Systems and methods for vertical federated learning training for inference for artificial intelligence and / or machine learning models

Abstract

A method (1000) by first node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500) using Vertical Federated Learning, VFL, to train an Artificial Intelligence, AI, and / or Machine Learning, ML, model, includes receiving (1002), from a second node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500), information comprising and / or indicating at least one of: a type of the AI and / or ML model, interoperability information indicating a type of training method that is supported by the second node for the AI and / or ML model, and at least one intermediate result.

Description

[0001] P111793WO01 (017997.4195) PATENT APPLICATION

[0002] 1

[0003] SYSTEMS AND METHODS FOR VERTICAL FEDERATED LEARNING TRAINING FOR INFERENCE FOR ARTIFICIAL INTELLIGENCE AND / OR MACHINE LEARNING

[0004] MODELS

[0005] TECHNICAL FIELD

[0006] The present disclosure relates, in general, to wireless communications and, more particularly, systems and methods for Vertical Federated Learning (VFL) training for inference for Artificial Intelligence (Al) and / or Machine Learning (ML) models.

[0007] BACKGROUND

[0008] Artificial Intelligence (Al) and Machine Learning (ML) have been investigated as promising tools to optimize the design of air-interface in wireless communication networks in both academia and industry.

[0009] Building an AI / ML model includes several development steps where the actual training of the Al model is just one step in a training pipeline. An important part in AI / ML development is AI / ML model lifecycle management, which is illustrated in FIGURE 1. More specifically, FIGURE 1 provides an illustration of training and interference pipelines, and their interactions within a model lifecycle management procedure.

[0010] As depicted in FIGURE 1, the Al model lifecycle management typically consists of:

[0011] • A training (re-training) pipeline, o With data ingestion referring to gathering raw (training) data from a data storage. After data ingestion, there may also be a step that controls the validity of the gathered data. o With data pre-processing referring to some feature engineering applied to the gathered data. For example, it may include data normalization and possibly data transformation required for the input data to the AI / ML model. o With the model training steps. o With model evaluation referring to benchmarking the performance to some baseline. The iterative steps of model training and model P111793WO01 (017997.4195) PATENT APPLICATION

[0012] 2 evaluation continue until the acceptable level of performance (as previously exemplified) is achieved. o With model registration referring to registering the AI / ML model, including any corresponding AI / ML-mcta data that provides information on how the AI / ML model was developed, and possibly AI / ML model evaluations performance outcomes.

[0013] • A deployment stage to make the trained (or re -trained) AI / ML model part of the inference pipeline,

[0014] • An inference pipeline, o With data ingestion referring to gathering raw (inference) data from a data storage. o With a data pre-processing stage that is typically identical to corresponding processing that occurs in the training pipeline. o With model operational referring to using the trained and deployed model in an operational mode. o With data & model monitoring referring to validating that the inference data are from a distribution that aligns well with the training data, as well as monitoring model outputs for detecting any performance or operational drifts.

[0015] • A drift detection stage that informs about any drifts in the model operations .

[0016] In a first scenario, we assume that AI / ML models operating with the existing standard airinterface are placed at the User Equipment (UE) side. A UE uses the AI / ML models to generate output that is reported to a centralized node in the network for positioning the UE location.

[0017] In a second scenario, we assume AI / ML models operating with the existing standard airinterface are placed at different Transmit / Receive Points (TRPs). A TRP uses the AI / ML models to generate output that is reported to a centralized node in the network for positioning operations to determine the UE location.

[0018] Unlike traditional centralized learning approaches, where data is pooled together in a single location, or Horizontal Federated Learning (HFL), where different entities contribute similar types of data about different samples, Vertical Federated Learning (VFL) allows for the collaborative training of machine learning models across entities that hold different types of information about the same entities or events. P111793WO01 (017997.4195) PATENT APPLICATION

[0019] 3

[0020] FIGURE 2 illustrates a procedure used by Network Data Analytics Function (NWDAF) to subscribe / unsubscribe Network Functions (NFs) in order to be notified for data collection on related event(s), using Event Exposure Services as listed in Table 6.2.2.1-1. Depending on local regulation requirements, user consent for UE-related data collection and usage of collected data may be required. User consent is defined for a specific purpose such as, for example, analytics or model training. NWDAF checks user consent, taking the purpose for data collection and usage of these data into account.

[0021] Technical Report 23.700.84-1.0.0 concluded that it will be possible to do UE Positioning using an AI / ML Model in Release 19. This AI / ML Model may be initially available at a Location Management Function (LMF). Then the LMF performs data collection to train the AI / ML model. The following are principles listed in the conclusions in the TR23.700.84-1.0.0:

[0022] Conclusions for KI#2: 5GC Support for Vertical Federated Learning

[0023] P#2.1: VFL related new functionalities include:

[0024] P#2.1. 1 VFL Server: An Network Data Analytics Function (NWDAF) or Application Function (AF) that integrates local training results for the local ML model update in VFL training process. It also coordinates the VFL training process by discovering and selecting VFL clients. In VFL inference process, The VFL server aggregates local inference results from VFL clients to generate the final VFL inference result and sends the final VFL inference result to the consumer. Only one VFL server exists for each VFL process.

[0025] P#2.1.2 VFL Client: An NWDAF or AF that holds the local dataset and performs local training and inference as asked by VFL Server. There can be multiple VFL Clients in VFL trainingAnLF and inference.

[0026] P#2.1.3 VFL process is associated with an analytics Identifier (ID) (i.e. VFL training of models for analytics ID, VFL inference for an analytics ID) when VFL server is the NWDAF. P111793WO01 (017997.4195) PATENT APPLICATION

[0027] 4

[0028] P#2.1.4 VFL process is associated with internal AF process (i.e. VFL training of models for internal AF process, VFL inference for internal AF process) when VFL server is AF.

[0029] P#2.1.5: 5thGeneration Core (5GC) shall support VFL, i.e. a federated learning (FL) technique without exchanging / sharing local data set or ML models, in the following scenarios:

[0030] - VFL among NWDAFs in a single Public Land Mobile Network (PLMN).

[0031] - VFL between AF and NWDAF(s) in a single PLMN.

[0032] The NWDAFs as VFL Server determines based on internal logic and the operator's policy whether or not to use VLF to provide a particular Analytics ID.

[0033] P#2.1.6: NWDAF and AF are the only NFs that may act with any of the above VFL functionalities, i.e. VFL Server and VFL client.

[0034] P#2.2: For registration and discovery of VFL entities:

[0035] P#2.2.1 The NWDAF as VFL client shall register to Network Resource Function (NRF) with NF profde including VFL capability information (VFL capability type (i.e., VFL Clients) ). For an untrusted AF as the VFL client, the Network Enforcement Function (NEF) registers based on configuration at the NRF within its Network Function (NF) profile information about the AF as specified in clause 6.2.2.3 of 3GPP TS 23.288 and includes as part of the information about the AF an VFL capability information (VFL capability type (i.e. VFL Clients) ).

[0036] P#2.2.2 The NWDAF as VFL server will select NWDAF(s) as candidate VFL client(s) and / or AF(s) (via Network Enforcement Function (NEF) profile for untrusted AF) from NRF for VFL training process. P111793WO01 (017997.4195) PATENT APPLICATION

[0037] 5

[0038] P#2.2.3 The AF (via NEF in case of untrusted AF) as VFL server will select candidate NWDAF(s) as VFL client(s) from NRF for VFL training process.

[0039] NOTE 1: Whether and how Vendor specific feature information in P#2.2.1 is included will be determined in the normative phase.

[0040] P#2.3: For sample alignment for VFL:

[0041] P#2.3.1: For NWDAF acting as VFL Server, NWDAF triggers sample alignment, may query NWDAFs or AFs acting as VFL clients to query availability of samples, and generates the interclause of samples.

[0042] P#2.3.2 In case of VFL between the NWDAF as VFL server and AF as VFL client, the NEF may be involved in sample alignment with mapping of information (e.g., internal versus external information) , then NWDAF as VFL server will determine the final list of participants supporting the samples for VFL training.

[0043] P#2.3.3: For AF acting as VFL Server and NWDAFs acting as VFL clients, AF performs sample alignment, selects samples to be used in the training process and generates the interclause of samples.

[0044] P#2.3.4: Feature description information may be registered to the NRF or locally configured in the NWDAF or the AF, negotiated between VFL server and VFL client when performing feature alignment. Feature alignment is optional in a VFL process.

[0045] NOTE 2: Whether and how the feature alignment is supported will be determined in normative phase. P111793WO01 (017997.4195) PATENT APPLICATION

[0046] 6

[0047] NOTE 3: The procedure to perform sample alignment is not agreed yet, i.e. it can be done in a standalone procedure or part of the preparation phase, which will be determined in normative phase.

[0048] NOTE 4: How NEF assists sample alignment in P#2.3.3 will be determined in normative phase.

[0049] NOTE 5: For P#2.3.2 and P#2.3.3, whether and how NEF supports to be involved in generating the intersection of candidate samples is to be discussed in normative phase.

[0050] P#2.4: For VFL training process:

[0051] P#2.4. 1 : Either the NWDAF or the AF can act as VFL server and initiate VFL training process with the VFL client(s).

[0052] P#2.4.2: If an untrusted AF is involved in the VFL training process, their interactions with the NWDAF(s) are via NEF. When the NWDAF acts as VFL Server, the NWDAF can receive labels from an AF.

[0053] P#2.4.3: An identifier is allocated by a VFL server, which is used to correlate the participants during the VFL training and subsequent VFL inference processes and it is associated with the distributed ML Models in the VFL joint model training process.

[0054] P#2.4.4: VFL Clients compute the intermediate results for their local ML models involved in the VFL training and provide reports with the intermediate results to the AF or NWDAF acting as VFL server.

[0055] P#2.4.5: VFL clients may also provide intermediate results (e.g. gradient information, loss information) to other VFL clients as instructed by the VFL server. P111793WO01 (017997.4195) PATENT APPLICATION

[0056] 7

[0057] P#2.4.6: An AF or NWDAF acting as VFL server aggregates intermediate results from VFL client(s), trains a local model, computes intermediate results based on its local ML model, and sends the intermediate results towards VFL clients involved in the joint VFL training process.

[0058] P#2.4.7: VFL server may compute different intermediate training information (e.g. gradient information, loss information) for updating its own local model and the models of VFL clients during the VFL training process after processing the received intermediate results (that may include convergence reports), sends the updates to the VFL client(s), and the VFL server / client(s) update their local ML model based on the received information. The VFL server determines when the VFL training process terminates, then it will inform the VFL Clients that the training ends.

[0059] NOTE 6: Whether further specification on how the AF provides the label to NWDAF is needed in P#2.4.2 will be discussed in normative work.

[0060] NOTE 7: Whether and how P#2.4.5 is supported will be determined in normative phase.

[0061] NOTE 8: How the initial model is provided to VFL clients is to be discussed in normative phase.

[0062] P#2.5: For VFL inference process:

[0063] P#2.5.1 : NWDAF acting as VFL server scenario, the NF consumer of analytics will obtain the required output based on the VFL inference process coordinated by VFL server, which is generated via the VFL inference process between VFL sever and corresponding AFs or NWDAF acting as VFL client(s): P111793WO01 (017997.4195) PATENT APPLICATION

[0064] 8

[0065] For NWDAF acting as VFL server, NWDAF triggers the VFL inference phase after receiving the subscription or request for analytics and delivers the analytics to the NF consumer.

[0066] P#2.5.2: For the AF acting as a VFL server, the AF acting as VFL server can also start an inference process with corresponding NWDAF acting as VFL client(s). The AF may be triggered to start the inference by a 5GC consumer (i.e. NWDAF containing Analytic Logical Function (AnLF)). Interactions are viaNEF if AF is untrusted.

[0067] P#2.5.3: Before performing the VFL inference, the NWDAF acting as VFL server determines corresponding AF(s) and / or NWDAF(s) as VFL client(s) for the inference process based on the same identifier used in the VFL training process. Related details will be specified in the normative phase.

[0068] P#2.5.4: The VFL inference process may be controlled by a set of requirements, i.e. whether the determination if all VFL participants associated with the same identifier are needed in the VFL inference process may be based e.g. on accuracy requirements, the VFL signalling and load cost, contribution weights of each client, and temporal availability of output from VFL participants.

[0069] P#2.5.5: When performing VFL model performance monitoring, inference data can be used for model retraining, which is aligned with R18.

[0070] NOTE 9: Whether and how P#2.5.4 is supported will be determined in normative phase.

[0071] NOTE 10: Whether the NWDAF supports both AnLF and Model Training Logical Function (MTLF) or not during VFL training and inference is to be discussed in normative phase. P111793WO01 (017997.4195) PATENT APPLICATION

[0072] 9

[0073] NOTE 11 : How multiple NWDAFs are involved in VFL when AF is acting as VFL server is defined in normative phase.

[0074] NOTE 12: The details of the accuracy monitoring related to VFL process will be defined in the normative phase.

[0075] NOTE 13: The details of (optionally new) services and detailed list of parameters to enable the VFL processes will be defined in the normative phase.

[0076] NOTE 14: Details of the ML model storage will be defined in the normative phase.

[0077] An example procedure where an NWDF acts as an FL Server with FVL capabilities is depicted and disclosed with regard to Figure 6.23.2.1-1 of Section 6.23.2 of 3GPP TR 23.700-84 VI .0.0. Another example procedure where an NWDAF acts as an FL server with VFL capabilities is depicted and disclosed with regard to Figure 6.23.2.4.1-1 of Section 6.23.2.4.1 of 3GPP TR 23.700-84 Vl.0.0.

[0078] An example procedure where an AF acts as an FL Server with VFL capabilities is depicted and disclosed with regard to Figure 6.23.2.2-1 of Section 6.23.2.2 of 3GPP TR 23.700-84 Vl.0.0. Another example procedure where an AF acts as a FL server with VFL capabilities is depicted and disclosed with regard to Figure 6.23.2.4.2-1 of Section 6.23.2.4.2 of 3GPP TR 23.700-84 Vl.0.0.

[0079] There currently exist certain challenge(s), however. For example, a problem of the published conclusions in the TR 23.700-84 is that the participants may use VFL to train different type of models, and each of these models require exchanging different information between participants. As such, the conclusions in TR 23.700-84 do not enable VFL when different ML Model types are supported.

[0080] SUMMARY

[0081] Certain aspects of the disclosure and their embodiments may provide solutions to these or other challenges.

[0082] According to certain embodiments, a method by first node using VFL to train an Al and / or ML model includes receiving, from a second node, information comprising and / or indicating at P111793WO01 (017997.4195) PATENT APPLICATION

[0083] 10 least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result.

[0084] According to certain embodiments, a first node using VFL to train an Al and / or ML model is configured to receive, from a second node, information comprising and / or indicating at least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result.

[0085] According to certain embodiments, a method by second node using VFL to train an Al and / or ML model includes transmitting, to a first node, information comprising and / or indicating at least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result.

[0086] According to certain embodiments, a second node using VFL to train an Al and / or ML model is configured to transmit, to a first node, information comprising and / or indicating at least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result.

[0087] Certain embodiments may provide one or more of the following technical advantage (s). For example, certain embodiments may provide a technical advantage of allowing VFL for ML models based on Neural Networks (NNs). As another example, certain embodiments may provide a technical advantage of complying with the conclusions in the TR 23.700-84.

[0088] Other advantages may be readily apparent to one having skill in the art. Certain embodiments may have none, some, or all of the recited advantages.

[0089] BRIEF DESCRIPTION OF THE DRAWINGS

[0090] For a more complete understanding of the disclosed embodiments and their features and advantages, reference is now made to the following description, taken in conjunction with the accompanying drawings, in which:

[0091] FIGURE 1 illustrates training and interference pipelines, and their interactions within a model lifecycle management procedure;

[0092] FIGURE 2 illustrates a procedure used by NWDAF to subscribe / unsubscribe at NFs in order to be notified for data collection on a related event;

[0093] FIGURE 3 illustrates an example registration and discover procedure for VFL, according to certain embodiments, according to certain embodiments;

[0094] FIGURE 4 illustrates an example preparation procedure used to check if the VFL Client(s) P111793WO01 (017997.4195) PATENT APPLICATION

[0095] 11 can meet the ML Model training requirement, according to certain embodiments;

[0096] FIGURE 5 illustrates a general procedure for FVL, according to certain embodiments;

[0097] FIGURE 6 illustrates an example procedure for distributed inference, according to certain embodiments; FIGURE 7 illustrates an example method by first node using VFL to train an Al and / or

[0098] ML model, according to certain embodiments;

[0099] FIGURE 8 illustrates another example method by first node using VFL to train an Al and / or ML model, according to certain embodiments;

[0100] FIGURE 9 illustrates an example method by second node using VFL to train an Al and / or ML model, according to certain embodiments;

[0101] FIGURE 10 illustrates another example method by second node using VFL to train an Al and / or ML model, according to certain embodiments;

[0102] FIGURE 11 illustrates an example communication system, according to certain embodiments; FIGURE 12 illustrates an example UE, according to certain embodiments;

[0103] FIGURE 13 illustrates an example network node, according to certain embodiments; and

[0104] FIGURE 14 illustrates a virtualization environment in which functions implemented by some embodiments may be virtualized, according to certain embodiments.

