] The information related to TL for vertical FL can be added to describe the ML model content for the purpose of training in clause 6.2F.2 TS 23.288.

[0073] Embodiments of the proposed enhancements may include the NWDAF FL: (i) registration phase, (ii) discovery phase, (iii) preparation phase and (iv) execution phase. An overview of the registration, discovery, preparation, and execution procedure for VFL and transfer VFL is illustrated in Figure 3.

[0074] Figure 4a illustrates a first part of a signalling diagram 400 for a method of federated learning in accordance with aspects of the present disclosure.

[0075] The signalling diagram 400 comprises an NWDAF MTLF FL Server 410, a first NWDAF MTLF FL Client 422, N NWDAF MTLF FL Client(s) 424 and an NRF 430.

[0076] The method comprises steps 471 to 485. Figure 4a illustrates steps 471 to 476.

[0077] Steps 471 to 473 may correspond to an NWDAF registration procedure 440.

[0078] In steps 471 to 473, the NWDAF MTLF FL Server 410 or NWDAF MTLF FL Client 422 / 424 registers its NF profile with the NRF 430. The NF profile may comprise at least one of an NWDAF NF Type (see clause 5.2.7.2.2 of TS 23.502), Analytics ID(s), Address information of NWDAF, Service Area, FL capability type information (e.g., FL server or FL client) and / or a Time interval for supporting FL (and hence TL). The NF profile may further comprise one or more Feature IDs (for example, an Event IDs, PM or KPI) and / or the respective TL capabilities (for example, one or more weights, objects of interest and / or rules).

[0079] The NWDAF MTLF FL Server 410 may serve vertical FL scenarios and additionally (vertical federated) TL.

[0080] Steps 474 to 476 may correspond to an NWDAF Discovery procedure 450.

[0081] In steps 474 to 476, the NWDAF MTLF FL Server 410 may determine that the residing ML model requires FL based on an operator policy. For example, the operator policy may comprise at least one of a pre-configured list of ML models, Analytic ID and / or Service Area / DNAI. The desired data may not be obtainable directly from a data producer NF (e.g., due to privacy reasons). Alternatively, the NWDAF MTLF FL Server 410 may receive a request from another entity that the residing ML model requires FL based on the operator policy.

[0082] The NWDAF MTLF FL Server 410 may not have an option for horizontal FL. Alternatively, or additionally, the NWDAF MTLF FL Server 410 may have a requirement to combine vertical FL and vertical federated TL. Therefore, the NWDAF MTLF FL Server410 may discover and select other NWDAF(s) MTLF FL Clients 422 / 424 from NRF 430 by invoking the Nnrf_NFDiscovery_Request service operation (or message).

[0083] The Nnrf NFDiscovery Request service operation (or message) may comprise at least one of: an Analytic ID of the ML model required, FL capability Type (e.g., NWDAF MTLF FL client 422 / 424), Service Area, NF type(s) of data sources from which the NWDAF MTLF FL Client 422 / 424 is able to collect data for local model training, Time Period of Interest, ML Model and Feature Interoperability Indicator as well as Feature IDs (Event IDs, PM, KPIs), timestamp related to the last re-training process, and the respective TL capabilities (weights, objects of interest and rules).

[0084] Figure 4b illustrates a second part of a signalling diagram 400 for a method of federated learning in accordance with aspects of the present disclosure.

[0085] Steps 477 to 479 may correspond to an NWDAF FL preparation procedure.

[0086] In steps 477 to 479, the NWDAF MTLF FL Server 410 sends a FL preparation request to the NWDAF MTLF FL Client(s) 422 / 424. The FL preparation request may comprise a Nnwdaf MLModelTraining Subscribe orNnwdaf MLModelTraininglnfo Request service operation (or message) with the ML Preparation Flag, to determine if the NWDAF MTLF FL Client(s) 422 / 424 can meet the ML model training requirement (or training requirement).

[0087] The ML model training requirement may include at least one of: an Analytics ID, ML Model and Feature Interoperability information, Available data requirement, Availability time requirement (time span needed for the FL process), TL Feature ID(s), TL Feature adaptation capabilities and timestamp indicating how “fresh” is the current retraining.

[0088] Each NWDAF MTLF FL Client 422 / 424 checks if it can meet the ML model TL training requirement and decides whether to join the VFL process based on an operator policy and / or implementation. Then each NWDAF MTLF FL Client 422 / 424 invokes Nnwdaf_MLModelTraining_Notify or Nnwdaf_MLModelTraining_Subscribe response service operation or Nnwdaf MLModelTraininglnfo Request response service operation toindicate whether it will join the VFL procedure and may include the reason in the response message if it cannot join.

[0089] The FL preparation procedure may be skipped if the NWDAF MTLF FL Server 410 can decide which NWDAF MTLF FL Client(s) 422 / 424 may support the VFL procedure to be performed. For example, the NWDAF MTLF FL Server 410 may decide which NWDAF MTLF FL Client(s) 422 / 424 may support the VFL procedure to be performed based on other obtained information or based on local configuration.

[0090] Steps 480 to 485 may correspond to a NWDAF execution 460 procedure. The NWDAF Execution 460 procedure may be repeated until the time or performance criteria are met.

[0091] In step 480, the NWDAF MTLF FL Server 410 determines the final list of NWDAF MTLF FL Clients 422 / 424 and their roles (for example, the role may be VFL or vertical federated TL. This may be determined based on at least one of: a timestamp related to a “fresh” training; an assessment of the applicability of TL capabilities; and / or a time availability for completing the training process of NWDAF MTLF FL Client(s) 422 / 424 to be involved in the VFL procedures based on the information received and other information (if available).

[0092] If any NWDAF MTLF FL Client 422 / 424 abandons the VFL process for any reason, the NWDAF MTLF FL Server 410 may select another NWDAF MTLF FL Client 422 / 424 from the list or it may perform all or part of the preparation procedure again. For example, the NWDAF MTLF FL Server 410 may perform steps 487 to 489 again, to identify new NWDAF MTLF FL Client(s) 422 / 424.

[0093] In steps 481 to 483, the NWDAF MTLF FL Server 410 sends aNnwdaf MLModelTraining Subscribe or Nnwdaf MLModelTraininglnfo Request service operation (or message) to the selected NWDAF MTLF FL Client(s) 422 / 424. The selected NWDAF MTLF FL Client(s) 422 / 424 participates in the FL and may indicate (for example, by means of a flag) whether they need to perform: (i) VFL including local model training to determine the interim local ML model information based on the input VFL model parameters, Feature ID, considering time schedule requirements or (ii) TL based on theinput VFL model parameters, Feature ID in corelation with TL capabilities and time schedule requirements.

[0094] Each NWDAF MTLF FL Client 422 / 424 either: (i) in VFL further trains the ML model based on its own data and reports the interim local ML model information per indicated Feature ID or (ii) in TL it reports the indicated trained Feature IDs related information. In both cases, the reporting may use the Nnwdaf MLModelTraining Notify or Nnwdaf MLModelTraininglnfo Request response services (or messages). These messages may also include (for example, in the case of the VFL) the status report of VFL training that includes a local ML model metric (for example, accuracy) and Training Input Data Information.

[0095] The NWDAF MTLF FL Client 422 / 424 may not be able to complete the VFL training of the interim local ML model within the maximum response time provided. The NWDAF MTLF FL Client 422 / 424 may send the Delay Event Notification that includes the delay event indication, an optional cause code (e.g., training failure, more time necessary) and the expected time to complete the training.

[0096] In step 484, the NWDAF MTLF FL Server 410 aggregates all the local ML model Feature information and / or TL related to specified Feature IDs to update the global ML model (or global model). The NWDAF MTLF FL Server 410 may then compute a global metric. For example, the global metric may be accuracy. The NWDAF MTLF FL Server 410 may update the global ML model every time it receives local ML model information. The NWDAF MTLF FL Server 410 may perform an update when all NWDAF MTLF FL clients 422 / 424 provide local updates and transfer knowledge. If a maximum time is provided to perform VFL training, then the NWDAF MTLF FL Server 410 may determine how to deal with the NWDAF MTLF FL clients 422 / 424 that have not provided a response within that time limit.

[0097] In step 485, based on the subscription request sent from a VFL consumer (which may contain an upper bound of the training time or a performance target, e.g., confidence or accuracy levels), the NWDAF MTLF FL Server 410 may update or terminate the current VFL training process. Alternatively, the NWDAF execution 460 procedure may be repeated.

[0098] Figure 5 illustrates an example of a UE 500 in accordance with aspects of the present disclosure. The UE 500 may include a processor 502, a memory 504, a controller 506, and a transceiver 508. The processor 502, the memory 504, the controller 506, or the transceiver 508, or various combinations thereof or various components thereof may be examples of means for performing various aspects of the present disclosure as described herein. These components may be coupled (e.g., operatively, communicatively, functionally, electronically, electrically) via one or more interfaces.

[0099] The processor 502, the memory 504, the controller 506, or the transceiver 508, or various combinations or components thereof may be implemented in hardware (e.g., circuitry). The hardware may include a processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or other programmable logic device, or any combination thereof configured as or otherwise supporting a means for performing the functions described in the present disclosure.

[0100] The processor 502 may include an intelligent hardware device (e.g., a general- purpose processor, a DSP, a CPU, an ASIC, an FPGA, or any combination thereof). In some implementations, the processor 502 may be configured to operate the memory 504. In some other implementations, the memory 504 may be integrated into the processor 502. The processor 502 may be configured to execute computer-readable instructions stored in the memory 504 to cause the UE 500 to perform various functions of the present disclosure.

[0101] The memory 504 may include volatile or non-volatile memory. The memory 504 may store computer-readable, computer-executable code including instructions when executed by the processor 502 cause the UE 500 to perform various functions described herein. The code may be stored in a non-transitory computer-readable medium such the memory 504 or another type of memory. 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.

[0102] In some implementations, the processor 502 and the memory 504 coupled with the processor 502 may be configured to cause the UE 500 to perform one or more of thefunctions described herein (e.g., executing, by the processor 502, instructions stored in the memory 504). For example, the processor 502 may support wireless communication at the UE 500 in accordance with examples as disclosed herein.

[0103] The controller 506 may manage input and output signals for the UE 500. The controller 506 may also manage peripherals not integrated into the UE 500. In some implementations, the controller 506 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some implementations, the controller 506 may be implemented as part of the processor 502.

[0104] In some implementations, the UE 500 may include at least one transceiver 508. In some other implementations, the UE 500 may have more than one transceiver 508. The transceiver 508 may represent a wireless transceiver. The transceiver 508 may include one or more receiver chains 510, one or more transmitter chains 512, or a combination thereof.

[0105] A receiver chain 510 may be configured to receive signals (e.g., control information, data, packets) over a wireless medium. For example, the receiver chain 510 may include one or more antennas for receive the signal over the air or wireless medium. The receiver chain 510 may include at least one amplifier (e.g., a low-noise amplifier (LN A)) configured to amplify the received signal. The receiver chain 510 may include at least one demodulator configured to demodulate the receive signal and obtain the transmitted data by reversing the modulation technique applied during transmission of the signal. The receiver chain 510 may include at least one decoder for decoding the processing the demodulated signal to receive the transmitted data.