[0105] P111793WO01 (017997.4195) PATENT APPLICATION

[0106] 12

[0107] DETAILED DESCRIPTION

[0108] Some of the embodiments contemplated herein will now be described more fully with reference to the accompanying drawings. Embodiments are provided by way of example to convey the scope of the subject matter to those skilled in the art.

[0109] As used herein, ‘node’ can be a network node or a UE or any other node in a communications network. Examples of network nodes are NodeB, base station (BS), multistandard radio (MSR) radio node such as MSR BS, eNodeB (eNB), gNodeB (gNB), Master eNB (MeNB), Secondary eNB (SeNB), integrated access backhaul (IAB) node, network controller, radio network controller (RNC), base station controller (BSC), relay, donor node controlling relay, base transceiver station (BTS), Central Unit (e.g. in a gNB), Distributed Unit (e.g. in a gNB), Baseband Unit, Centralized Baseband, C-RAN, access point (AP), transmission points, transmission nodes, Remote Radio Unit (RRU), Remote Radio Head (RRH), nodes in distributed antenna system (DAS), core network node (e.g. Mobile Switching Center (MSC), Mobility Management Entity (MME), etc.), Operations & Maintenance (O&M), Operations Support System (OSS), Self-Organizing Network (SON), positioning node (e.g. E-SMLC), etc. The terms network node and radio network node are used interchangeably herein.

[0110] Another example of a node is user equipment (UE), which is a non-limiting term and refers to any type of wireless device communicating with a network node and / or with another UE in a cellular or mobile communication system. Examples of UE are target device, device to device (D2D) UE, vehicular to vehicular (V2V), machine type UE, Machine-Type Communications (MTC) UE or UE capable of machine to machine (M2M) communication, Personal Digital Assistant (PDA), Tablet, mobile terminals, smart phone, laptop embedded equipment (LEE), laptop mounted equipment (LME), Unified Serial Bus (USB) dongles, etc.

[0111] The term radio access technology (RAT), may refer to any RAT such as, for example, Universal Terrestrial Radio Access Network (UTRA), Evolved Universal Terrestrial Radio Access Network (E-UTRA), narrow band internet of things (NB-IoT), WiFi, Bluetooth, next generation RAT, NR, 4G, 5G, etc. Any of the equipment denoted by the terms node, network node or radio network node may be capable of supporting a single or multiple RATs.

[0112] The term signal or radio signal used herein can be any physical signal or physical channel. Examples of downlink (DL) physical signals are reference signal (RS) such as Primary Synchronization Signal (PSS), Secondary Synchronization Signal (SSS), Channel State Information-Reference Signal (CSI-RS), Demodulation Reference Signal (DMRS) signals in P111793WO01 (017997.4195) PATENT APPLICATION

[0113] 13

[0114] SS / PBCH block (SSB), discovery reference signal (DRS), Cell Specific Reference Signal (CRS), Positioning Reference Signal (PRS), etc. RS may be periodic. For example, RS occasions carrying one or more RSs may occur with certain periodicity (e.g., 20 ms, 40 ms, etc.). The RS may also be aperiodic.

[0115] Each SSB carries New Radio-Primary Synchronization Signal (NR-PSS), New RadioSecondary Synchronization Signal (NR-SSS) and New Radio-Physical Broadcast Channel (NR- PBCH) in four successive symbols. One or multiple Synchronization Signal Blocks (SSBs) are transmitted in one SSB burst which is repeated with certain periodicity such as, for example, 5 ms, 10 ms, 20 ms, 40 ms, 80 ms, and 160 ms. The UE is configured with information about SSB on cells of certain carrier frequency by one or more SS / PBCH block measurement timing configuration (SMTC) configurations. The SMTC configuration comprising parameters such as SMTC periodicity, SMTC occasion length in time or duration, SMTC time offset with regard to reference time (e.g., serving cell’s SFN) etc. Therefore, SMTC occasion may also occur with certain periodicity (e.g., 5 ms, 10 ms, 20 ms, 40 ms, 80 ms, and 160 ms). Examples of uplink (UL) physical signals are reference signals such as Sounding Reference Signals (SRS), Demodulation Reference Signals (DMRS), etc. The term physical channel refers to any channel carrying higher layer information e.g. data, control etc. Examples of physical channels are Physical Broadcast Channel (PBCH), Physical Downlink Control Channel (PDCCH), Physical Downlink Shared Channel (PDSCH), Physical Uplink Shared Channel (PUSCH), Physical Uplink Control Channel (PUCCH), Physical Uplink Shared Channel (PUSCH), Short PUSCH (sPUCCH), Short PDSCH (sPDSCH), Short PUCCH (sPUCCH), Short PUSCH (sPUSCH), MTC PDCCH (MPDCCH), Narrowband PBCH (NPBCH), Narrowband PDCCH (NPDCCH), Narrowband PDSCH (NPDSCH), Narrowband PUSCH (NPUSCH), Enhanced PDCCH (E-PDCCH), etc.

[0116] The term time resource used herein may correspond to any type of physical resource or radio resource expressed in terms of length of time. Examples of time resources are symbol, time slot, subframe, radio frame, transmission time interval (TTI), interleaving time, slot, sub-slot, mini-slot, system frame number (SFN) cycle, hyper-SFN (H-SFN) cycle, etc.

[0117] The term Analytics Accuracy Information represents a performance measure of an analytics identifier (ID) provided by an NWDAF containing AnLF, which is composed of the number of correct predictions of the analytics ID out of all predictions and the corresponding number of samples. P111793WO01 (017997.4195) PATENT APPLICATION

[0118] 14

[0119] The term Analytics Feedback Information indicates that the consumer NF has taken action(s) influenced by the previously provided analytics, which may or may not affect the ground truth data.

[0120] The term ML feature refers to the input data an ML model uses for training or inference.

[0121] The term ML Model Accuracy Information represents a performance measure of a ML Model provided by an NWDAF containing Model Training Logic Function (MTLF), which is composed of the number of correct predictions by the ML Model out of all predictions and the corresponding number of samples.

[0122] The Term VFL Client refers to an NF that holds the local dataset and performs local training and inference as requested by the VFL Server. There can be multiple VFL Clients in the VFL training and the VFL inference procedure.

[0123] The term VFL procedure refers to the procedure to train an ML Model or to perform inference of an ML Model using VFL techniques. When the VFL Server is the NWDAF, the VFL procedure is related to an ML Model for an Analytics ID.

[0124] The term VFL Server refers to an NF that integrates local training results for the local ML model updated in the VFL training procedure. In the VFL training procedure, the VFL Server also performs discovery and selection of VFL Clients. In the VFL inference procedure, the VFL server combines local inference results from VFL clients to generate the final VFL inference result and sends the final VFL inference result to the consumer.

[0125] The term VFL Active participant refers to the same as in Clause 3.1. In addition, the VFL Active Participant is also involved in other VFL tasks such as inference.

[0126] The term VFL Passive participant refers to the same as in Clause 3. 1. In addition, the VFL Active Participant is also involved in other VFL tasks such as inference, whether its feature set overlaps with the feature set of other passive participants or not is determined during the preparation phase fortraining. There can be multiple passive participants in VFL.

[0127] Herein, the output from passive participants, during prediction / forward propagation in NNs, is referred to as Activations. Activations refers to the output produced when performing forward propagation by one layer that is input to the next layer of a NN. The activation is produced by the VFL participant then sent to the active VFL participant as input to the process to produce output. P111793WO01 (017997.4195) PATENT APPLICATION

[0128] 15

[0129] Herein, output from active participant towards passive participants during re- calibration / backward propagation in NNs is called Gradients, which refers to the result produced when performing backward propagation.

[0130] The term Loss function refers to the difference between the output and the real label and is used to update the active participant's model part and sending gradients to passive participants to be able to update their local models.

[0131] The term ML feature refers to features that are the input data for a ML model. For each Analytics ID defined in 3GPP TS 23.288, a list of input data is defined. An example of the ML feature for a video Quality of Experience (QoE) estimation task would be underlying Quality of Service (QoS) parameters such as packet delay, packet loss, or throughput. Some ML features may not be available for all VFL participants or may not be specified such as, for example, those that are vendor specific.

[0132] As used herein, sample alignment ensures that all the VFL participants have common samples when training ML models.

[0133] Herein, the VFL coordinator coordinates the training process in VFL. For example, the VFL coordinator handles the task of selecting the participants in the VFL process.

[0134] For the purpose of aligning with terminology used for FL in 3GPP TS 23.288, the VFL Server with VFL capability is used to refer to the NF that plays the role of the VFL Coordinator and active VFL participant and FL Client with VFL Capabilities is used to refer to the NF that plays the role of passive VFL participant.

[0135] In order to train a model that is built using Neural Networks (NNs), one option is to use a distributed method such as VFL. The information that each participant (e.g., NWDAF and / or AF) needs to exchange during training is specific to the type of model that needs to be trained. When using such a distributed method like VFL, each participant may have available the ML Model. Such ML Model may be NN-based and / or provisioned by a coordinator, for example. Alternatively, each participant may not have a NN-based ML Model available, but each may select the NN-based model based on the requirements received from the coordinator.

[0136] According to certain embodiments, each participant registers in NRF its NF profile that includes the capability to train a NN-based ML model. Additionally, each participant registers information, which may be referred herein to as interoperability information, that defines the intermediate matrix that is exchanged between participants including both the number of nodes in P111793WO01 (017997.4195) PATENT APPLICATION

[0137] 16 the matrix and the type of content of each of the nodes in the intermediate matrix. As used herein, the term intermediate matrix contains intermediate information and / or intermediate result(s).

[0138] For example, according to certain embodiments, a participant may provide Nnrf NFManagement NFRegister service operation information. In a particular embodiment, for example, required input information associated with Nnrf NFManagement NFRegister service operation may include: NF type, NF instance ID, Fully Qualified Domain Name (FQDN) or Internet Protocol (IP) address of NF, Names of supported NF services (if applicable), and Public Land Mobile Network Identifier (PLMN ID) if, for example, NF needs to be discovered by other PLMNs / Standalone Non-Public Networks (SNPN)s.

[0139] In a particular embodiment, for example, if the consumer is Network Data Analytics Function (NWDAF) containing MTLF with FL capability for FL type VFL includes the type of ML Model set to NN-networks, optional input information associated with Nnrf NFManagement NFRegister service operation may include information indicating the dimensions of the matrix that can expose to other participants and the contents of each of the node of the exposed matrix.

[0140] For example, during the VFL training process, the VFL server triggers training an NN- based model with each VFL client. The VFL server then selects the candidate participants based on its NF profile and provides requirements for the VFL training process to each candidate participant. In a particular embodiment, for example, the requirements may include the dimensionality (i.e., number of nodes of the matrix that is exchanged between the VFL server and the VFL participants) and the type of content per node in the matrix.

[0141] As another example, according to certain embodiments, a participant may provide Nnwdaf MLModelTraining Sub scribe service operation information. In a particular embodiment, for example, required input information associated with Nnwdaf MLModelTraining Subscribe service operation information may include Analytics ID as defined in Table 7.1-2 of 3GPP TS 23.288 and a Notification Target Address (+ Notification Correlation ID).

[0142] In a particular embodiment, optional input information associated with Nnwdaf MLModelTraining Sub scribe service operation may include interoperability information containing the VFL training method for neural networks based ML Models, for example.

[0143] In another particular embodiment, the optional input information associated with Nnwdaf MLModelTraining Sub scribe service operation may include additional interoperability information that depends on the VFL training method. For example, when VFL training is for NN- P111793WO01 (017997.4195) PATENT APPLICATION

[0144] 17 based ML Models, the additional interoperability information may include the dimensionality of the intermediate results, such as, for example, including one or more of:

[0145] • number of nodes for the intermediate matrix, which indicates the number of samples and a results per sample, and

[0146] • per each of the nodes in the intermediate matrix, the type of content of the node.

[0147] In another particular embodiment, the optional input information associated with Nnwdaf MLModelTraining Sub scribe service operation may include intermediate results that carries information between the VFL participants. For example, in a particular embodiment, when the NN-model is trained then, the intermediate results information may include:

[0148] • an indicator that defines what it is provided in the matrix such as, for example, gradients that are provided; and / or

[0149] • an intermediate matrix container that includes the matrix such as, for example, gradients per batch of samples sent in the container.

[0150] As another example, according to certain embodiments, a participant may provide Nnwdaf MLModelTraining Notify service operation information. In a particular embodiment, for example, required input information associated with Nnwdaf MLModelTraining Notify service operation may include Notification Correlation Information, which may include a parameter indicating the Notification Correlation identifier (ID) that has been assigned by the consumer during ML Model training.

[0151] In a particular embodiment, optional input information associated with Nnwdaf MLModelTraining Notify service operation may include Interoperability information of which the VFL client can agree such as, for example, that the NN-based ML Model is accepted and / or can be trained.

[0152] In another particular embodiment, optional information associated with Nnwdaf MLModelTraining Notify service operation may include additional interoperability information that depends on the VFL training method. For example, in a particular embodiment, when VFL training is for NN-based ML Models, the additional interoperability information may include dimensionality of the intermediate results such as, for example, one or more of:

[0153] • number of nodes for the intermediate matrix accepted by the VFL client, which indicates the number of accepted samples and its content; and / or P111793WO01 (017997.4195) PATENT APPLICATION

[0154] 18

[0155] • per each of the nodes in the intermediate matrix, the type of content of the node accepted by the VFL client.

[0156] In another particular embodiment, optional information associated with Nnwdaf MLModelTraining Notify service operation may include intermediate results that carries information between the VFL participant. For example, when the NN-model is trained, the intermediate results may include one or more of:

[0157] • an indicator that defines what it is provided in the matrix e .g . activations are provided; and / or

[0158] • an intermediate matrix container that includes the matrix such as for example, activations per batch of samples sent in the container.

[0159] As another example, according to certain embodiments, a participant may provide Nnwdaf AnalyticsSubscription Subscribe information. For example, a participant may provide information indicating that the participant subscribes to NWDAF analytics. Optionally, the participant may provide Analytics Accuracy Information with specific parameters.

[0160] In a particular embodiment, for example, required input information associated with Nnwdaf AnalyticsSubscription Subscribe information may include one or more of:

[0161] • (Set of) Analytics ID(s) as defined in Table 7.1-2 of 3GPP TS 23.288;

[0162] • Target of Analytics Reporting (such as, for example, sample IDs);

[0163] • Notification Target Address (+ Notification Correlation ID); and / or

[0164] • Analytics Reporting Parameters (including Analytics target period, etc.).

[0165] As another example, according to certain embodiments, a participant may provide Nnwdaf AnalyticsSubscription Notify information. In a particular embodiment, for example, required input information associated with Nnwdaf AnalyticsSubscription Notify may include Notification Correlation Information. In a particular embodiment, this parameter indicates the Notification Correlation ID that has been assigned by the consumer during analytics subscription.

[0166] In a particular embodiment, for example, optional input information associated with Nnwdaf AnalyticsSubscription Notify may include similar information to that described above with regard to NnwdafyMLModelTraining Notify.

[0167] NWDAF Discovery and Selection P111793WO01 (017997.4195) PATENT APPLICATION

[0168] 19

[0169] The NWDAF service consumer selects an NWDAF that supports requested analytics information and required analytics capabilities and / or requested ML Model Information by using the NWDAF discovery principles defined in Clause 6.3.13 of 3GPP TS 23.501.