[0106] A transmitter chain 512 may be configured to generate and transmit signals (e.g., control information, data, packets). The transmitter chain 512 may include at least one modulator for modulating data onto a carrier signal, preparing the signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes like phase-shift keying (PSK) or quadrature amplitude modulation (QAM). The transmitter chain 512 may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level suitablefor transmission over the wireless medium. The transmitter chain 512 may also include one or more antennas for transmitting the amplified signal into the air or wireless medium.

[0107] Figure 6 illustrates an example of a processor 600 in accordance with aspects of the present disclosure. The processor 600 may be an example of a processor configured to perform various operations in accordance with examples as described herein. The processor 600 may include a controller 602 configured to perform various operations in accordance with examples as described herein. The processor 600 may optionally include at least one memory 604, which may be, for example, an L1 / L2 / L3 cache. Additionally, or alternatively, the processor 600 may optionally include one or more arithmetic-logic units (ALUs) 606. One or more of 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).

[0108] The processor 600 may be a processor chipset and include a protocol stack (e.g., a software stack) executed by the processor chipset to perform various operations (e.g., receiving, obtaining, retrieving, transmitting, outputting, forwarding, storing, determining, identifying, accessing, writing, reading) in accordance with examples as described herein. The processor chipset may include one or more cores, one or more caches (e.g., memory local to or included in the processor chipset (e.g., the processor 600) or other memory (e.g., random access memory (RAM), read-only memory (ROM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), static RAM (SRAM), ferroelectric RAM (FeRAM), magnetic RAM (MRAM), resistive RAM (RRAM), flash memory, phase change memory (PCM), and others).

[0109] The controller 602 may be configured to manage and coordinate various operations (e.g., signalling, receiving, obtaining, retrieving, transmitting, outputting, forwarding, storing, determining, identifying, accessing, writing, reading) of the processor 600 to cause the processor 600 to support various operations in accordance with examples as described herein. For example, the controller 602 may operate as a control unit of the processor 600, generating control signals that manage the operation of various components of the processor 600. These control signals include enabling or disabling functional units, selecting data paths, initiating memory access, and coordinating timing of operations.

[0110] The controller 602 may be configured to fetch (e.g., obtain, retrieve, receive) instructions from the memory 604 and determine subsequent instruction(s) to be executed to cause the processor 600 to support various operations in accordance with examples as described herein. The controller 602 may be configured to track memory address of instructions associated with the memory 604. The controller 602 may be configured to decode instructions to determine the operation to be performed and the operands involved. For example, the controller 602 may be configured to interpret the instruction and determine control signals to be output to other components of the processor 600 to cause the processor 600 to support various operations in accordance with examples as described herein. Additionally, or alternatively, the controller 602 may be configured to manage flow of data within the processor 600. The controller 602 may be configured to control transfer of data between registers, arithmetic logic units (ALUs), and other functional units of the processor 600.[OHl] The memory 604 may include one or more caches (e.g., memory local to or included in the processor 600 or other memory, such RAM, ROM, DRAM, SDRAM, SRAM, MRAM, flash memory, etc. In some implementations, the memory 604 may reside within or on a processor chipset (e.g., local to the processor 600). In some other implementations, the memory 604 may reside external to the processor chipset (e.g., remote to the processor 600).

[0112] The memory 604 may store computer-readable, computer-executable code including instructions that, when executed by the processor 600, cause the processor 600 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. The controller 602 and / or the processor 600 may be configured to execute computer-readable instructions stored in the memory 604 to cause the processor 600 to perform various functions. For example, the processor 600 and / or the controller 602 may be coupled with or to the memory 604, the processor 600, the controller 602, and the memory 604 may be configured to perform various functions described herein. In some examples, the processor 600 may include multiple processors and the memory 604 may include multiple memories. One or more of the multiple processors may be coupled with one or more of the multiplememories, which may, individually or collectively, be configured to perform various functions herein.

[0113] The one or more ALUs 606 may be configured to support various operations in accordance with examples as described herein. In some implementations, the one or more ALUs 606 may reside within or on a processor chipset (e.g., the processor 600). In some other implementations, the one or more ALUs 606 may reside external to the processor chipset (e.g., the processor 600). One or more ALUs 606 may perform one or more computations such as addition, subtraction, multiplication, and division on data. For example, one or more ALUs 606 may receive input operands and an operation code, which determines an operation to be executed. One or more ALUs 606 be configured with a variety of logical and arithmetic circuits, including adders, subtractors, shifters, and logic gates, to process and manipulate the data according to the operation. Additionally, or alternatively, the one or more ALUs 606 may support logical operations such as AND, OR, exclusive-OR (XOR), not-OR (NOR), and not- AND (NAND), enabling the one or more ALUs 606 to handle conditional operations, comparisons, and bitwise operations.

[0114] The processor 600 may support wireless communication in accordance with examples as disclosed herein. The processor 600 may be configured to support a means for sending, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model and receiving, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0115] Alternatively, the processor 600 may be configured to support a means for receiving, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and sending, to the first network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0116] Figure 7 illustrates an example of a NE 700 in accordance with aspects of the present disclosure. The NE 700 may include a processor 702, a memory 704, a controller706, and a transceiver 708. The processor 702, the memory 704, the controller 706, or the transceiver 708, or various combinations thereof or various components thereof may be examples of means for performing various aspects of the present disclosure as described herein. These components may be coupled (e.g., operatively, communicatively, functionally, electronically, electrically) via one or more interfaces.