[0170] Different deployments may require different discovery and selection parameters. Different ways to perform discovery and selection mechanisms depend on different types of analytics / data (NF related analytics / data and UE related analytics / data). NF related refers to analytics / data that do not require a Subscription Permanent Identifier (SUPI) nor group of SUPIs (e.g., NF load analytics). UE related refers to analytics / data that requires SUPI or group of SUPIs (e.g., UE mobility analytics).

[0171] To discover an NWDAF containing AnLF using the NRF:

[0172] - If the analytics is related to NF(s) and the NWDAF service consumer (other than an NWDAF) cannot provide an Area of Interest for the requested data analytics, the NWDAF service consumer may select an NWDAF with large serving area from the candidate NWDAFs from a discovery response. Alternatively, in case the consumer receives NWDAF(s) with aggregation capability, the consumer preferably selects an NWDAF with aggregation capability with large serving area.

[0173] NOTE 1: If the selected NWDAF cannot provide the requested data analytics, e.g., due to the NF(s) to be contacted being out of serving area of the NWDAF, the selected NWDAF might reject the analytics request / subscription or it might query the NRF with the service area of the NF to be contacted to determine another target NWDAF.

[0174] - If the analytics is related to UE(s) and the NWDAF service consumer (other than an NWDAF) cannot provide an Area of Interest for the requested data analytics, the NWDAF service consumer may select an NWDAF with large serving area from the candidate NWDAFs from a discovery response. Alternatively, in case the consumer receives NWDAF(s) with aggregation capability, the consumer preferably selects an NWDAF with aggregation capability with large serving area.

[0175] NOTE 2: If a selected NWDAF cannot provide analytics for the requested UE(s) (e.g., the NWDAF serves a different serving area), the selected NWDAF might reject the analytics request / subscription or it might P111793WO01 (017997.4195) PATENT APPLICATION

[0176] 20 determine the Application Management Function (AMF) serving the UE as specified in Clause 6.2.2. 1, request UE location information from the AMF and query the NRF with the tracking area where the UE is located to discover another target NWDAF serving the area where the UE(s) is located.

[0177] - If the analytics are related to UE(s) and if NWDAF instances indicate weights for Tracking Area Identifiers (TAIs) in their NF profile (see Clause 6.3.13 of3GPP TS 23.501), the NWDAF service consumer may use the weights for TAIs to decide which NWDAF to select.

[0178] If the NWDAF service consumer needs to discover an NWDAF containing an AnLF with analytics accuracy checking capability, the consumer may query NRF providing also the analytics accuracy checking capability in the discovery request.

[0179] If the NWDAF service consumer needs to discover an NWDAF that is able to collect data from particular data sources identified by their NF Set IDs or NF types or to collect data from particular NWDAF Serving Area, the consumer may query NRF providing the NF Set IDs or NF types or Area of Interest in the discovery request:

[0180] NOTE 3: The NF Set ID or NF Type of a data source serving a particular UE, can be determined as indicated in Table 5A.2-1.

[0181] In order to discover an NWDAF that has registered in UDM for a given UE:

[0182] - NWDAF service consumers or other NWDAFs interested in UE related data or analytics, if supported, may make a query to UDM to discover an NWDAF instance that is already serving the given UE.

[0183] If an NWDAF service consumer needs to discover NWDAFs with data collection exposure capability, the NWDAF service consumer may discover, via NRF, the NWDAF(s) that provide the Nnwdaf DataManagement service and their associated NF type of data sources or their associated NF Set ID of data sources or NWDAF Serving Area information as defined in Clause 6.3.13 of 3GPP TS 23.501.

[0184] In order to discover an NWDAF containing MTLF via NRF: P111793WO01 (017997.4195) PATENT APPLICATION

[0185] 21

[0186] When one or more trained ML Models are available for one or more Analytics ID(s) the NWDAF containing MTLF shall include the Analytics ID(s) that is(are) supported per service in the registration towards NRF. The NWDAF containing MTLF may wait to register in NRF the above services until at least one trained model is available. The NWDAF containing MTLF may provide to the NRF a list of Analytics IDs corresponding to the trained ML Models and possibly the ML Model Filter Information for the trained ML Model per Analytics ID(s), if available. In this Release of the specification, only the Single Network Slice Selection Assistance Information (S-NSSAI(s)) and Area(s) of Interest from the ML Model Filter Information for the trained ML Model per Analytics ID(s) may be registered into the NRF during the NWDAF containing MTLF registration. If the NWDAF containing MTLF supports ML Model interoperability, the NWDAF containing MTLF includes, in the registration to the NRF, an ML Model Interoperability indicator for each Analytics ID.

[0187] - The ML Model Interoperability indicator comprises a list of NWDAF providers (vendors) that are allowed to retrieve ML Models from this NWDAF containing MTLF. It also indicates that the NWDAF containing MTLF supports the interoperable ML Models requested by the NWDAFs from the vendors in the list.

[0188] NOTE 4: The S-NSSAI(s) and Area(s) of Interest from the ML Model Filter Information are within the indicated S-NSSAI and NWDAF Serving Area information in the NF profde of the NWDAF containing MTLF, respectively.

[0189] - During the discovery of NWDAF containing MTLF, a consumer (e.g. an NWDAF containing AnLF, an NWDAF containing MTLF as FL server or FL client) may include in the request the target NF type (i.e. NWDAF), the Analytics ID(s), the S-NSSAI(s), Area(s) of Interest of the Trained ML Model required, ML Model Interoperability indicator and NF consumer information. The NRF returns one or more candidate instances of NWDAF containing MTLF to the NF consumer and each candidate instance of NWDAF containing MTLF includes the Analytics ID(s), possibly the ML P111793WO01 (017997.4195) PATENT APPLICATION

[0190] 22

[0191] Model Filter Information for the available trained ML Models and ML Model Interoperability indicator, if available.

[0192] NOTE 5: NF consumer information such as Vendor ID is defined in stage 3.

[0193] - If the NWDAF service consumer needs to discover an NWDAF containing an MTLF with ML Model accuracy checking capability, the consumer may query NRF also providing the ML Model accuracy checking capability in the discovery request.

[0194] In order to discover an NWDAF containing MTLF with Federated Learning (FL) capability via NRF, in addition to the procedures described above for discovering NWDAF containing MTLF:

[0195] - An NWDAF containing MTLF supporting FL as a server shall additionally include FL capability type (i.e., FL server), may include the supported FL type (i.e., VFL or HFL), and may include Time interval supporting FL both as FL capability information and the supported FL type during the registration in NRF.

[0196] NOTE 6: When the supported FL type is not provided, the FL capability type refers to the NWDAF capability for FL server or FL client for HFL.

[0197] - An NWDAF containing MTLF supporting FL as a client shall additionally include FL capability type (i.e. FL client), may include the supported FL type (i.e. VFL or HFL) and may include Time interval supporting both FL as FL capability information and the supported FL type during the registration in NRF, and it may also include, NF type(s) and NWDAF Serving Area information and / or NF set ID(s) of the data source(s) where data can be collected as input for local model training.

[0198] NOTE 7: An NWDAF containing MTLF may indicate to support both

[0199] FL server and FL client in the FL capability for specific Analytics ID. An NWDAF containing MTLF may indicate to support both HFL and VFL in the supported FL type for a specific Analytics ID. P111793WO01 (017997.4195) PATENT APPLICATION

[0200] 23

[0201] - An NWDAF containing MTLF supports FL as client or as server, and supporting FL type as VFL may include the Interoperability information (e.g. type of VFL training method, dimensionality and support for ML model convergence reports), optionally supported Feature IDs as part of the NF profile registered in NRF.

[0202] - During the discovery of NWDAF containing MTLF as FL server, a consumer (e.g. a NWDAF containing MTLF) may include in the request the FL capability type as FL server, the supported FL type (i.e. VFL or HFL) and may include Time Period of Interest and ML Model Filter information for the trained ML Model(s) per Analytics ID(s), if available. The NRF returns one or more NF profiles of candidate instances of NWDAF satisfying the query parameters.

[0203] - During the discovery of NWDAF containing MTLF as FL client, a consumer (e.g. an FL server) may include in the request FL capability type as FL client and may provide the supported FL types (i.e. VFL or HFL) and may include Time Period of Interest, a list of NF type(s) and / or NF set ID(s) of the data source(s). The NRF returns one or more NF profiles of candidate instances of NWDAF satisfying the query parameters.

[0204] NOTE 8: The service consumer to discover an NWDAF containing

[0205] MTLF with FL capability is limited to NWDAF containing MTLF and / or AF in this Release.

[0206] A Policy Control Function (PCF) may learn which NWDAFs being used by AMF, SMF, and UPF for a specific UE, via signalling described in Clause 4.16 of 3GPP TS 23.502. This enables a PCF to select the same NWDAF instance that is already being used for a specific UE.

[0207] In the roaming architecture, the NWDAF with roaming exchange capability (RE-NWDAF) for requesting analytics or input data is discovered via the NRF. A consumer in the same Public Land Management Network (PLMN) as the RE-NWDAF discovers the RE-NWDAF(s) by querying for NWDAF(s) where the roaming exchange capability is indicated in its (their) NF profile. A consumer in a peer PLMN (i.e., RE-NWDAF) discovers the RE-NWDAF(s) by querying for NWDAF(s) in the target PLMN that is (are) supporting the specific services defined for roaming. A RE-NWDAF discovers the RE-NWDAF(s) in a different PLMN (i.e., Home P111793WO01 (017997.4195) PATENT APPLICATION

[0208] 24

[0209] PLMN (HPLMN) or Visited PLMN (VPLMN)) using the procedure defined in Clause 4.17.5 (if delegated discovery is not used) or Clause 4.17.10 (if delegated discovery is used) of 3GPP TS 23.502, where the detailed parameters are determined based on the analytics request or subscription from the consumer 5thGeneration Core (5GC) NF, operator policy, user consent, and / or local.

[0210] HFL Among Multiple NWDAFs

[0211] FL among multiple NWDAFs is a machine learning technique in core network that trains an ML Model across multiple decentralized entities holding local data set, without exchanging / sharing local data set. This approach stands in contrast to centralized machine learning techniques where all the local datasets are uploaded to one server, thus allowing to address critical issues such as data privacy, data security, data access rights. It is noted that HFL is supported among multiple NWDAFs, which means the local data set in different FL client NWDAFs have the same feature space for different samples (e.g. UE IDs).

[0212] For FL supported by multiple NWDAFs containing MTLF, there is one NWDAF containing MTLF acting as FL server (called FL server NWDAF for short) and multiple NWDAFs containing MTLF acting as FL client (called FL client NWDAF for short), the main functionality includes:

[0213] FL server NWDAF : discovers and selects FL client NWDAFs to participant in an FL procedure; requests FL client NWDAFs to do local model training and to report local model information;

[0214] - generates global ML Model by aggregating local model information from FL client NWDAFs; and sends the global ML Model back to FL client NWDAFs to perform an additional training iteration if needed.

[0215] FL client NWDAF: locally trains ML Model as tasked by the FL server NWDAF with the available local data set, which includes the data that may not be allowed to be shared with other FL client NWDAFs due to, for example, data privacy, data security, data access rights; P111793WO01 (017997.4195) PATENT APPLICATION

[0216] 25 reports the trained local ML Model information to the FL server NWDAF; receives the global ML Model from FL server NWDAF and perform an additional training iteration if needed.

[0217] FL server NWDAF or FL client NWDAF register to NRF with their FL capability information as described in Clause 5.2.

[0218] The NWDAF containing MTLF determines to train an ML Model either based on local configuration or when it receives a request from NWDAF containing AnLF. The NWDAF containing MTLF further determines whether the ML Model should be trained via FL mechanism based on Analytic ID, Service Area / DNAI or when data cannot be obtained directly from data producer NF (e.g., due to data privacy, data security). The NWDAF containing AnLF is not aware whether the ML Model is trained based on FL or not.

[0219] If the NWDAF containing MTLF can act as an FL server for the ML Model training, then FL procedure is initiated by the NWDAF containing MTLF as FL server NWDAF directly.

[0220] If the NWDAF containing MTLF determines to train an ML Model based on local configuration and the FL mechanism is required, but the NWDAF containing MTLF can't act as an FL server, the NWDAF containing MTLF discovers an FL server NWDAF as described in Clause 5.2 and requests the FL server NWDAF to provide the trained ML Model as described in Clause 6.2C.2.2. The FL server NWDAF may determine to initiate FL procedure before providing the ML Model.

[0221] If the ML Model training is triggered by the request from NWDAF containing AnLF, the NWDAF containing MTLF determines the FL mechanism is required but it cannot act as an FL server, the NWDAF containing MTLF should discover an FL server NWDAF as described in Clause 5.2 and request the FL server NWDAF to provide the trained ML Model as described in Clause 6.2C.2.2. The Notification Target Address and the Notification Correlation ID from the NWDAF containing AnLF is provided in the request message sent to the FL server NWDAF. The FL server NWDAF may determine to initiate FL procedure before providing the ML Model. The FL server NWDAF sends the ML Model information to the notification endpoint (e.g., the NWDAF containing AnLF) after the ML Model training success.

[0222] The security procedure on authorizing FL server to initiate FL procedure on the FL client(s) is described in Annex X, clause X.9 of 3GPP TS 33.501. The security procedure authorizing an P111793WO01 (017997.4195) PATENT APPLICATION

[0223] 26

[0224] MTLF to request ML Models on behalf of an AnLF to another MTLF (e.g., FL server NWDAF) is described in Annex X, clause X.10 of 3GPP TS 33.501.

[0225] Before FL procedure is initiated by FL server NWDAF, appropriate FL client NWDAFs should be discovered by FL server NWDAF as described in Clause 5.2.

[0226] When starting an FL procedure, the FL server NWDAF is to provide an initial model to each FL client NWDAF, and then each FL client NWDAF is to perform local model training using its local data set. The detailed procedure for FL among Multiple NWDAFs is described in Clause 6.2C.

[0227] VFL

[0228] VFL is a machine learning technique that trains an ML Model across multiple NFs. Each NF trains their local ML Model with its local data for the same samples. Only NWDAF containing MTLF and AF are the NFs that train the ML Model using VFL, possibility via NEF. When NEF is involved, the NEF may provide a candidate list of VFL Clients that fulfil certain criteria provided by the AF.

[0229] A FL Server refers to a NF that has the FL capability type set to FL server and the supported FL type set to VFL.

[0230] VFL includes the following procedures:

[0231] - Registration of the NF profde including a list of VFL related information to NRF. Registration of the NWDAF profde to NRF is described in Clause 5.2. For an untrusted AF, the NEF registers based on configuration at the NRF within its NF profile information about the AF as specified in Clause 6.2.2.3 of 3GPP TS 23.288 and includes as part of the information about the AF an VFL capability information. The procedure for registration and discovery of VFL server and VFL client is described in Clause 6.2x.2.1 of 3GPP TS 23.288.

[0232] - Preparation for VFL including sample alignment to ensure that all the VFL participants has common samples when training ML models, as described in Clause 6.2x.2.2 of 3GPP TS 23.288.

[0233] - VFL training as described in Clause 6.2x.2.3 of 3GPP TS 23.288.

[0234] - Distributed inference as described in Clause 6.2x.2.5 of 3GPP TS 23.288. P111793WO01 (017997.4195) PATENT APPLICATION

[0235] 27

[0236] Contents of ML Model Training

[0237] The consumers of the ML Model training services (i.e., an NWDAF containing MTLF and / or an AF) may provide the input parameters in Nnwdaf MLModelTraining Subscribe or Nnwdaf MLModelTraininglnfo Request service operations as listed below:

[0238] - Analytics ID: identifies the analytics for which the ML Model is requested to be trained.

[0239] For HFL, ML Model Interoperability Information as defined in Clause 6.2A.2 of 3GPP TS 23.288.

[0240] (Only for Nnwdaf MLModelTraining Subscribe) A Notification Target Address (+ Notification Correlation ID) as defined in 3GPP TS 23.502 Clause 4.15.1, allowing to correlate notifications received from the NWDAF containing MTLF with the subscription.

[0241] [OPTIONAL] For HFL, ML Model Information (as defined in Clause 6.2A.2 of 3GPP TS 23.288).

[0242] - [OPTIONAL] For HFL, ML Model file .

[0243] NOTE 1 : It is up to NWDAF implementation to determine whether to include ML Model file in input parameters considering ML Model file size, etc.