[0117] The processor 702, the memory 704, the controller 706, or the transceiver 708, or various combinations or components thereof may be implemented in hardware (e.g., circuitry). The hardware may include a processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or other programmable logic device, or any combination thereof configured as or otherwise supporting a means for performing the functions described in the present disclosure.

[0118] The processor 702 may include an intelligent hardware device (e.g., a general- purpose processor, a DSP, a CPU, an ASIC, an FPGA, or any combination thereof). In some implementations, the processor 702 may be configured to operate the memory 704. In some other implementations, the memory 704 may be integrated into the processor 702. The processor 702 may be configured to execute computer-readable instructions stored in the memory 704 to cause the NE 700 to perform various functions of the present disclosure.

[0119] The memory 704 may include volatile or non-volatile memory. The memory 704 may store computer-readable, computer-executable code including instructions when executed by the processor 702 cause the NE 700 to perform various functions described herein. The code may be stored in a non-transitory computer-readable medium such the memory 704 or another type of memory. 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.

[0120] In some implementations, the processor 702 and the memory 704 coupled with the processor 702 may be configured to cause the NE 700 to perform one or more of the functions described herein (e.g., executing, by the processor 702, instructions stored in the memory 704). For example, the processor 702 may support wireless communication at theNE 700 in accordance with examples as disclosed herein. The NE 700 may be configured to support a means for sending, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model and receiving, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0121] Alternatively, the NE 700 may be configured to support a means for receiving, from a first network entity, a first request message, the first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and sending, to the first network entity, a first response message comprising a knowledge transfer for supporting the first network entity with training of the global model for the first feature.

[0122] The controller 706 may manage input and output signals for the NE 700. The controller 706 may also manage peripherals not integrated into the NE 700. In some implementations, the controller 706 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some implementations, the controller 706 may be implemented as part of the processor 702.

[0123] In some implementations, the NE 700 may include at least one transceiver 708. In some other implementations, the NE 700 may have more than one transceiver 708. The transceiver 708 may represent a wireless transceiver. The transceiver 708 may include one or more receiver chains 710, one or more transmitter chains 712, or a combination thereof.

[0124] A receiver chain 710 may be configured to receive signals (e.g., control information, data, packets) over a wireless medium. For example, the receiver chain 710 may include one or more antennas for receive the signal over the air or wireless medium. The receiver chain 710 may include at least one amplifier (e.g., a low-noise amplifier (LN A)) configured to amplify the received signal. The receiver chain 710 may include at least one demodulator configured to demodulate the receive signal and obtain the transmitted data by reversing the modulation technique applied during transmission of the signal. The receiver chain 710 may include at least one decoder for decoding the processing the demodulated signal to receive the transmitted data.

[0125] A transmitter chain 712 may be configured to generate and transmit signals (e.g., control information, data, packets). The transmitter chain 712 may include at least one modulator for modulating data onto a carrier signal, preparing the signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes like phase-shift keying (PSK) or quadrature amplitude modulation (QAM). The transmitter chain 712 may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level suitable for transmission over the wireless medium. The transmitter chain 712 may also include one or more antennas for transmitting the amplified signal into the air or wireless medium.

[0126] Figure 8 illustrates a flowchart of a method 800 in accordance with aspects of the present disclosure. The operations of the method may be implemented by a NE as described herein. In some implementations, the NE may execute a set of instructions to control the function elements of the NE to perform the described functions.

[0127] At 802, the method 800 may include sending, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model. The operations of 802 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 802 may be performed by an NE as described with reference to Figure 7.

[0128] At 804, the method 800 may include receiving, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature. The operations of 804 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 804 may be performed by an NE as described with reference to Figure 7.

[0129] It should be noted that the method 800 described herein describes a possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0130] Figure 9 illustrates a flowchart of a method 900 in accordance with aspects of the present disclosure. The operations of the method 900 may be implemented by a NE as described herein. In some implementations, the NE may execute a set of instructions to control the function elements of the NE to perform the described functions.

[0131] At 902, the method 900 may include receiving, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model. The operations of 902 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 902 may be performed by a NE as described with reference to Figure 7.

[0132] At 904, the method 900 may include sending, to the first network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

[0133] The operations of 900 may be performed in accordance with examples as described herein. In some implementations, aspects of the operations of 900 may be performed by a NE as described with reference to Figure 7.

[0134] It should be noted that the method 900 described herein describes a possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0135] There is provided a first network entity for wireless communication, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the first network entity to: send, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receive, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature. Such a first network entity tends to reduce latency in training of the global model.

[0136] The first response message may comprise a knowledge transfer for supporting the first network entity with training of the global model for the first feature.

[0137] The first network entity may be a first analytics network function. The first network entity may be a Network Data Analytics Function, NWDAF. The first network entity may be a Model Training Logical Function, MTLF. The first network entity may be a Federated Learning, FL, server. The first network entity may be a NWDAF MTLF FL Server.

[0138] The second network entity may be a second analytics network function. The second network entity may be a Network Data Analytics Function, NWDAF. The second network entity may be a Model Training Logical Function, MTLF. The second network entity may be a Federated Learning, FL, Client. The second network entity may be a NWDAF MTLF FL Client.

[0139] The first request message may be a Nnwdaf MLModelTraining Subscribe message. The first request message may be a Nnwdaf MLModelTraininglnfo Request message. The first request message may further comprise an identifier of the first feature.

[0140] The knowledge transfer may comprise information from the first local model trained using a local data set for the first feature for vertical federated learning of the global model for the first feature. The knowledge transfer may comprise the first local model for transfer learning of the global model for the first feature.

[0141] The knowledge transfer may comprise any information relating to the first local model for the first feature. The knowledge transfer may comprise raw data from the first local model for the first feature. The knowledge transfer may comprise updates from the first local model for the first feature. The knowledge transfer may comprise gradients from the first local model for the first feature. The knowledge transfer may comprise parameters from the first local model for the first feature. The knowledge transfer may comprise information about how the first local model's performance for the first feature. The knowledge transfer may comprise information relating to how the first local model’s performance for the first feature can be improved based on its local data set. The knowledge transfer may be used to update the global model for the first feature.

[0142] Transfer learning may comprise transfer of knowledge from the first local model to the global model. Transfer learning may comprise aggregation of knowledge from the first local model with the global model.

[0143] The knowledge transfer may comprise transfer learning information. The transfer learning information may comprise at least one of: weights, parameters or transfer rules. The transfer learning information may comprise weights from the first local model trained on the respective local data set for the first feature. The transfer learning information may comprise parameters from the first local model trained on the respective local data set for the first feature. The transfer learning information may comprise transfer rules from the first local model trained on the respective local data set for the first feature.

[0144] The global model may be a global Machine Learning (ML) model. The global model may be a machine learning model that is trained across one or more second network entities. Each second network entity may hold different features of data that is vertically partitioned. The global model may be trained in a collaborative manner across the one or more second network entities. The one or more second network entities may learn from each other's data without sharing it directly.

[0145] The global model may be used to perform analytics identified by a first analytics identifier, ID. The first local model may perform analytics identified by a second analytics ID. The first analytics ID may be different to the second analytics ID. The transfer learning information may comprise a weight for adapting knowledge from the first analytics ID to the second analytics ID. The transfer learning information may comprise a weight for adapting knowledge from the second analytics ID to the first analytics ID.

[0146] Each of the one or more second network entities may comprise a first local model. The first local model may be a machine learning model. Each of the one or more second network entities may comprise a local data set. The first local model may be trained locally on the respective local data set. The local data set may be vertically partitioned data.

[0147] The global model may aggregate knowledge from each of the one or more second network entities local data set. Each first local model may be trained independently on the respective local data set. Each of the one or more second network entities may trainits own first local model using its respective local data set without sharing the raw data with another of the one or more second network entities.

[0148] Aggregation of knowledge from each of the one or more second network entities local data set may comprise model averaging or model weighting.

[0149] The at least one processor coupled with the at least one memory may be further configured to cause the first network entity to: send, to a network repository function, a second request message for discovering the second network entity, wherein the second request message comprises an identifier of the first feature. The second request message may further comprise a flag that identifies support for VFL and / or (Vertical Federated) TL. The second request message may further comprise an identifier related to training capabilities of the first local model. The second request message may be a Nnrf_NFDiscovery_Request message.

[0150] The at least one processor coupled with the at least one memory is further configured to cause the first network entity to: receive, from the network repository function, a response to the second request message, wherein the response to the second request message comprises one or more identifiers of one or more second network entities.

[0151] The one or more second network entity(s) may be capable of supporting the first network entity with transfer learning of the global model for the first feature. The one or more second network entity(s) may be capable of supporting the first network entity with vertical federated learning of the global model for the first feature. The one or more second network entity(s) may be capable of supporting the first network entity with vertical federated learning of the global model for the first feature and / or transfer learning of the global model for the first feature. The response to the second request message may be a Nnrf_NFDiscovery_Request Response message.

[0152] The at least one processor coupled with the at least one memory may be further configured to cause the first network entity to: send, to the second network entity, a third request message comprising a training requirement.

[0153] The third request message may be a federated learning preparation request. The third request message may be a Nnwdaf MLModelTraining Subscribe message. The third request message may be a Nnwdaf MLModelTraininglnfo Request message.

[0154] The training requirement may comprise an identifier of the first feature. The training requirement may comprise a machine learning, ML, preparation flag. The training requirement may be a ML model training requirement.

[0155] The training requirement may comprise an indication for the second network entity to support the first network entity with training of the global model for the first feature using the first local model by vertical federated learning. The training requirement may comprise an indication for the second network entity to support the first network entity with training of the global model for the first feature using the first local model by transfer learning.

[0156] The at least one processor coupled with the at least one memory may be further configured to cause the first network entity to: receive, from the second network entity, a response to the third request message comprising an indication whether the second network entity can support the training requirement.

[0157] The response to the third request message may be a Nnwdaf_MLModelTraining_Notify message. The response to the third request message may be a Nnwdaf MLModelTraining Subscribe message. The response to the third request message may be a Nnwdaf MLModelTraininglnfo Request message.

[0158] The response to the third request message may comprise an indication that the second network entity will support the first network entity with training of the global model for the first feature using the first local model by vertical federated learning. The response to the third request message may comprise an indication that the second network entity will support the first network entity with training of the global model for the first feature using the first local model by transfer learning.

[0159] The knowledge transfer may comprise information for adapting the first feature to create a second feature. The first feature may be identical to the second feature. The second feature may be different to first feature. The information for adapting the firstfeature to create a second feature may comprise at least one of: a weight with values for adapting a trained feature from a first analytics ID to a second analytics ID; an indication of specific objects of interest related to the first feature and the second feature; or an indication of rules that relate to the first feature and second feature.

[0160] The at least one processor coupled with the at least one memory may be further configured to cause the first network entity to: select the second feature based on at least one of: a timestamp indicating the freshness of information relating to the first feature; information for adapting the first feature to create the second feature; and / or vendor interoperability information for adapting the first feature to create the second feature.