[0244] [OPTIONAL] ML Model identifier: For HFL, identifies the provided ML Model. For VFL identifies the server’s ML model.

[0245] [OPTIONAL] ML Preparation Flag: identifies whether the request is for preparing Federated Learning or executing Federated Learning.

[0246] [OPTIONAL] For HFL, ML Model Accuracy Check Flag: identifies that the request is for using the local training data as the testing dataset to calculate the Model Accuracy of the global ML Model provided by the NWDAF service consumer acting as the FL Server NWDAF.

[0247] [OPTIONAL] For HFL, ML Correlation ID: identifies the Federated Learning procedure for training the ML Model. This parameter is included when the service is used for FL.

[0248] [OPTIONAL] For HFL, Data Availability requirement. This is the requirement on data availability for the ML Model training, e.g., FL Server NWDAF sends the requirement in preparation request to a FL Client P111793WO01 (017997.4195) PATENT APPLICATION

[0249] 28

[0250] NWDAF for selecting the FL Client NWDAF which can meet the data availability requirement. The following may be included:

[0251] Event ID list to be collected for local model training;

[0252] Dataset statistical properties as defined in Clause 6.1.3;

[0253] Time window of the data samples; and / or Minimum number of data samples.

[0254] [OPTIONAL] FL Availability time requirement. This is the requirement on availability time for the ML Model training, e.g., FL Server NWDAF sends the requirement in preparation request to FL Client NWDAF for selecting the FL Client NWDAF which is available in the required time for training ML Model.

[0255] [OPTIONAL] Training Filter Information: enables to select which data for the ML Model training is requested, e.g., S-NSSAI, Area of Interest. Parameter types in the Training Filter Information are the same as or subset of parameter types in the ML Model Filter Information which are defined in Clause 6.2A.2 of 3GPP TS 23.288.

[0256] [OPTIONAL] Target of Training Reporting: indicates the object(s) for which data for ML Model training is requested, i.e. group of UEs identified by a list of Intemal-Group-Ids or any UE (i.e. all UEs). This parameter contains the samples for VFL training when samples are UEs, it is also used during the preparation phase for sample alignment.

[0257] [OPTIONAL] Use case context: indicates the context of use of ML Model. [OPTIONAL] Training Reporting Information with the following parameters:

[0258] Maximum response time: indicates maximum time for waiting notifications (i.e., model training results).

[0259] [OPTIONAL] Iteration round ID: indicates the iteration round number of current ML Model training.

[0260] [OPTIONAL] Expiry time.

[0261] [OPTIONAL] Indication of skipping the current FL round.

[0262] [OPTIONAL] During VFL preparation phase, Interoperability information containing the VFL training method (e.g., neural networks, etc). P111793WO01 (017997.4195) PATENT APPLICATION

[0263] 29

[0264] [OPTIONAL] During VFL preparation phase, additional interoperability information that depends on the VFL training method. When VFL training is NNs, the additional interoperability information, suggested by server includes the dimensionality of the intermediate results, which defines the number of samples, number of nodes for the intermediate matrix, the type of content in the intermediate matrix container, exchanged between FL Clients and FL Server (e.g., activations, gradients, etc). The supported model convergence reports.

[0265] - During VFL training, Intermediate results that carries information between the VFL participants (server or client). The Intermediate results contain when the VFL training is "neural networks”:

[0266] Intermediate matrix indicator may be set to, for example, gradients and / or other agreed content in the preparation phase.

[0267] Intermediate matrix container which includes the matrix of, for example, gradients per batch of samples sent in the container.

[0268] [OPTIONAL] model convergence reports which are used to support the converge of a model during training and may contain one or multiple of the below parameters, The server sends it to the clients that may use it to adapt the internal ML model training settings (Settings are implementation specifics or set by configuration and some examples of possible settings are given below):

[0269] Expected number of iterations left.

[0270] Convergence speed such as Fast / slow / almost done convergence. This may trigger the receiver to e.g., set the learning rate.

[0271] Under or over fitting. This may trigger the receiver to, for example, expand or reduce the input data to be collected.

[0272] - Not converging at all. This may trigger the receiver to, for example, change the ML Model.

[0273] Stuck in a local minimum or similar. May trigger the receiver to introduce random noise to its local model. P111793WO01 (017997.4195) PATENT APPLICATION

[0274] 30

[0275] - Addition of random noise or similar in Intermediate results. This is sent to inform the receiver, that local monitoring of convergence may diverge from expected.

[0276] The NWDAF containing MTLF provides to the consumer of the ML Model training service operations as described in Clause 7.10 and clause 7.11, the output information in notification or response as listed below:

[0277] (Only for Nnwdaf_MLModelTraining_Notify) The Notification Correlation Information.

[0278] [OPTIONAL] ML Model Information (as defined in Clause 6.2A.2 of 3GPP TS 23.288).

[0279] [OPTIONAL] ML Model identifier: identifies the provisioned ML Model.

[0280] [OPTIONAL] Global ML Model Accuracy information: The model metric value of the global ML Model and optionally the used metric, which is calculate by the FL Client NWDAF using the local training data as the testing dataset.

[0281] [OPTIONAL] Status report of FL training: Accuracy information of local model and Training Input Data Information (e.g., areas covered by the data set, sampling ratio, maximum / minimum of value of each dimension, etc.), which are generated by the FL Client NWDAF during FL procedure.

[0282] NOTE 2: The parameters in Training Input Data Information are up to the implementation.

[0283] [OPTIONAL] ML Correlation ID. This parameter may be included when the service is used for Federated Learning.

[0284] [OPTIONAL] Iteration round ID: indicates the iteration round number of ML Model training indicated by the FL Server NWDAF.

[0285] [OPTIONAL] Delay Event Notification with the following parameters:

[0286] Delay event indication: this parameter indicates that FL Client NWDAF is not able to complete the training of the interim local ML Model within the maximum response time provided by the FL Server NWDAF.

[0287] [OPTIONAL] Cause code (e.g. local ML Model training failure, more time necessary for local ML Model training, etc.). P111793WO01 (017997.4195) PATENT APPLICATION

[0288] 31

[0289] [OPTIONAL] Expected time to complete the training: Indicates to the FL Server NWDAF that expected remaining training time and may be provided with Delay Event Notification.

[0290] [OPTIONAL] For VFL preparation phase, agreeable Interoperability information

[0291] [OPTIONAL] For VFL preparation phase, maximum agreeable dimensionality of the Intermediate results

[0292] [OPTIONAL] For VFL preparation phase, the list of samples agreeable by VFL client.

[0293] [OPTIONAL] For VFL, model convergence reports (during preparation phase the listed parameters indicates which parameters are supported to be received from server. The client shall not send any other parameter than indicated as supported by server in the request)

[0294] [OPTIONAL] For VFL training, Intermediate results that carries information between the VFL participants (server or client). The Intermediate results contain when the VFL training is "neural networks”:

[0295] Intermediate matrix indicator may be set to, for example, activations and / or other agreed content in the preparation phase.

[0296] Intermediate matrix container which includes a matrix of, for example, activations per batch of sample sent in the container.

[0297] [OPTIONAL] model convergence reports which are used to support the converge of a model during training and may contain one or multiple of the parameters mentioned above in the request from the server, The reports are only sent if the client perform convergence monitoring. The reports may be used in server to either replace monitoring or support monitoring at server.

[0298] NOTE x: Addition of random noise or similar in Intermediate results is only added after server has sent a report that will initiate a setting in client that introduces such action. P111793WO01 (017997.4195) PATENT APPLICATION

[0299] 32

[0300] VFL Procedures

[0301] Registration and Discovery Procedure for VFL

[0302] FIGURE 3 illustrates an example registration and discover procedure 100 for VFL, according to certain embodiments. In a particular embodiment, a VFL Server 105 can either be an AF or an NWDAF. If the VFL Server 105 is an NWDAF, the FL client(s) 110 can be either AF(s) or NWDAF(s). If VFL Server 105 is an AF the client(s) can be NWDAF(s).

[0303] As depicted in FIGURE 3, reference numerals 130 to 140 include steps for an example registration procedure:

[0304] • 130. FL Server 105 or FL Client(s) 110 register to NRF 120 with its NF profde, which includes NWDAF NF Type (see Clause 5.2.7.2.2 of 3GPP TS 23.502), Analytics ID(s), FL capability type information (i.e., FL server 105 and / or FL client 110 and / or FL type (i.e., HFL and / or VFL)), Interoperability information, and Time interval supporting FL as described in Clause 5.2.

[0305] • 135: The NRF 120 stores the NF profdes.

[0306] • 140: The NRF 120 sends a.Nnrf_ NFManagement NFRegister Response to the VFL Server 105.

[0307] In particular embodiments, if AFs are untrusted, an NEF 115 is used to register information of the AF instead of the AF(s) as described in Clause 6.2.2.3 of 3GPP TS 23.288. It is noted that further extensions are needed to show when any of the VFL participants are an untrusted AF. In this case, the procedure will contain an NEF 115.

[0308] As depicted in FIGURE 3, reference numerals 145 to 155 include steps for an example discovery procedure:

[0309] 145-155. Server determines ML Model requires FL based on, for example, operator policy (e.g., pre-configured list of ML Models), Analytics ID or data cannot be obtained directly from data producer NF (e.g., due to privacy reasons).

[0310] If the VFL Server is an NWDAF containing MTLF and cannot perform as VFL Server NWDAF, the MTLF first discovers and selects VFL Server from NRF by invoking the Nnrf NFDiscovery Request service operation. The following criteria might be used: Analytic ID of the ML Model P111793WO01 (017997.4195) PATENT APPLICATION

[0311] 33 required, Model filter information as defined in Clause 6.2A.2 of 3GPP TS 23.288, FL capability Type (i.e., FL server), Time Period of Interest. NOTE: The above step may include a request to an AF to become VFL Server

[0312] Once the VFL Server (the requested or the selected one) is determined, the VFL Server discovers and selects client(s)

[0313] If VFL Server is an NWDAF, it uses NRF by invoking the Nnrf NFDiscovery Request service operation, at step 145. The following criteria might be used: Analytic ID of the ML Model required, FL capability Type (i.e., FL client), NF type(s) of data sources from which the FL Client NWDAF is able to collect data for local ML Model training, Time Period of Interest. At step 150, the NRF 120 gathers and provides the result. At step 155, the NRF 120 sends a Nnrf Disocver Request Response at VFL Server 105.

[0314] Preparation Procedure for VFL

[0315] FIGURE 4 illustrates an example preparation procedure 200 used to check if the FL Client(s) can meet the ML Model training requirement, according to certain embodiments. The procedure includes the negotiation, between VFL Server 205 and FL client(s) 210 on what of the Interoperability Information shall be used such as, for example, sample alignment and, optionally, feature alignment. As depicted, FIGURE 4 includes signaling between FL Server 205, FL Client(s) 210, and an NEF 215. It is noted that features can be non-specified and privacy protected and, therefore, feature alignment is a simple alignment using information registered in NRF (e.g., data sources info or proprietary Feature ID). If the VFL Server 205 is an AF, the VFL Server 205 may use UE member selection to select candidate UEs before starting the procedure illustrated in FIGURE 4. Further extensions may be needed to show when any of the VFL participants are an untrusted AF, and in such a case, the procedure contains NEF 215.

[0316] Furthermore, it is recognized that the FL preparation procedure 200 (i.e., steps 230-240) can be skipped if the VFL Server 205 decides that the FL Client(s) 210 support the FL procedure to be performed such as, for example, based on information acquired from previous FL procedures or from the NRF or based on local configuration. P111793WO01 (017997.4195) PATENT APPLICATION

[0317] 34

[0318] At step 230, VFL Server 205 sends FL preparation request to the FL Client(s) 210, using Nnf MLModelTraining Subscribe or Nnf MLModelTraininglnfo Request service with the ML Preparation Flag. In various particular embodiments, the VFL Server 205 may add one or more of Interoperability Information, sample IDs, Analytics ID, and, optionally, Feature ID. It is noted that, in a particular embodiment, Nnf MLModelTraining Subscribe can either be Nnwdaf_ MLModelTraining Subscribe or Naf_ MLModelTraining SubscribQ .

[0319] At step 235, FL Client(s) 210 check whether the FL Client(s) 210 can meet the ML Model training requirement. If the FL client(s) 210 cannot meet the requirements for any reason, FL Client(s) 210 may decide which requirements the FL Client(s) 210 can agree to.

[0320] At step 240, FL Client(s) 210 invokes Nnf MLModelTraining Notijy or Nnf MLModelTraining Subscribe response or Nnf MLModelTraininglnfo Request response service operation to indicate to the VFL Server 205 whether it accepts the requirements, request new requirements or whether it will not join. In a particular embodiment, FL Client(s) 210 include, in the response, the requirements the FL Client(s) 210 can accept or may add a reason if the FL Client(s) 210 cannot join the FL process.

[0321] In a particular embodiment, if the notify or response includes new requested requirements, one or more of the steps are repeated.

[0322] In a particular embodiment, FL Server NWDAF determines the FL Client(s) to be involved in the FL procedures based on the information received in step 6 in FIGURE 6.2X.2. 1-1 of 3GPP TS 23.288 and other information received in step 3 (if available).

[0323] Procedure for VFL

[0324] FIGURE 5 illustrates a general procedure 300 for FVL training, according to certain embodiments. As depicted, FIGURE 5 includes signaling between a FL Server 305, FL Client(s) 310, NEF 315, NRF 320, and NF 325. It is noted that further extensions are needed to show when any of the VFL participants are untrusted AF(s). In that case, the procedure will contain a NEF 315, and the service operations going via NEF 315 include the same service operation but as an Nnef service operation in between. It is also noted that further extensions are needed to the procedure, which only describes the FL Server 305 being NWDAF and FL Client(s) 310 are also NWDAFs. Extensions include when untrusted AF is included either as FL Server 305 or FL Client(s) 310, the service operations are changed accordingly to be an NFaf service operation. P111793WO01 (017997.4195) PATENT APPLICATION

[0325] 35

[0326] As depicted in FIGURE 5, at step 330, FL Server 305 selects FL Clients(s) 310 as described in clause 6.2X.2.1 and clause 6.2X.2.2 of 3GPP TS 23.288.

[0327] At step 335, FL Server 305 sends a Nnwdaf MLModelTraining Subscribe or NnwdafyMLModelTraininglnfo Request to the selected FL Clients(s) 310, which participate in the FL to perform the local model training. The request includes ML Model ID, sample IDs, and may include Interoperability Information and maximum response time, the FL Client(s) 310 has to report the intermediate results to the FL Server 305 before the maximum response time elapses.

[0328] At an optional step 340, each FL Client 310 collects its local data.

[0329] At step 345, during the FL training procedure, each FL Client 310 trains the local ML Model and reports the Intermediate result to the FL Server 305 in Nnwdaf MLModelTraining Notify or NnwdafyMLModelTraininglnfo Request response. The Intermediate results include Intermediate matrix indicator per Intermediate matrix container and may include model convergence reports.

[0330] It is noted that there can be multiple intermediate result containers in one message, in a particular embodiment. Further, in a particular embodiment, the intermediate content indicators indicate what is included in each container.

[0331] The Intermediate results, which is sent from the FL Client(s) 310 to the FL Server 305 during the FL training process 300, is the information needed by the FL Server 305 to generate the intermediate or final output.

[0332] It is noted that the output is the output for the Analytics ID, in a particular embodiment. During FL, it is intermediate. The output becomes final when training is done.

[0333] In a particular embodiment, a FL Client 310 may indicate in a message that the FL Client 310 will leave the FL.

[0334] In a particular embodiment, the FL Server 305 may inform FL Client(s) 310 to cease the ML Model training by sending termination request and to report back the current Intermediate results.

[0335] At step 350, the FL Server 305 generates the intermediate output. The FL Server 305 may also compute the loss and / or gradients, in a particular embodiment.

[0336] In a particular embodiment, at an optional step 355, the FL Server 305 may terminate the current FL training process.

[0337] If the FL Server 305 decides to stop the FL Training process, steps 360 and 365 are skipped. P111793WO01 (017997.4195) PATENT APPLICATION

[0338] 36

[0339] At step 360, if the FL procedure 300 continues, the FL Server 350 determines which FL Client(s) 310 to continue the FL with and sends Nnwdaf MLModelTraining Subscribe or Nnwdaf MLModelTraininglnfo Request that includes the intermediate result to the FL client(s) 310. In a particular embodiment, the Intermediate results include Intermediate matrix indicator per intermediate matrix container and may include model convergence reports and sample IDs (if changed), in various embodiments.