[0161] The first feature may comprise at least one of: an event ID; a performance measurement; a key performance indicator; an analytics ID; a time range for collecting data; a user equipment type; a user mobility profile; a user contract profile; a mean opinion score related to application performance; a load level; a congestion level; a quality of service level; a quality of experience level; or user experience feedback in relation to an application.

[0162] There is provided processor for wireless communication, comprising: at least one controller coupled with at least one memory and configured to cause the processor to: send, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receive, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature. Such a processor tends to reduce latency in training of the global model.

[0163] The first response message may comprise a knowledge transfer for supporting the first network entity with training of the global model for the first feature.

[0164] There is provided a method performed by a first network entity, the method comprising: sending, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receiving, from the second network entity, a first response message comprising a knowledge transfer for training of theglobal model for the first feature. Such a method tends to reduce latency in training of the global model.

[0165] The knowledge transfer may comprise information from the first local model trained using local data for the first feature for vertical federated learning of the global model for the first feature. The knowledge transfer may comprise the first local model for transfer learning of the global model for the first feature.

[0166] Transfer learning may comprise transfer of knowledge from the first local model to the global model. Transfer learning may comprise aggregation of knowledge from the first local model with the global model.

[0167] The knowledge transfer may further comprise transfer learning information. The transfer learning information may comprise at least one of: weights, parameters or transfer rules. The global model may perform analytics identified by a first analytics identifier, ID. The first local model may perform analytics identified by a second analytics ID. The first analytics ID may be different to the second analytics ID. The transfer learning information may comprise a weight for adapting knowledge from the first analytics ID to the second analytics ID. The transfer learning information may comprise a weight for adapting knowledge from the second analytics ID to the first analytics ID.

[0168] The method may further comprise sending, to a network repository function, a second request message for discovering the second network entity, wherein the second request message comprises an identifier of the first feature. The second request message may further comprise a flag that identifies support for VFL and / or (Vertical Federated) TL. The second request message may further comprise an identifier related to training capabilities of the first local model. The second request message may be a Nnrf_NFDiscovery_Request message.

[0169] The method may further comprise receiving, from the network repository function, a response to the second request message, wherein the response to the second request message comprises one or more identifiers of one or more second network entities.

[0170] The one or more second network entity(s) may be capable of supporting the first network entity with transfer learning of the global model for the first feature. The one ormore second network entity(s) may be capable of supporting the first network entity with vertical federated learning of the global model for the first feature. The one or more second network entity(s) may be capable of supporting the first network entity with vertical federated learning of the global model for the first feature and / or transfer learning of the global model for the first feature. The response to the second request message may be a Nnrf_NFDiscovery_Request Response message.

[0171] The method may further comprise sending, to the second network entity, a third request message comprising a training requirement. The third request message may be a federated learning preparation request. The third request message may be aNnwdaf MLModelTraining Subscribe message. The third request message may be a Nnwdaf MLModelTraininglnfo Request message.

[0172] The training requirement may comprise an identifier of the first feature. The training requirement may comprise a machine learning, ML, preparation flag. The training requirement may be a ML model training requirement.

[0173] The training requirement may comprise an indication for the second network entity to support the first network entity with training of the global model for the first feature using the first local model by vertical federated learning. The training requirement may comprise an indication for the second network entity to support the first network entity with training of the global model for the first feature using the first local model by transfer learning.

[0174] The method may further comprise receiving, from the second network entity, a response to the third request message comprising an indication whether the second network entity can support the training requirement. The response to the third request message may be a Nnwdaf MLModelTraining Notify message. The response to the third request message may be a Nnwdaf MLModelTraining Subscribe message. The response to the third request message may be a Nnwdaf MLModelTraininglnfo Request message.

[0175] The response to the third request message may comprise an indication that the second network entity will support the first network entity with training of the global model for the first feature using the first local model by vertical federated learning. The responseto the third request message may comprise an indication that the second network entity will support the first network entity with training of the global model for the first feature using the first local model by transfer learning.

[0176] The knowledge transfer may comprise information for adapting the first feature to create a second feature.

[0177] The first feature may be identical to the second feature. The second feature may be different to first feature. The information for adapting the first feature to create a second feature may comprise at least one of a weight with values for adapting a trained feature from a first analytics ID to a second analytics ID; an indication of specific objects of interest related to the first feature and the second feature; or an indication of rules that relate to the first feature and second feature.

[0178] The method may further comprise selecting the second feature based on at least one of a timestamp indicating the freshness of information relating to the first feature; information for adapting the first feature to create the second feature; and / or vendor interoperability information for adapting the first feature to create the second feature.

[0179] The first feature may comprise at least one of an event ID; a performance measurement; a key performance indicator; an analytics ID; a time range for collecting data; a user equipment type; a user mobility profile; a user contract profile; a mean opinion score related to application performance; a load level; a congestion level; a quality of service level; a quality of experience level; or user experience feedback in relation to an application.

[0180] There is provided a second network entity for wireless communication, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the second network entity to: receive, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and send, to the first network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature. Such a second network entity tends to reduce latency in training of the global model.

[0181] The first response message may comprise a knowledge transfer for supporting the first network entity with training of the global model for the first feature.

[0182] The at least one processor coupled with the at least one memory may be further configured to cause the second network entity to: send, to a network repository function, a fourth request message for registering a profile of the second network entity, wherein the fourth request message comprises an identifier of the first feature. The fourth request message may further comprise a flag that identifies support for VFL and / or (Vertical Federated) TL.

[0183] The fourth request message may be a Nnrf_NFManagement_NRRegister request message. The profile of the second network entity may comprise at least one of a second network entity type; one or more Analytics ID(s); address information of the second network entity, Service Area, FL capability type information or time interval supporting FL, an identifier of the first feature or transfer learning capabilities. The transfer learning capabilities may comprise at least one of one or more weights, one or more objects of interest or one or more rules.