[0340] At step 365, each FL Client 310 performs training locally using info received by the FL Server 305 at step 360.

[0341] In particular embodiments, steps 340-365 are repeated until the training termination condition (e.g., maximum number of iterations, or the result of loss function is lower than a threshold) is reached or the FL Server 305 requests to pause.

[0342] In a particular embodiment, the FL Server 305 requests the FL client(s) 310 to stop the FL procedure by invoking Nnwdaf MLModelTraining Unsubscribe service with a cause code that the FL process has finished. If a stop is requested, the FL client(s) 310 terminate the local model training and stores at least the local Model.

[0343] It is noted that the local Model is connected inside the client to the Model ID received from the FL Server 305. The connection is used for inference, in a particular embodiment.

[0344] Distributed Inference

[0345] The inference process occurs once the training process has been done. Specifically, the VFL Server knows that there is an ML Model for an AnalyticsID trained and determines which ones from participants in the training will participate in the inference. The VFL Server has kept the client NF IDs from training and can initiate inference from the client NF IDs. The VFL Server and FL clients (if NWDAFs) are a combined NWDAF containing both MTLF and AnLF. The differences between the VFL training and inference are that there is no check of the labels and, as such, no intermediate results are sent to the FL Clients to repeat the process.

[0346] If the VFL Server has no suitable model available, in the response to the Analytics Subscription it responds to consumer that training of model is needed. Server starts training. If VFL is needed, step 1 is performed in Figure 6.2X.2.2-1 of 3GPP TS 23.288. After the training process is complete, the VFL Server NWDAF will send Nnwdaf AnalyticsSubscibe Notify that includes the output to the consumer. P111793WO01 (017997.4195) PATENT APPLICATION

[0347] 37

[0348] It is noted that the consumer can, at any time, unsubscribe to the Analytics subscription. Whether the VFL Server decides to stop the FL is up to VFL Server to decide.

[0349] FIGURE 6 illustrates an example procedure 400 for distributed inference, according to certain embodiments. As depicted, FIGURE 6 includes signaling between a consumer 402, an FL Server 405, FL Client(s) 410, anNEF 415, anNRF 420, and an NF 425. When any of the Consumer 402, FL Server 405, and FL Client(s) 410 are untrusted AF(s), the procedure 400 contains NEF 415, and the service operations going via NEF 415 include the same service operation but as an Nnef service operation in between. In the depicted procedure 400, the FL Server 405 is NWDAF and FL Client(s) 410 are also NWDAFs.

[0350] At step 430, the Consumer 402 sends a subscription request to FL Server 405 to retrieve Analytics. In a particular embodiment, and as depicted, the subscription request is a Nnwdaf AnalyticsSubscription message and includes an Analytics ID.

[0351] The FL Server 405 knows that there is an ML Model for an Analytics ID trained.

[0352] At step 435, FL Server 405 selects FL Clients(s) 410 using input from earlier training described in 6.2X.2 of 3GPP TS 23.288. The FL Server 405 may not select all FL Client(s) 410 such as, for example, depending on their contribution to the training result or if NWDAF decides to do inference independently of AF(s).

[0353] At step 440, FL Server 405 sends a. Nnwdaf AnalyticsSubscription Subscribe message to the selected FL Clients(s) 410. The request includes ML Model ID and samples IDs, in a particular embodiment.

[0354] In a particular embodiment, at an optional step 445, each FL Client 410 collects its local data.

[0355] At step 450, during distributed inference procedure, each FL Client 410 performs inference using the local ML Model and reports the intermediate result to the FL Server 405 in a Nnwdaf AnalyticsSubscription Notify message .

[0356] The Intermediate results, which is sent from the FL Client(s) 410 to the FL Server 405 during the FL training process, is the information needed by the FL Server 405 to generate the final output.

[0357] At step 455, the FL Server 405 generates the final output.

[0358] At step 460, the FL Server 405 sends a Nnwdaf ' AnalyticsSubscription Notify message to update the Consumer 402 with the requested Analytics. P111793WO01 (017997.4195) PATENT APPLICATION

[0359] 38

[0360] VFL Capabilities within NF Profile

[0361] As disclosed in Section 6.23.2.3 of 3GPP TR 23.700-84 VI.0.0, the passive participants such as the AF or the NEF (on behalf of the AF) or the NWDAF registers its VFL capabilities into NRF:

[0362] FL capability information (extending existing one), whether the AF or NWDAF can act as a FL Server with VFL Capabilities or FL Client with VFL Capabilities or both. (MANDATORY).

[0363] If FL capability information indicates that the AF or NWDAF can perform VFL then the AF or NWDAF provide the list of Analytics IDs that can train. Per Analytics ID the AF provides:

[0364] Interoperability information such as, VFL training method (e.g. neural networks, XGBoost, etc.). This is a non-specified parameter comparable to ML Model Interoperability Information defined in 3GPP TS 23.288 [5],

[0365] Dimensionality of the intermediate results (e.g. maximum number of samples and number of nodes).

[0366] NOTE: The intermediate results are sent in a container, similar as ML Model File parameter in Nnwdaf_MLModelTraining.

[0367] Whether the FL Client with VFL Capabilities / passive participant can share labels or not (when supervised learning applies).

[0368] Whether the FL Client / passive participant can perform loss calculation with the FL Server. whether the FL client or the FL server can share model convergence reports during the training phase. whether the AF, as FL server (with VFL capabilities) can share the impact on the output of the intermediate results provided by NWDAF during training.

[0369] (Optional). The list of supported ML features.

[0370] Other parameters that are listed in clause 6.2A.1 of 3GPP TS 23.502 [3], P111793WO01 (017997.4195) PATENT APPLICATION

[0371] 39

[0372] Training procedure

[0373] The training process in VFL including agreement between the FL server and the FL clients on the training method is illustrated below. The training process is repeated until the server decides to terminate it based on the local constraints that are set by the server and distributed to the clients.

[0374] When the ML Model is trained, the active participant notifies the passive participants that there is a trained ML Model available and that the training process has finalized. If the active or passive participant detect that there are changes e.g. in the input data, then it may trigger a retraining, not shown in the figure.

[0375] FIGURE 7 illustrates a method 900 by first node using VFL to train an Al and / or ML model, according to certain embodiments. As illustrated the method begins at step 902 when the first node receives, from a second node, information comprising and / or indicating at least one of:

[0376] • a type of the Al and / or ML model,

[0377] • interoperability information indicating a type of training method that is supported by the second node for the Al and / or ML model, and

[0378] • at least one intermediate result.

[0379] Optionally, in a particular embodiment, the method 900 further includes using the information to train the Al and / or ML model, at step 904.

[0380] FIGURE 8 illustrates another example method 1000 by first node using VFL to train an A, and / or ML model, according to certain embodiments. As illustrated, the method begins at step 1002 when the first node receives, from a second node, information comprising and / or indicating at least one of:

[0381] • a type of the Al and / or ML model,

[0382] • interoperability information, and

[0383] • at least one intermediate result.

[0384] In a particular embodiment, the interoperability information indicates a type of training method that is supported by the second node for the Al and / or ML model and / or agreed by the first node and second node.

[0385] In a particular embodiment, the interoperability information includes at least one of: an indicator of at least one gradient associated with the at least one intermediate result, and an indication of a number of gradients per batch of samples in an intermediate matrix container that includes the intermediate matrix. P111793WO01 (017997.4195) PATENT APPLICATION

[0386] 40

[0387] In a particular embodiment, the interoperability information includes at least one of: an indicator of at least one activations associated with the at least one intermediate result, and an indication of a number of activations per batch of samples in an intermediate matrix container that includes the intermediate matrix.

[0388] In a particular embodiment, the information is used to train the Al and / or ML model.

[0389] In a particular embodiment, the Al and / or ML model is neural network-based.

[0390] In a particular embodiment, the information and / or the interoperability information includes at least one parameter associated with the intermediate matrix to be exposed.

[0391] In a further particular embodiment, the at least one parameter associated with the intermediate matrix includes at least one dimension of the intermediate matrix, and the at least one dimension of the intermediate matrix comprises at least one of: a number of a plurality of nodes in an intermediate matrix; a content of each of a plurality of nodes of an intermediate matrix; and a type of content of each of the plurality of the nodes in the intermediate matrix.

[0392] In a further particular embodiment, the at least one parameter associated with the intermediate matrix comprises a number of activations per batch of samples in a container.

[0393] In a particular embodiment, the information and / or the interoperability information comprises at least one of: at least one analytics identifier, at least one sample identifier, at least one notification target address, at least one notification correlation identifier, and at least one analytics reporting parameter.

[0394] In a particular embodiment, the first node includes an NRF, and the second node includes an NF.

[0395] In a particular embodiment, the information comprises Nnrf NFManagement NFRegister service operation information.

[0396] In a particular embodiment, the first node is operating as a VFL client, and the second node is operating as a VFL server.

[0397] In a further particular embodiment, the information includes NnrfyMLModelTraining Subscribe service operation information.

[0398] In a particular embodiment, the first node is operating as a VFL server, and the second node is operating as a VFL client.

[0399] In a further particular embodiment, the information includes Nnwdaf MLModelTraining Notify service operation information. P111793WO01 (017997.4195) PATENT APPLICATION

[0400] 41

[0401] In a particular embodiment, the first node includes an NWDAF, which includes an AnLF, and the second node includes a consumer.

[0402] In a further particular embodiment, the information includes Nnwdaf AnalyticsSubscription Subscribe operation information.

[0403] In a particular embodiment, the first node includes a consumer, and the second node includes a NWDAF, which includes an AnLF.

[0404] In a further particular embodiment, the information comprises Nnwdaf AnalyticsSubscription Notify operation information.

[0405] In a particular embodiment, the first node or second node is operating as a VFL client and comprises a UE or a gNB.

[0406] In a particular embodiment, the information is received during a registration procedure.

[0407] In a particular embodiment, the information is received in response to or with a Federated Learning preparation request.

[0408] In a further particular embodiment, the method includes determining one or more FL parameters that the first node supports, and transmitting, to the second node, a FL response indicating the one or more FL parameters that the first node supports.

[0409] FIGURE 9 illustrates a method 1100 by second node using VFL to train an Al and / or ML model, according to certain embodiments. As illustrated, the method 1100 optionally begins at step 1102 when the second node determines that an Al and / or ML model requires a type of training. At step 1104, the second node transmits to a first node, information comprising and / or indicating at least one of:

[0410] • a type of the Al and / or ML model,

[0411] • interoperability information indicating a type of training method that is supported by the second node for the Al and / or ML model, and

[0412] • at least one intermediate result.

[0413] FIGURE 10 illustrates another example method 1200 by second node using VFL to train an Al and / or ML model, according to certain embodiments. As depicted, the method begins at step 1202 when the second node transmits, to the first node, information comprising and / or indicating at least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result. P111793WO01 (017997.4195) PATENT APPLICATION

[0414] 42

[0415] In a particular embodiment, the interoperability information indicates a type of training method that is supported by the second node for the Al and / or ML model and / or agreed by the first node and second node.

[0416] In a particular embodiment, the interoperability information includes at least one of: an indicator of at least one gradient associated with the at least one intermediate result, and an indication of a number of gradients per batch of samples in an intermediate matrix container that includes the intermediate matrix.

[0417] In a particular embodiment, the interoperability information comprises at least one of: an indicator of at least one activations associated with the at least one intermediate result, and an indication of a number of activations per batch of samples in an intermediate matrix container that includes the intermediate matrix.

[0418] In a particular embodiment, the second node determines that the Al and / or ML model requires the type of training, and wherein the information is transmitted to the first node based on determining that the Al and / or ML model requires the type of training.

[0419] In a particular embodiment, the second node uses the information to train the Al and / or ML model.

[0420] In a particular embodiment, the Al and / or ML model is neural network-based.

[0421] In a particular embodiment, the information and / or the interoperability information includes at least one parameter associated with an intermediate matrix to be exposed.

[0422] In a further particular embodiment, the at least one parameter associated with the intermediate matrix includes at least one dimension of the intermediate matrix, and the at least one dimension of the intermediate matrix includes at least one of: a number of a plurality of nodes in an intermediate matrix; a content of each of a plurality of nodes of an intermediate matrix; and a type of content of each of the plurality of the nodes in the intermediate matrix.

[0423] In a particular embodiment, the at least one parameter associated with the intermediate matrix includes a number of activations per batch of samples in a container.

[0424] In a particular embodiment, the information and / or the interoperability information comprises at least one of: at least one analytics identifier, at least one sample identifier, at least one notification target address, at least one notification correlation identifier, and at least one analytics reporting parameter.

[0425] In a particular embodiment, the first node includes an NRF, and the second node includes an NF. P111793WO01 (017997.4195) PATENT APPLICATION

[0426] 43

[0427] In a further particular embodiment, the information comprises NnrfyNFManagement NFRegister service operation information.

[0428] In a particular embodiment, the first node is operating as a VFL client, and the second node is operating as a VFL server.

[0429] In a further particular embodiment, the information comprises NnrfyMLModelTraining Subscribe service operation information.

[0430] In a particular embodiment, the first node is operating as a VFL server, and the second node is operating as a VFL client.

[0431] In a further particular embodiment, the information comprises Nnwdaf MLModelTraining Notify service operation information.

[0432] In a particular embodiment, the first node includes an NWDAF comprising an AnLF, and the second node includes a consumer.

[0433] In a further particular embodiment, the information comprises Nnwdaf AnalyticsSubscription Subscribe operation information.

[0434] In a particular embodiment, the first node includes a consumer, and the second node includes a NWDAF comprising an AnLF.

[0435] In a particular embodiment, the information includes Nnwdaf AnalyticsSubscription Notify operation information.

[0436] In a particular embodiment, the first node or second node is operating as a VFL client and is a UE or a gNB.

[0437] In a particular embodiment, the information is transmitted during a registration procedure for registering the first node with the second node.

[0438] In a particular embodiment, the information is transmitted in response to or with a FL preparation request.

[0439] In a particular embodiment, the second node receives, from the first node, a FL response indicating the one or more FL parameters that the first node supports.

[0440] FIGURE 11 shows an example of a communication system 1300 in accordance with some embodiments. In the example, the communication system 1300 includes a telecommunication network 1302 that includes an access network 1304, such as a radio access network (RAN), and a core network 1306, which includes one or more core network nodes 1308. The access network 1304 includes one or more access network nodes, such as network nodes 1310A and 1310B (one or more of which may be generally referred to as network nodes 1310), or any other similar 3rd P111793WO01 (017997.4195) PATENT APPLICATION

[0441] 44

[0442] Generation Partnership Project (3GPP) access node or non-3GPP access point. The network nodes 1310 facilitate direct or indirect connection of user equipment (UE), such as by connecting UEs 1312A, 1312B, 1312C, and 1312D (one or more of which may be generally referred to as UEs 1312) to the core network 1306 over one or more wireless connections.

[0443] Example wireless communications over a wireless connection include transmitting and / or receiving wireless signals using electromagnetic waves, radio waves, infrared waves, and / or other types of signals suitable for conveying information without the use of wires, cables, or other material conductors. Moreover, in different embodiments, the communication system 1300 may include any number of wired or wireless networks, network nodes, UEs, and / or any other components or systems that may facilitate or participate in the communication of data and / or signals whether via wired or wireless connections. The communication system 1300 may include and / or interface with any type of communication, telecommunication, data, cellular, radio network, and / or other similar type of system.

[0444] The UEs 1312 may be any of a wide variety of communication devices, including wireless devices arranged, configured, and / or operable to communicate wirelessly with the network nodes 1310 and other communication devices. Similarly, the network nodes 1310 are arranged, capable, configured, and / or operable to communicate directly or indirectly with the UEs 1312 and / or with other network nodes or equipment in the telecommunication network 1302 to enable and / or provide network access, such as wireless network access, and / or to perform other functions, such as administration in the telecommunication network 1302.