[0184] The at least one processor coupled with the at least one memory may be further configured to cause the second network entity to: receive, from the network repository function, a response to the fourth request message.

[0185] The response to the fourth request message may be a Nnrf_NFManagement_NRRegister response message.

[0186] The knowledge transfer may comprise information from the first local model trained using local data for the first feature for vertical federated learning of the global model for the first feature.

[0187] The knowledge transfer may comprise the first local model for transfer learning of the global model for the first feature.

[0188] The knowledge transfer may further comprise transfer learning information.

[0189] The at least one processor coupled with the at least one memory may be further configured to cause the second network entity to: receive, from the first network entity, a third request message comprising a training requirement.

[0190] The at least one processor coupled with the at least one memory may be further configured to cause the second network entity to: send, to the first network entity, a response to the third request message comprising an indication whether the second network entity can support the training requirement.

[0191] The knowledge transfer may comprise information for adapting the first feature to create a second feature.

[0192] The first feature may comprise at least one of: an event ID; a performance measurement; a key performance indicator; an analytics ID; a time range for collecting data; a user equipment type; a user mobility profile; a user contract profile; a mean opinion score related to application performance; a load level; a congestion level; a quality of service level; a quality of experience level; or user experience feedback in relation to an application.

[0193] There is further provided a processor for wireless communication, comprising: at least one controller coupled with at least one memory and configured to cause the processor to: receive, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and send, to the first network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature. Such a processor tends to reduce latency in training of the global model.

[0194] There is further provided a method performed by a second network entity, the method comprising: receiving, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and sending, to the first network entity, a first response message comprising a knowledge transfer fortraining of the global model for the first feature. Such a method tends to reduce latency in training of the global model.

[0195] The method may further comprise sending, to a network repository function, a fourth request message for registering a profile of the second network entity, wherein the fourth request message comprises an identifier of the first feature. The fourth request message may further comprise a flag that identifies support for VFL and / or (Vertical Federated) TL. The method may further comprise receiving, from the network repository function, a response to the fourth request message.

[0196] The knowledge transfer may comprise information from the first local model trained using local data for the first feature for vertical federated learning of the global model for the first feature. The knowledge transfer may comprise the first local model for transfer learning of the global model for the first feature. The knowledge transfer may further comprise transfer learning information.

[0197] The method may further comprise receiving, from the first network entity, a third request message comprising a training requirement. The method may further comprise sending, to the first network entity, a response to the third request message comprising an indication whether the second network entity can support the training requirement.

[0198] The knowledge transfer may comprise information for adapting the first feature to create a second feature.

[0199] The first feature may comprise at least one of: an event ID; a performance measurement; a key performance indicator; an analytics ID; a time range for collecting data; a user equipment type; a user mobility profile; a user contract profile; a mean opinion score related to application performance; a load level; a congestion level; a quality of service level; a quality of experience level; or user experience feedback in relation to an application.

[0200] Vertical Federated Learning (VFL) relies on the capability and availability of the respective client to perform training of specified features with local data. That may not always be the case and it may introduce latency. At the same time certain features related to the same or even different Analytics IDs and ML models may also be transferred to speedup learning but there is no indication when and under which conditions and which knowledge can be transferred between a vertical federated server and a respective set of clients.

[0201] Examples described herein generally relate to an apparatus and method for introducing transfer learning in VFL scenarios allowing feature transfer among different Analytics IDs. Transfer learning is based on the concept that related features can be adapted to serve the needs of another problem. This can be achieved by introducing a weight, identify specific objects of interest for knowledge transfer or specific rules that relate different features or bring the feature relation experience of one domain to another. The clients may register in the repository their transfer learning capabilities, which may be evaluated by the federated learning server to make a choice based on time of a “fresh” retraining, the feature related information and interoperability.

[0202] In the past, wireless communications systems do not deal with transfer learning in a vertical federated learning environment and do not consider any means to achieve this.

[0203] Examples described herein may relate to a registration and discovery phase that allows server and client to register to a repository the transfer learning capabilities to allow vertical federated learning servers to perform discovery. Examples described herein may relate to a preparation phase that allows the server to check the transfer learning capabilities of the potential clients. Examples described herein may relate to an execution phase where the transfer leaning is performed to assist the ML model training.

[0204] There is provided herein, an apparatus and a method for assisting a first analytics network function to perform feature based vertical federated learning and / or feature based transfer learning with a second analytics function.

[0205] The first analytics network function may: select a second analytics network function for transfer learning of a feature, perform preparation to determine if the selected feature related to a second analytics network function conforms the feature capability and availability requirements, request to the selected second analytics network function to perform transfer leaning for a feature, receive from a second analytics network function a requested feature, aggregate and update the global model until a condition is reached.

[0206] The feature may be at least one of the following: a measurement type that can be produced from collected data, one or more attributes related to the output report of an analytics service, a time range related to collected data, a device type, a user profile, a network condition, an application and / or service performance level and / or a user experience information.

[0207] The first analytics network function may indicate a transfer leaning request related to a feature using a related flag or identifier.

[0208] The second analytics network function may register its transfer leaning capabilities related to a feature with a repository function (or NRF).

[0209] The second analytics function may provide the same or different analytic services.

[0210] The first analytics function and the second analytics function may exchange the same features and / or different features.

[0211] The feature may be an event and / or a performance measurement and / or a key performance indicator. The feature may be transferred and adapted to serve the purpose of a first analytics function service.