[0445] In the depicted example, the core network 1306 connects the network nodes 1310 to one or more hosts, such as host 1316. These connections may be direct or indirect via one or more intermediary networks or devices. In other examples, network nodes may be directly coupled to hosts. The core network 1306 includes one more core network nodes (e.g., core network node 1308) that are structured with hardware and software components. Features of these components may be substantially similar to those described with respect to the UEs, network nodes, and / or hosts, such that the descriptions thereof are generally applicable to the corresponding components of the core network node 1308. Example core network nodes include functions of one or more of a Mobile Switching Center (MSC), Mobility Management Entity (MME), Home Subscriber Server (HSS), Access and Mobility Management Function (AMF), Session Management Function (SMF), Authentication Server Function (AUSF), Subscription Identifier De-concealing function P111793WO01 (017997.4195) PATENT APPLICATION

[0446] 45

[0447] (SIDF), Unified Data Management (UDM), Security Edge Protection Proxy (SEPP), Network Exposure Function (NEF), and / or a User Plane Function (UPF).

[0448] The host 1316 may be under the ownership or control of a service provider other than an operator or provider of the access network 1304 and / or the telecommunication network 1302, and may be operated by the service provider or on behalf of the service provider. The host 1316 may host a variety of applications to provide one or more service. Examples of such applications include live and pre-recorded audio / video content, data collection services such as retrieving and compiling data on various ambient conditions detected by a plurality of UEs, analytics functionality, social media, functions for controlling or otherwise interacting with remote devices, functions for an alarm and surveillance center, or any other such function performed by a server.

[0449] As a whole, the communication system 1300 of FIGURE 11 enables connectivity between the UEs, network nodes, and hosts. In that sense, the communication system may be configured to operate according to predefined rules or procedures, such as specific standards that include, but are not limited to: Global System for Mobile Communications (GSM); Universal Mobile Telecommunications System (UMTS); Long Term Evolution (LTE), and / or other suitable 2G, 3G, 4G, 5G standards, or any applicable future generation standard (e.g., 6G); wireless local area network (WLAN) standards, such as the Institute of Electrical and Electronics Engineers (IEEE) 802.11 standards (WiFi); and / or any other appropriate wireless communication standard, such as the Worldwide Interoperability for Microwave Access (WiMax), Bluetooth, Z-Wave, Near Field Communication (NFC) ZigBee, LiFi, and / or any low-power wide-area network (LPWAN) standards such as LoRa and Sigfox.

[0450] In some examples, the telecommunication network 1302 is a cellular network that implements 3GPP standardized features. Accordingly, the telecommunications network 1302 may support network slicing to provide different logical networks to different devices that are connected to the telecommunication network 1302. For example, the telecommunications network 1302 may provide Ultra Reliable Low Latency Communication (URLLC) services to some UEs, while providing Enhanced Mobile Broadband (eMBB) services to other UEs, and / or Massive Machine Type Communication (mMTC)ZMassive loT services to yet further UEs.

[0451] In some examples, the UEs 1312 are configured to transmit and / or receive information without direct human interaction. For instance, a UE may be designed to transmit information to the access network 1304 on a predetermined schedule, when triggered by an internal or external event, or in response to requests from the access network 1304. Additionally, a UE may be P111793WO01 (017997.4195) PATENT APPLICATION

[0452] 46 configured for operating in single- or multi-RAT or multi-standard mode. For example, a UE may operate with any one or combination of Wi-Fi, NR (New Radio) and LTE, i.e. being configured for multi -radio dual connectivity (MR-DC), such as E-UTRAN (Evolved-UMTS Terrestrial Radio Access Network) New Radio - Dual Connectivity (EN-DC).

[0453] In the example, the hub 1314 communicates with the access network 1304 to facilitate indirect communication between one or more UEs (e.g., UE 1312c and / or 1312d) and network nodes (e.g., network node 1310b). In some examples, the hub 1314 may be a controller, router, content source and analytics, or any of the other communication devices described herein regarding UEs. For example, the hub 1314 may be a broadband router enabling access to the core network 1306 for the UEs. As another example, the hub 1314 may be a controller that sends commands or instructions to one or more actuators in the UEs. Commands or instructions may be received from the UEs, network nodes 1310, or by executable code, script, process, or other instructions in the hub 1314. As another example, the hub 1314 may be a data collector that acts as temporary storage for UE data and, in some embodiments, may perform analysis or other processing of the data. As another example, the hub 1314 may be a content source. For example, for a UE that is a VR headset, display, loudspeaker or other media delivery device, the hub 1314 may retrieve VR assets, video, audio, or other media or data related to sensory information via a network node, which the hub 1314 then provides to the UE either directly, after performing local processing, and / or after adding additional local content. In still another example, the hub 1314 acts as a proxy server or orchestrator for the UEs, in particular in if one or more of the UEs are low energy loT devices.

[0454] The hub 1314 may have a constant / persi stent or intermittent connection to the network node 1310b. The hub 1314 may also allow for a different communication scheme and / or schedule between the hub 1314 and UEs (e.g., UE 1312c and / or 1312d), and between the hub 1314 and the core network 1306. In other examples, the hub 1314 is connected to the core network 1306 and / or one or more UEs via a wired connection. Moreover, the hub 1314 may be configured to connect to an M2M service provider over the access network 1304 and / or to another UE over a direct connection. In some scenarios, UEs may establish a wireless connection with the network nodes 1310 while still connected via the hub 1314 via a wired or wireless connection. In some embodiments, the hub 1314 may be a dedicated hub - that is, a hub whose primary function is to route communications to / from the UEs from / to the network node 1310b. In other embodiments, the hub 1314 may be a non-dedicated hub - that is, a device which is capable of operating to route P111793WO01 (017997.4195) PATENT APPLICATION

[0455] 47 communications between the UEs and network node 1310b, but which is additionally capable of operating as a communication start and / or end point for certain data channels.

[0456] FIGURE 12 shows a UE 1400 in accordance with some embodiments. As used herein, a UE refers to a device capable, configured, arranged and / or operable to communicate wirelessly with network nodes and / or other UEs. Examples of a UE include, but are not limited to, a smart phone, mobile phone, cell phone, voice over IP (VoIP) phone, wireless local loop phone, desktop computer, personal digital assistant (PDA), wireless cameras, gaming console or device, music storage device, playback appliance, wearable terminal device, wireless endpoint, mobile station, tablet, laptop, laptop-embedded equipment (LEE), laptop-mounted equipment (LME), smart device, wireless customer-premise equipment (CPE), vehicle-mounted or vehicle embedded / integrated wireless device, etc. Other examples include any UE identified by the 3rd Generation Partnership Project (3GPP), including a narrow band internet of things (NB-IoT) UE, a machine type communication (MTC) UE, and / or an enhanced MTC (eMTC) UE.

[0457] A UE may support device -to-device (D2D) communication, for example by implementing a 3GPP standard for sidelink communication, Dedicated Short-Range Communication (DSRC), vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), or vehicle-to-everything (V2X). In other examples, a UE may not necessarily have a user in the sense of a human user who owns and / or operates the relevant device. Instead, a UE may represent a device that is intended for sale to, or operation by, a human user but which may not, or which may not initially, be associated with a specific human user (e.g., a smart sprinkler controller). Alternatively, a UE may represent a device that is not intended for sale to, or operation by, an end user but which may be associated with or operated for the benefit of a user (e.g., a smart power meter).

[0458] The UE 1400 includes processing circuitry 1402 that is operatively coupled via a bus 1404 to an input / output interface 1406, a power source 1408, amemory 1410, a communication interface 1412, and / or any other component, or any combination thereof. Certain UEs may utilize all or a subset of the components shown in FIGURE 12. The level of integration between the components may vary from one UE to another UE. Further, certain UEs may contain multiple instances of a component, such as multiple processors, memories, transceivers, transmitters, receivers, etc.

[0459] The processing circuitry 1402 is configured to process instructions and data and may be configured to implement any sequential state machine operative to execute instructions stored as machine-readable computer programs in the memory 1410. The processing circuitry 1402 may be implemented as one or more hardware -implemented state machines (e.g., in discrete logic, field- P111793WO01 (017997.4195) PATENT APPLICATION

[0460] 48 programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), etc.); programmable logic together with appropriate firmware; one or more stored computer programs, general-purpose processors, such as a microprocessor or digital signal processor (DSP), together with appropriate software; or any combination of the above. For example, the processing circuitry 1402 may include multiple central processing units (CPUs).

[0461] In the example, the input / output interface 1406 may be configured to provide an interface or interfaces to an input device, output device, or one or more input and / or output devices. Examples of an output device include a speaker, a sound card, a video card, a display, a monitor, a printer, an actuator, an emitter, a smartcard, another output device, or any combination thereof. An input device may allow a user to capture information into the UE 1400. Examples of an input device include a touch-sensitive or presence-sensitive display, a camera (e.g., a digital camera, a digital video camera, a web camera, etc.), a microphone, a sensor, a mouse, a trackball, a directional pad, a trackpad, a scroll wheel, a smartcard, and the like. The presence-sensitive display may include a capacitive or resistive touch sensor to sense input from a user. A sensor may be, for instance, an accelerometer, a gyroscope, a tilt sensor, a force sensor, a magnetometer, an optical sensor, a proximity sensor, a biometric sensor, etc., or any combination thereof. An output device may use the same type of interface port as an input device. For example, a Universal Serial Bus (USB) port may be used to provide an input device and an output device.

[0462] In some embodiments, the power source 1408 is structured as a battery or battery pack. Other types of power sources, such as an external power source (e.g., an electricity outlet), photovoltaic device, or power cell, may be used. The power source 1408 may further include power circuitry for delivering power from the power source 1408 itself, and / or an external power source, to the various parts of the UE 1400 via input circuitry or an interface such as an electrical power cable. Delivering power may be, for example, for charging of the power source 1408. Power circuitry may perform any formatting, converting, or other modification to the power from the power source 1408 to make the power suitable for the respective components of the UE 1400 to which power is supplied.

[0463] The memory 1410 may be or be configured to include memory such as random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic disks, optical disks, hard disks, removable cartridges, flash drives, and so forth. In one example, the memory 1410 includes one or more application programs 1414, P111793WO01 (017997.4195) PATENT APPLICATION

[0464] 49 such as an operating system, web browser application, a widget, gadget engine, or other application, and corresponding data 1416. The memory 1410 may store, for use by the UE 1400, any of a variety of various operating systems or combinations of operating systems.

[0465] The memory 1410 may be configured to include a number of physical drive units, such as redundant array of independent disks (RAID), flash memory, USB flash drive, external hard disk drive, thumb drive, pen drive, key drive, high-density digital versatile disc (HD-DVD) optical disc drive, internal hard disk drive, Blu-Ray optical disc drive, holographic digital data storage (HDDS) optical disc drive, external mini-dual in-line memory module (DIMM), synchronous dynamic random access memory (SDRAM), external micro-DIMM SDRAM, smartcard memory such as tamper resistant module in the form of a universal integrated circuit card (UICC) including one or more subscriber identity modules (SIMs), such as a USIM and / or ISIM, other memory, or any combination thereof. The UICC may for example be an embedded UICC (eUICC), integrated UICC (iUICC) or a removable UICC commonly known as ‘SIM card.’ The memory 1410 may allow the UE 1400 to access instructions, application programs and the like, stored on transitory or non-transitory memory media, to off-load data, or to upload data. An article of manufacture, such as one utilizing a communication system may be tangibly embodied as or in the memory 1410, which may be or comprise a device-readable storage medium.

[0466] The processing circuitry 1402 may be configured to communicate with an access network or other network using the communication interface 1412. The communication interface 1412 may comprise one or more communication subsystems and may include or be communicatively coupled to an antenna 1422. The communication interface 1412 may include one or more transceivers used to communicate, such as by communicating with one or more remote transceivers of another device capable of wireless communication (e.g., another UE or a network node in an access network). Each transceiver may include a transmitter 1418 and / or a receiver 1420 appropriate to provide network communications (e.g., optical, electrical, frequency allocations, and so forth). Moreover, the transmitter 1418 and receiver 1420 may be coupled to one or more antennas (e.g., antenna 1422) and may share circuit components, software or firmware, or alternatively be implemented separately.

[0467] In the illustrated embodiment, communication functions of the communication interface 1412 may include cellular communication, Wi-Fi communication, LPWAN communication, data communication, voice communication, multimedia communication, short-range communications such as Bluetooth, near-field communication, location-based communication such as the use of P111793WO01 (017997.4195) PATENT APPLICATION

[0468] 50 the global positioning system (GPS) to determine a location, another like communication function, or any combination thereof. Communications may be implemented in according to one or more communication protocols and / or standards, such as IEEE 802.11, Code Division Multiplexing Access (CDMA), Wideband Code Division Multiple Access (WCDMA), GSM, LTE, New Radio (NR), UMTS, WiMax, Ethernet, transmission control protocol / intemet protocol (TCP / IP), synchronous optical networking (SONET), Asynchronous Transfer Mode (ATM), QUIC, Hypertext Transfer Protocol (HTTP), and so forth.

[0469] Regardless of the type of sensor, a UE may provide an output of data captured by its sensors, through its communication interface 1 412, via a wireless connection to a network node. Data captured by sensors of a UE can be communicated through a wireless connection to a network node via another UE. The output may be periodic (e.g., once every 15 minutes if it reports the sensed temperature), random (e.g., to even out the load from reporting from several sensors), in response to a triggering event (e.g., when moisture is detected an alert is sent), in response to a request (e.g., a user initiated request), or a continuous stream (e.g., a live video feed of a patient).

[0470] As another example, a UE comprises an actuator, a motor, or a switch, related to a communication interface configured to receive wireless input from a network node via a wireless connection. In response to the received wireless input the states of the actuator, the motor, or the switch may change. For example, the UE may comprise a motor that adjusts the control surfaces or rotors of a drone in flight according to the received input or to a robotic arm performing a medical procedure according to the received input.

[0471] A UE, when in the form of an Internet of Things (loT) device, may be a device for use in one or more application domains, these domains comprising, but not limited to, city wearable technology, extended industrial application and healthcare. Non-limiting examples of such an loT device are a device which is or which is embedded in: a connected refrigerator or freezer, a TV, a connected lighting device, an electricity meter, a robot vacuum cleaner, a voice controlled smart speaker, a home security camera, a motion detector, a thermostat, a smoke detector, a door / window sensor, a flood / moisture sensor, an electrical door lock, a connected doorbell, an air conditioning system like a heat pump, an autonomous vehicle, a surveillance system, a weather monitoring device, a vehicle parking monitoring device, an electric vehicle charging station, a smart watch, a fitness tracker, a head-mounted display for Augmented Reality (AR) or Virtual Reality (VR), a wearable for tactile augmentation or sensory enhancement, a water sprinkler, an animal- or itemtracking device, a sensor for monitoring a plant or animal, an industrial robot, an Unmanned Aerial P111793WO01 (017997.4195) PATENT APPLICATION

[0472] 51

[0473] Vehicle (UAV), and any kind of medical device, like a heart rate monitor or a remote controlled surgical robot. A UE in the form of an loT device comprises circuitry and / or software in dependence of the intended application of the loT device in addition to other components as described in relation to the UE 1400 shown in FIGURE 12.

[0474] As yet another specific example, in an loT scenario, a UE may represent a machine or other device that performs monitoring and / or measurements, and transmits the results of such monitoring and / or measurements to another UE and / or a network node. The UE may in this case be an M2M device, which may in a 3GPP context be referred to as an MTC device. As one particular example, the UE may implement the 3GPP NB-IoT standard. In other scenarios, a UE may represent a vehicle, such as a car, a bus, a truck, a ship and an airplane, or other equipment that is capable of monitoring and / or reporting on its operational status or other functions associated with its operation.

[0475] In practice, any number of UEs may be used together with respect to a single use case. For example, a first UE might be or be integrated in a drone and provide the drone’s speed information (obtained through a speed sensor) to a second UE that is a remote controller operating the drone. When the user makes changes from the remote controller, the first UE may adjust the throttle on the drone (e.g. by controlling an actuator) to increase or decrease the drone’s speed. The first and / or the second UE can also include more than one of the functionalities described above. For example, a UE might comprise the sensor and the actuator, and handle communication of data for both the speed sensor and the actuators.

[0476] FIGURE 13 shows a network node 1500 in accordance with some embodiments. As used herein, network node refers to equipment capable, configured, arranged and / or operable to communicate directly or indirectly with a UE and / or with other network nodes or equipment, in a telecommunication network. Examples of network nodes include, but are not limited to, access points (APs) (e.g., radio access points), base stations (BSs) (e.g., radio base stations, Node Bs, evolved Node Bs (eNBs) and NRNodeBs (gNBs)).

[0477] Base stations may be categorized based on the amount of coverage they provide (or, stated differently, their transmit power level) and so, depending on the provided amount of coverage, may be referred to as femto base stations, pico base stations, micro base stations, or macro base stations. A base station may be a relay node or a relay donor node controlling a relay. A network node may also include one or more (or all) parts of a distributed radio base station such as centralized digital units and / or remote radio units (RRUs), sometimes referred to as Remote Radio P111793WO01 (017997.4195) PATENT APPLICATION

[0478] 52

[0479] Heads (RRHs). Such remote radio units may or may not be integrated with an antenna as an antenna integrated radio. Parts of a distributed radio base station may also be referred to as nodes in a distributed antenna system (DAS).

[0480] Other examples of network nodes include multiple transmission point (multi-TRP) 5G access nodes, multi-standard radio (MSR) equipment such as MSR BSs, network controllers such as radio network controllers (RNCs) or base station controllers (BSCs), base transceiver stations (BTSs), transmission points, transmission nodes, multi-cell / multicast coordination entities (MCEs), Operation and Maintenance (O&M) nodes, Operations Support System (OSS) nodes, Self-Organizing Network (SON) nodes, positioning nodes (e.g., Evolved Serving Mobile Location Centers (E-SMLCs)), and / or Minimization of Drive Tests (MDTs).

[0481] The network node 1500 includes a processing circuitry 1502, a memory 1504, a communication interface 1506, and a power source 1508. The network node 1500 may be composed of multiple physically separate components (e.g., a NodeB component and a RNC component, or a BTS component and a BSC component, etc.), which may each have their own respective components. In certain scenarios in which the network node 1500 comprises multiple separate components (e.g., BTS and BSC components), one or more of the separate components may be shared among several network nodes. For example, a single RNC may control multiple NodeBs. In such a scenario, each unique NodeB and RNC pair, may in some instances be considered a single separate network node. In some embodiments, the network node 1500 may be configured to support multiple radio access technologies (RATs). In such embodiments, some components may be duplicated (e.g., separate memory 1504 for different RATs) and some components may be reused (e.g., a same antenna 1510 may be shared by different RATs). The network node 1500 may also include multiple sets of the various illustrated components for different wireless technologies integrated into network node 1500, for example GSM, WCDMA, LTE, NR, WiFi, Zigbee, Z-wave, LoRaWAN, Radio Frequency Identification (RFID) or Bluetooth wireless technologies. These wireless technologies may be integrated into the same or different chip or set of chips and other components within network node 1500.

[0482] The processing circuitry 1502 may comprise a combination of one or more of a microprocessor, controller, microcontroller, central processing unit, digital signal processor, application-specific integrated circuit, field programmable gate array, or any other suitable computing device, resource, or combination of hardware, software and / or encoded logic operable P111793WO01 (017997.4195) PATENT APPLICATION

[0483] 53 to provide, either alone or in conjunction with other network node 1500 components, such as the memory 1504, to provide network node 1500 functionality.

[0484] In some embodiments, the processing circuitry 1502 includes a system on a chip (SOC). In some embodiments, the processing circuitry 1502 includes one or more of radio frequency (RF) transceiver circuitry 1512 and baseband processing circuitry 1514. In some embodiments, the radio frequency (RF) transceiver circuitry 1512 and the baseband processing circuitry 1514 may be on separate chips (or sets of chips), boards, or units, such as radio units and digital units. In alternative embodiments, part or all of RF transceiver circuitry 1512 and baseband processing circuitry 1514 may be on the same chip or set of chips, boards, or units.

[0485] The memory 1504 may comprise any form of volatile or non-volatile computer-readable memory including, without limitation, persistent storage, solid-state memory, remotely mounted memory, magnetic media, optical media, random access memory (RAM), read-only memory (ROM), mass storage media (for example, a hard disk), removable storage media (for example, a flash drive, a Compact Disk (CD) or a Digital Video Disk (DVD)), and / or any other volatile or non-volatile, non-transitory device-readable and / or computer-executable memory devices that store information, data, and / or instructions that may be used by the processing circuitry 1502. The memory 1504 may store any suitable instructions, data, or information, including a computer program, software, an application including one or more of logic, rules, code, tables, and / or other instructions capable of being executed by the processing circuitry 1502 and utilized by the network node 1500. The memory 1504 may be used to store any calculations made by the processing circuitry 1502 and / or any data received via the communication interface 1506. In some embodiments, the processing circuitry 1502 and memory 1504 is integrated.

[0486] The communication interface 1506 is used in wired or wireless communication of signaling and / or data between a network node, access network, and / or UE. As illustrated, the communication interface 1506 comprises port(s) / terminal(s) 1516 to send and receive data, for example to and from a network over a wired connection. The communication interface 1506 also includes radio front-end circuitry 1518 that may be coupled to, or in certain embodiments a part of, the antenna 1510. Radio front-end circuitry 1518 comprises filters 1520 and amplifiers 1522. The radio frontend circuitry 1518 may be connected to an antenna 1510 and processing circuitry 1502. The radio front-end circuitry may be configured to condition signals communicated between antenna 1510 and processing circuitry 1502. The radio front-end circuitry 1518 may receive digital data that is to be sent out to other network nodes or UEs via a wireless connection. The radio front-end P111793WO01 (017997.4195) PATENT APPLICATION

[0487] 54 circuitry 1518 may convert the digital data into a radio signal having the appropriate channel and bandwidth parameters using a combination of filters 1520 and / or amplifiers 1522. The radio signal may then be transmitted via the antenna 1510. Similarly, when receiving data, the antenna 1510 may collect radio signals which are then converted into digital data by the radio front-end circuitry 1518. The digital data may be passed to the processing circuitry 1502. In other embodiments, the communication interface may comprise different components and / or different combinations of components.

[0488] In certain alternative embodiments, the network node 1500 does not include separate radio front-end circuitry 1518, instead, the processing circuitry 1502 includes radio front-end circuitry and is connected to the antenna 1510. Similarly, in some embodiments, all or some of the RF transceiver circuitry 1512 is part of the communication interface 1506. In still other embodiments, the communication interface 1506 includes one or more ports or terminals 1516, the radio frontend circuitry 1518, and the RF transceiver circuitry 1512, as part of a radio unit (not shown), and the communication interface 1506 communicates with the baseband processing circuitry 1514, which is part of a digital unit (not shown).

[0489] The antenna 1510 may include one or more antennas, or antenna arrays, configured to send and / or receive wireless signals. The antenna 1510 may be coupled to the radio front-end circuitry 1518 and may be any type of antenna capable of transmitting and receiving data and / or signals wirelessly. In certain embodiments, the antenna 1510 is separate from the network node 1500 and connectable to the network node 1500 through an interface or port.

[0490] The antenna 1510, communication interface 1506, and / or the processing circuitry 1502 may be configured to perform any receiving operations and / or certain obtaining operations described herein as being performed by the network node. Any information, data and / or signals may be received from a UE, another network node and / or any other network equipment. Similarly, the antenna 1510, the communication interface 1506, and / or the processing circuitry 1502 may be configured to perform any transmitting operations described herein as being performed by the network node. Any information, data and / or signals may be transmitted to a UE, another network node and / or any other network equipment.

[0491] The power source 1508 provides power to the various components of network node 1500 in a form suitable for the respective components (e.g., at a voltage and current level needed for each respective component). The power source 1508 may further comprise, or be coupled to, power management circuitry to supply the components of the network node 1500 with power for P111793WO01 (017997.4195) PATENT APPLICATION

[0492] 55 performing the functionality described herein. For example, the network node 1500 may be connectable to an external power source (e.g., the power grid, an electricity outlet) via an input circuitry or interface such as an electrical cable, whereby the external power source supplies power to power circuitry of the power source 1508. As a further example, the power source 1508 may comprise a source of power in the form of a battery or battery pack which is connected to, or integrated in, power circuitry. The battery may provide backup power should the external power source fail.

[0493] Embodiments of the network node 1500 may include additional components beyond those shown in FIGURE 13 for providing certain aspects of the network node’s functionality, including any of the functionality described herein and / or any functionality necessary to support the subject matter described herein. For example, the network node 1500 may include user interface equipment to allow input of information into the network node 1500 and to allow output of information from the network node 1500. This may allow a user to perform diagnostic, maintenance, repair, and other administrative functions for the network node 1500.

[0494] FIGURE 14 is a block diagram illustrating a virtualization environment 1600 in which functions implemented by some embodiments may be virtualized. In the present context, virtualizing means creating virtual versions of apparatuses or devices which may include virtualizing hardware platforms, storage devices and networking resources. As used herein, virtualization can be applied to any device described herein, or components thereof, and relates to an implementation in which at least a portion of the functionality is implemented as one or more virtual components. Some or all of the functions described herein may be implemented as virtual components executed by one or more virtual machines (VMs) implemented in one or more virtual environments 1600 hosted by one or more of hardware nodes, such as a hardware computing device that operates as a network node, UE, core network node, or host. Further, in embodiments in which the virtual node does not require radio connectivity (e.g., a core network node or host), then the node may be entirely virtualized. In some embodiments, the virtualization environment 1600 includes components defined by the O-RAN Alliance, such as an O-Cloud environment orchestrated by a Service Management and Orchestration Framework via an O-2 interface. Virtualization may facilitate distributed implementations of a network node, UE, core network node, or host.

[0495] Applications 1602 (which may alternatively be called software instances, virtual appliances, network functions, virtual nodes, virtual network functions, etc.) are run in the P111793WO01 (017997.4195) PATENT APPLICATION

[0496] 56 virtualization environment 1600 to implement some of the features, functions, and / or benefits of some of the embodiments disclosed herein.

[0497] Hardware 1604 includes processing circuitry, memory that stores software and / or instructions executable by hardware processing circuitry, and / or other hardware devices as described herein, such as a network interface, input / output interface, and so forth. Software may be executed by the processing circuitry to instantiate one or more virtualization layers 1606 (also referred to as hypervisors or virtual machine monitors (VMMs)), provide VMs 1608a and 1608b (one or more of which may be generally referred to as VMs 1608), and / or perform any of the functions, features and / or benefits described in relation with some embodiments described herein. The virtualization layer 1606 may present a virtual operating platform that appears like networking hardware to the VMs 1608.

[0498] The VMs 1608 comprise virtual processing, virtual memory, virtual networking or interface and virtual storage, and may be run by a corresponding virtualization layer 1606. Different embodiments of the instance of a virtual appliance 1602 may be implemented on one or more of VMs 1608, and the implementations may be made in different ways. Virtualization of the hardware is in some contexts referred to as network function virtualization (NFV). NFV may be used to consolidate many network equipment types onto industry standard high volume server hardware, physical switches, and physical storage, which can be located in data centers, and customer premise equipment.

[0499] In the context of NFV, a VM 1608 may be a software implementation of a physical machine that runs programs as if they were executing on a physical, non-virtualized machine. Each of the VMs 1608, and that part of hardware 1604 that executes that VM, be it hardware dedicated to that VM and / or hardware shared by that VM with others of the VMs, forms separate virtual network elements. Still in the context of NFV, a virtual network function is responsible for handling specific network functions that run in one or more VMs 1608 on top of the hardware 1604 and corresponds to the application 1602.

[0500] Hardware 1604 may be implemented in a standalone network node with generic or specific components. Hardware 1604 may implement some functions via virtualization. Alternatively, hardware 1604 may be part of a larger cluster of hardware (e.g. such as in a data center or CPE) where many hardware nodes work together and are managed via management and orchestration 1610, which, among others, oversees lifecycle management of applications 1602. In some embodiments, hardware 1604 is coupled to one or more radio units that each include one or more P111793WO01 (017997.4195) PATENT APPLICATION

[0501] 57 transmiters and one or more receivers that may be coupled to one or more antennas. Radio units may communicate directly with other hardware nodes via one or more appropriate network interfaces and may be used in combination with the virtual components to provide a virtual node with radio capabilities, such as a radio access node or a base station. In some embodiments, some signaling can be provided with the use of a control system 1612 which may alternatively be used for communication between hardware nodes and radio units.

[0502] Although the computing devices described herein (e.g., UEs, network nodes, hosts) may include the illustrated combination of hardware components, other embodiments may comprise computing devices with different combinations of components. It is to be understood that these computing devices may comprise any suitable combination of hardware and / or software needed to perform the tasks, features, functions and methods disclosed herein. Determining, calculating, obtaining or similar operations described herein may be performed by processing circuitry, which may process information by, for example, converting the obtained information into other information, comparing the obtained information or converted information to information stored in the network node, and / or performing one or more operations based on the obtained information or converted information, and as a result of said processing making a determination. Moreover, while components are depicted as single boxes located within a larger box, or nested within multiple boxes, in practice, computing devices may comprise multiple different physical components that make up a single illustrated component, and functionality may be partitioned between separate components. For example, a communication interface may be configured to include any of the components described herein, and / or the functionality of the components may be partitioned between the processing circuitry and the communication interface. In another example, non-computationally intensive functions of any of such components may be implemented in software or firmware and computationally intensive functions may be implemented in hardware.

[0503] In certain embodiments, some or all of the functionality described herein may be provided by processing circuitry executing instructions stored on in memory, which in certain embodiments may be a computer program product in the form of a non-transitory computer-readable storage medium. In alternative embodiments, some or all of the functionality may be provided by the processing circuitry without executing instructions stored on a separate or discrete device-readable storage medium, such as in a hard-wired manner. In any of those particular embodiments, whether executing instructions stored on a non-transitory computer-readable storage medium or not, the P111793WO01 (017997.4195) PATENT APPLICATION

[0504] 58 processing circuitry can be configured to perform the described functionality. The benefits provided by such functionality are not limited to the processing circuitry alone or to other components of the computing device, but are enjoyed by the computing device as a whole, and / or by end users and a wireless network generally.

[0505] EXAMPLE EMBODIMENTS

[0506] Group A Example Embodiments

[0507] Example Embodiment 1. A method (100) by first node using Vertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, the method comprising: receiving (102), from a second node, information comprising and / or indicating at least one of: a type of the Al and / or ML model (e.g., NN-networks), interoperability information indicating a type of training method that is supported by the second node for the Al and / or ML model (e.g., NN-based ML Model is accepted / can be trained), and at least one intermediate result.

[0508] Example Embodiment 2. The method of Example Embodiment 1, comprising: using the information to train the Al and / or ML model.

[0509] Example Embodiment 3. The method of any one of Example Embodiments 1 to 2, wherein the Al and / or ML model is neural network-based.

[0510] Example Embodiment 4. The method of any one of Example Embodiments 1 to 3, wherein the interoperability information indicating the type of training method indicates whether the second node supports Vertical Federated Learning and / or Horizontal Federated Learning.

[0511] Example Embodiment 5. The method of any one of Example Embodiments 1 to 4, wherein the interoperability information indicates the type of training method that is supported by the second node for the Al and / or ML model and / or a that is agreed to by the second node and the first node.

[0512] Example Embodiment 6. The method of any one of Example Embodiments 1 to 4, wherein the information and / or interoperabililty information comprises at least one parameter associated with an intermediate matrix to be exposed.

[0513] Example Embodiment 7. The method of Example Embodiment 6, wherein the at least one parameter associated with the intermediate matrix comprises at least one dimension of the intermediate matrix. P111793WO01 (017997.4195) PATENT APPLICATION

[0514] 59

[0515] Example Embodiment 8. The method of Example Embodiment 7, wherein the at least one dimension of the intermediate matrix comprises at least one of: a number of a plurality of nodes in an intermediate matrix; a content of each of a plurality of nodes of an intermediate matrix; and a type of content of each of the plurality of the nodes in the intermediate matrix.

[0516] Example Embodiment 9. The method of any one of Example Embodiments 6 to 8, wherein the at least one parameter associated with the intermediate matrix comprises a number of activations per batch of samples in a container.

[0517] Example Embodiment 10. The method of any one of Example Embodiments 1 to 9, wherein the information and / or interoperabililty information comprises at least one of: analytics identifier, a notification target address, at least one parameter associated with the at least one intermediate result associated with the Al and / or ML model, a number of samples, a number of immediate results per sample, an indicator of at least one gradient associated with the at least one intermediate result, an indication of a number of gradients per batch of samples in an intermediate matrix container that includes the intermediate matrix.