[0212] The feature may be transferred and adapted using at least one of the following: a weight that can modify the value and / or range of a feature, an identifier for indicating the objects of interest from which knowledge shall be transferred, a set of one or more rules that relate one feature to another feature, a set of one or more rules that relate a first domain knowledge of specified features to be applied to second domain.

[0213] The first analytics network function may select a feature to be transferred from a second analytics network function considering at least one of the following criteria: a timestamp that indicates the freshness of the feature information, a feature inter-related capability information among a first feature and a second feature that determines how a first feature can be converted and / or adapted and / or create a second feature, a vendor interoperability information related a specified feature that can be transferred.

[0214] It should be noted that the method described herein describes a possible implementation, and that the operations and the steps may be rearranged or otherwise modified and that other implementations are possible.

[0215] 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.

[0216] The following abbreviations are relevant in the field addressed by this document: 5G - 5thGeneration of Mobile Communication; AI / ML - Artificial Intelligence / Machine Learning; ADRF - Analytical Data Repository Function; AF - Application Function; AnLF - Analytics Logical Function; DCAF - Data Collection AF; DCCF - Data Collection Coordination Functionality; KPI - Key Performance Indicator; MF - Management Function; MFAF - Messaging Framework Adaptor Function; MnS - Management Service; MOS - Mean Opinion Score; MTLF - Model Training Logical Function; NF - Network Function; NRF - Network Repository Function; NWDAF - Network Data Analytics Function; 0AM - Operations, Administration and Maintenance; PDU - Packet Data Unit; PM - Performance Measurement; QoE - Quality of Experience; QoS - Quality of Service; TL - Transfer Learning; VFL - Vertical Federated Learning; UDM - User Data manager; UE - User Equipment.

Claims

CLAIMSWhat is claimed is:

1. A first network entity for wireless communication, comprising: at least one memory; and at least one processor coupled with the at least one memory and configured to cause the first network entity to: send, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receive, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

2. The first network entity of claim 1, wherein the knowledge transfer comprises information from the first local model trained using local data for the first feature for vertical federated learning of the global model for the first feature.

3. The first network entity of claim 1, wherein the knowledge transfer comprises the first local model for transfer learning of the global model for the first feature.

4. The first network entity of claim 3, wherein the knowledge transfer further comprises transfer learning information.

5. The first network entity of any one of claims 1 to 4, wherein the at least one processor coupled with the at least one memory is further configured to cause the first network entity to: send, to a network repository function, a second request message for discovering the second network entity, wherein the second request message comprises an identifier of the first feature and / or an identifier related to training capabilities of the first local model.

6. The first network entity of claim 5, wherein the at least one processor coupled with the at least one memory is further configured to cause the first network entity to:receive, from the network repository function, a response to the second request message, wherein the response to the second request message comprises one or more identifiers of one or more second network entities.

7. The first network entity of any one of claims 1 to 6, wherein the at least one processor coupled with the at least one memory is further configured to cause the first network entity to: send, to the second network entity, a third request message comprising a training requirement.

8. The first network entity of claim 7, wherein the at least one processor coupled with the at least one memory is further configured to cause the first network entity to: receive, from the second network entity, a response to the third request message comprising an indication whether the second network entity can support the training requirement.

9. The first network entity of any one of claims 1 to 8, wherein the knowledge transfer comprises information for adapting the first feature to create a second feature.

10. The first network entity of claim 9, wherein the at least one processor coupled with the at least one memory is further configured to cause the first network entity to: select the second feature based on at least one of: a timestamp indicating the freshness of information relating to the first feature; information for adapting the first feature to create the second feature; and / or vendor interoperability information for adapting the first feature to create the second feature.

11. The first network entity of any one of claims 1 to 10, wherein the first feature comprises at least one of: an event ID;a performance measurement; a key performance indicator; an analytics ID; a time range for collecting data; a user equipment type; a user mobility profile; a user contract profile; a mean opinion score related to application performance; a load level; a congestion level; a quality of service level; a quality of experience level; or user experience feedback in relation to an application.

12. A processor for wireless communication, comprising: at least one controller coupled with at least one memory and configured to cause the processor to: send, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receive, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

13. A method performed by a first network entity, the method comprising: sending, to a second network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and receiving, from the second network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

14. A second network entity for wireless communication, comprising:at least one memory; and at least one processor coupled with the at least one memory and configured to cause the second network entity to: receive, from a first network entity, a first request message comprising an indication for the second network entity to support the first network entity with training of a global model for a first feature using a first local model; and send, to the first network entity, a first response message comprising a knowledge transfer for training of the global model for the first feature.

15. The second network entity of claim 14, wherein the at least one processor coupled with the at least one memory is further configured to cause the second network entity to: send, to a network repository function, a fourth request message for registering a profile of the second network entity, wherein the fourth request message comprises an identifier of the first feature and / or an identifier related to training capabilities of the first local model.

16. The second network entity of claim 15, wherein the at least one processor coupled with the at least one memory is further configured to cause the second network entity to: receive, from the network repository function, a response to the fourth request message.

17. The second network entity of any one of claims 14 to 16, wherein the knowledge transfer comprises information from the first local model trained using local data for the first feature for vertical federated learning of the global model for the first feature.

18. The second network entity of any one of claims 14 to 16, wherein the knowledge transfer comprises the first local model for transfer learning of the global model for the first feature.

19. The second network entity of any one of claims 14 to 18, wherein the knowledge transfer further comprises transfer learning information.

20. The second network entity of any one of claims 14 to 19, wherein the at least one processor coupled with the at least one memory is further configured to cause the second network entity to: receive, from the first network entity, a third request message comprising a training requirement.

Images