[0518] Example Embodiment 11. The method of any one of Example Embodiments 1 to 10, wherein the information and / or interoperabililty information comprises at least one of: a notification correlation identifier, the Al and / or ML model that is accepted and / or can be trained by the VFL client, a number of nodes for the intermediate matrix that is accepted by the VFL client, a number of samples accepted by the VFL client, a type of content that is accepted by the VFL client for each of the plurality of nodes, an indicator of at least one activation provided in the intermediate matrix, and an indication of a number of activations per batch of samples in an intermediate matrix container that includes the intermediate matrix.

[0519] Example Embodiment 12. The method of any one of Example Embodiments 1 to 11, wherein the information and / or interoperabililty information comprises at least one of: at least one analytics identifier, at least one sample identifier, at least one notification target address, at least one notification correlation identifier, and at least one analytics reporting parameter.

[0520] Example Embodiment 13. The method of any one of Example Embodiments 1 to 12, wherein: the first node comprises a NRF, and the second node comprises a NF.

[0521] Example Embodiment 14. The method of Example Embodiment 13, wherein the information comprises Nnrf NFManagement NFRegister service operation information. P111793WO01 (017997.4195) PATENT APPLICATION

[0522] 60

[0523] Example Embodiment 15. The method of any one of Example Embodiments 1 to 12, wherein: the first node is operating as a VFL client, and the second node is operating as a VFL server.

[0524] Example Embodiment 16. The method of Example Embodiment 15, wherein the information comprises Nnrf MLModelTraining Sub scribe service operation information.

[0525] Example Embodiment 17. The method of any one of Example Embodiments 1 to 12, wherein: the first node is operating as a VFL server, and the second node is operating as a VFL client.

[0526] Example Embodiment 18. The method of Example Embodiment 17, wherein the information comprises Nnwdaf MLModelTraining Notify service operation information.

[0527] Example Embodiment 19. The method of any one of Example Embodiments 1 to 12, wherein: the first node comprises aNWDAF comprising an AnLF, and the second node comprises a consumer.

[0528] Example Embodiment 20. The method of Example Embodiment 19, wherein the information comprises Nnwdaf AnalyticsSubscription Subscribe operation information.

[0529] Example Embodiment 21. The method of any one of Example Embodiments 1 to 12, wherein: the first node comprises a consumer, and the second node comprises a NWDAF comprising an AnLF.

[0530] Example Embodiment 22. The method of Example Embodiment 21, wherein the information comprises Nnwdaf AnalyticsSubscription Notify operation information.

[0531] Example Embodiment 23. The method of any one of Example Embodiments 1 to 22, wherein the first node or second node operating as the VFL client comprises a User Equipment or a gNodeB.

[0532] Example Embodiment 24. The method of any one of Example Embodiments 1 to 23, wherein the information is received during a registration procedure for registering the first node with the second node.

[0533] Example Embodiment 25. The method of any one of Example Embodiments 1 to 24, wherein the information is received in response to or with a Federated Learning preparation request.

[0534] Example Embodiment 26. The method of Example Embodiment 25, comprising: determining one or more federated learning parameters that the first node supports, and P111793WO01 (017997.4195) PATENT APPLICATION

[0535] 61 transmitting, to the second node, a Federated Learning response indicating the one or more Federated Learning parameters that the first node supports.

[0536] Group B Embodiments

[0537] Example Embodiment 27. A method (200) by second node using Vertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, the method comprising: transmitting (202), to the first node, information comprising and / or indicating at least one of: a type of the Al and / or ML model (e.g., NN-networks), interoperability information indicating a type of training method that is supported by the second node for the Al and / or ML model (e.g., NN -based ML Model is accepted / can be trained), and at least one intermediate result.

[0538] Example Embodiment 28. The method of Claim 27, comprising determining that the Al and / or ML model requires the type of training, and wherein the information is transmitted to the first node based on determining that the Al and / or ML model requires the type of training.

[0539] Example Embodiment 29. The method of any one of Example Embodiments 27 to 28, comprising: using the information to train the Al and / or ML model.

[0540] Example Embodiment 30. The method of any one of Example Embodiments 27 to 29, wherein the Al and / or ML model is neural network-based.

[0541] Example Embodiment 31. The method of any one of Example Embodiments 27 to 30, wherein the interoperability information indicating the type of training method indicates whether the second node supports Vertical Federated Learning and / or Horizontal Federated Learning.

[0542] Example Embodiment 32. The method of any one of Example Embodiments 27 to 31, wherein the interoperability information indicates the type of training method that is supported by the second node for the Al and / or ML model and / or a that is agreed to by the second node and the first node.

[0543] Example Embodiment 33. The method of any one of Example Embodiments 27 to 32, wherein the information and / or interoperabililty information comprises at least one parameter associated with an intermediate matrix to be exposed.

[0544] Example Embodiment 34. The method of Example Embodiment 33, wherein the at least one parameter associated with the intermediate matrix comprises at least one dimension of the intermediate matrix.

[0545] Example Embodiment 35. The method of Example Embodiment 34, wherein the at least one dimension of the intermediate matrix comprises at least one of: a number of a plurality of P111793WO01 (017997.4195) PATENT APPLICATION

[0546] 62 nodes in an intermediate matrix; a content of each of a plurality of nodes of an intermediate matrix; and a type of content of each of the plurality of the nodes in the intermediate matrix.

[0547] Example Embodiment 36. The method of any one of Example Embodiments 33 to 35, wherein the at least one parameter associated with the intermediate matrix comprises a number of activations per batch of samples in a container.

[0548] Example Embodiment 37. The method of any one of Example Embodiments 27 to 36, wherein the information and / or interoperabililty information comprises at least one of: analytics identifier, a notification target address, at least one parameter associated with the at least one intermediate result associated with the Al and / or ML model, a number of samples, a number of immediate results per sample, an indicator of at least one gradient associated with the at least one intermediate result, and an indication of a number of gradients per batch of samples in an intermediate matrix container that includes the intermediate matrix.

[0549] Example Embodiment 38. The method of any one of Example Embodiments 27 to 37, wherein the information and / or interoperabililty information comprises at least one of: a notification correlation identifier, the Al and / or ML model that is accepted and / or can be trained by the VFL client, a number of nodes for the intermediate matrix that is accepted by the VFL client, a number of samples accepted by the VFL client, a type of content that is accepted by the VFL client for each of the plurality of nodes, an indicator of at least one activation provided in the intermediate matrix, and an indication of a number of activations per batch of samples in an intermediate matrix container that includes the intermediate matrix.

[0550] Example Embodiment 39. The method of any one of Example Embodiments 27 to 38, wherein the information and / or interoperabililty information comprises at least one of: at least one analytics identifier, at least one sample identifier, at least one notification target address, at least one notification correlation identifier, and at least one analytics reporting parameter.

[0551] Example Embodiment 40. The method of any one of Example Embodiments 27 to 39, wherein: the first node comprises a NRF, and the second node comprises a NF.

[0552] Example Embodiment 41. The method of Example Embodiment 40, wherein the information comprises Nnrf NFManagement NFRegister service operation information.

[0553] Example Embodiment 42. The method of any one of Example Embodiments 27 to 39, wherein: the first node is operating as a VFL client, and the second node is operating as a VFL server. P111793WO01 (017997.4195) PATENT APPLICATION

[0554] 63

[0555] Example Embodiment 43. The method of Example Embodiment 42, wherein the information comprises Nnrf MLModelTraining Sub scribe service operation information.

[0556] Example Embodiment 44. The method of any one of Example Embodiments 27 to 39, wherein: the first node is operating as a VFL server, and the second node is operating as a VFL client.

[0557] Example Embodiment 45. The method of Example Embodiment 44, wherein the information comprises Nnwdaf MLModelTraining Notify service operation information.

[0558] Example Embodiment 46. The method of any one of Example Embodiments 27 to 39, wherein: the first node comprises aNWDAF comprising an AnLF, and the second node comprises a consumer.

[0559] Example Embodiment 47. The method of Example Embodiment 46, wherein the information comprises Nnwdaf AnalyticsSubscription Subscribe operation information.

[0560] Example Embodiment 48. The method of any one of Example Embodiments 27 to 39, wherein: the first node comprises a consumer, and the second node comprises a NWDAF comprising an AnLF.

[0561] Example Embodiment 49. The method of Example Embodiment 48, wherein the information comprises Nnwdaf AnalyticsSubscription Notify operation information.

[0562] Example Embodiment 50. The method of any one of Example Embodiments 27 to 49, wherein the first node or second node operating as the VFL client comprises a User Equipment or a gNodeB.

[0563] Example Embodiment 51. The method of any one of Example Embodiments 27 to 50, wherein the information is transmitted during a registration procedure for registering the first node with the second node.

[0564] Example Embodiment 52. The method of any one of Example Embodiments 27 to 51, wherein the information is transmitted in response to or with a Federated Learning preparation request.

[0565] Example Embodiment 53. The method of Example Embodiment 52, comprising: receiving, from the first node, a Federated Learning response indicating the one or more Federated Learning parameters that the first node supports.

[0566] Group C Example Embodiments P111793WO01 (017997.4195) PATENT APPLICATION

[0567] 64

[0568] Example Embodiment 54. A first node using Vertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, the first node configured to perform any of the steps of any of the Group A Example embodiments.

[0569] Example Embodiment 55. A second node using Vertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, the second node configured to perform any of the steps of any of the Group B Example embodiments.

[0570] Example Embodiment 56. A node for using Vertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, comprising: processing circuitry configured to perform any of the steps of any of the Group A and Group B Example embodiments; and power supply circuitry configured to supply power to the processing circuitry.

Claims

1. P111793WO01 (017997.4195) PATENT APPLICATION65CLAIMS1. A method (1000) by first node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500) using Vertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, the method comprising: receiving (1002), from a second node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500), information comprising and / or indicating at least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result.

2. The method of Claim 1, wherein the interoperability information indicates a type of training method that is supported by the second node for the Al and / or ML model and / or agreed by the first node and second node.

3. The method of any one of Claims 1 to 2, wherein the interoperability information comprises at least one of: an indicator of at least one gradient associated with the at least one intermediate result, and an indication of a number of gradients per batch of samples in an intermediate matrix container that includes the intermediate matrix.

4. The method of any one of Claims 1 to 3, wherein the interoperability information comprises at least one of: an indicator of at least one activations associated with the at least one intermediate result, and an indication of a number of activations per batch of samples in an intermediate matrix container that includes the intermediate matrix.

5. The method of any one of Claims 1 to 4, comprising using the information to train the Al and / or ML model.

6. The method of any one of Claims 1 to 5, wherein the Al and / or ML model is neural network-based.P111793WO01 (017997.4195) PATENT APPLICATION667. The method of any one of Claims 1 to 6. wherein the information and / or the interoperability information comprises at least one parameter associated with the intermediate matrix to be exposed.

8. The method of Claim 7, wherein the at least one parameter associated with the intermediate matrix comprises at least one dimension of the intermediate matrix, and wherein the at least one dimension of the intermediate matrix comprises at least one of: a number of a plurality of nodes in an intermediate matrix; a content of each of a plurality of nodes of an intermediate matrix; and a type of content of each of the plurality of the nodes in the intermediate matrix.

9. The method of Claim 8, wherein the at least one parameter associated with the intermediate matrix comprises a number of activations per batch of samples in a container.

10. The method of any one of Claims 1 to 9, wherein the information and / or the interoperability information comprises at least one of: at least one analytics identifier, at least one sample identifier, at least one notification target address, at least one notification correlation identifier, and at least one analytics reporting parameter.

11. The method of any one of Claims 1 to 10, wherein: the first node comprises a Network Resource Function, NRF (120, 320, 420), and the second node comprises a Network Function, NF (325, 425).

12. The method of Claim 11, wherein the information comprises Nnrf NFManagement NFRegister service operation information.

13. The method of any one of Claims 1 to 10, wherein: the first node is operating as a VFL client (110, 210, 310, 410), andP111793WO01 (017997.4195) PATENT APPLICATION67 the second node is operating as a VFL server (105, 205, 305, 405).

14. The method of Claim 13, wherein the information comprises NnrfyMLModelTraining Subscribe service operation information.

15. The method of any one of Claims 1 to 10, wherein: the first node is operating as a VFL server (105, 205, 305, 405), and the second node is operating as a VFL client (110, 210, 310, 410).

16. The method of Claim 15, wherein the information comprises NnwdafyMLModelTraining Notify service operation information.

17. The method of any one of Claims 1 to 10, wherein: the first node comprises a Network Data Analysis Function, NWDAF, comprising an Analytics Logic Function, AnLF, and the second node comprises a consumer (402).

18. The method of Claim 17, wherein the information comprises Nnwdaf AnalyticsSubscription Subscribe operation information.

19. The method of any one of Claims 1 to 10, wherein: the first node comprises a consumer (402), and the second node comprises a Network Data Analysis Function, NWDAF, comprising an Analytics Logic Function, AnLF.

20. The method of Claim 19, wherein the information comprises Nnwdaf AnalyticsSubscription Notify operation information.

21. The method of any one of Claims 1 to 10, wherein the first node or second node is operating as a VFL client (110, 210, 310, 410) and comprises a User Equipment, UE (1312, 1400), or a gNodeB, gNB (1310, 1500).P111793WO01 (017997.4195) PATENT APPLICATION6822. The method of any one of Claims 1 to 21, wherein the information is received during a registration procedure.

23. The method of any one of Claims 1 to 22, wherein the information is received in response to or with a Federated Learning, FL, preparation request.

24. The method of Claim 23, comprising: determining one or more FL parameters that the first node supports, and transmitting, to the second node, a FL response indicating the one or more FL parameters that the first node supports.

25. A method (1200) by second node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500) using Vertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, the method comprising: transmitting (1202), to a first node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500), information comprising and / or indicating at least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result.

26. The method of Claim 25, wherein the interoperability information indicates a type of training method that is supported by the second node for the Al and / or ML model and / or agreed by the first node and second node.

27. The method of any one of Claims 25 to 26, wherein the interoperability information comprises at least one of: an indicator of at least one gradient associated with the at least one intermediate result, and an indication of a number of gradients per batch of samples in an intermediate matrix container that includes the intermediate matrix.

28. The method of any one of Claims 25 to 27, wherein the interoperability information comprises at least one of:P111793WO01 (017997.4195) PATENT APPLICATION69 an indicator of at least one activations associated with the at least one intermediate result, and an indication of a number of activations per batch of samples in an intermediate matrix container that includes the intermediate matrix.

29. The method of Claim 28, comprising determining that the Al and / or ML model requires the type of training, and wherein the information is transmitted to the first node based on determining that the Al and / or ML model requires the type of training.

30. The method of any one of Claims 25 to 29, comprising using the information to train the Al and / or ML model.

31. The method of any one of Claims 25 to 30, wherein the Al and / or ML model is neural network-based.

32. The method of any one of Claims 25 to 31, wherein the information and / or the interoperability information comprises at least one parameter associated with an intermediate matrix to be exposed.

33. The method of Claim 32, wherein the at least one parameter associated with the intermediate matrix comprises at least one dimension of the intermediate matrix, and wherein the at least one dimension of the intermediate matrix comprises at least one of: a number of a plurality of nodes in an intermediate matrix; a content of each of a plurality of nodes of an intermediate matrix; and a type of content of each of the plurality of the nodes in the intermediate matrix.

34. The method of any one of Claims 32 to 33, wherein the at least one parameter associated with the intermediate matrix comprises a number of activations per batch of samples in a container.

35. The method of any one of Claims 25 to 34, wherein the information and / or the interoperability information comprises at least one of: at least one analytics identifier,P111793WO01 (017997.4195) PATENT APPLICATION70 at least one sample identifier, at least one notification target address, at least one notification correlation identifier, and at least one analytics reporting parameter.

36. The method of any one of Claims 25 to 35, wherein: the first node comprises a Network Resource Function, NRF (120, 320, 420), and the second node comprises a Network Function, NF (325, 425).

37. The method of Claim 36, wherein the information comprises NnrfyNFManagement NFRegister service operation information.

38. The method of any one of Claims 25 to 35, wherein: the first node is operating as a VFL client (110, 210, 310, 410), and the second node is operating as a VFL server (105, 205, 305, 405).

39. The method of Claim 38, wherein the information comprises NnrfyMLModelTraining Subscribe service operation information.

40. The method of any one of Claims 25 to 35, wherein: the first node is operating as a VFL server (105, 205, 305, 405), and the second node is operating as a VFL client (110, 210, 310, 410).

41. The method of Claim 40, wherein the information comprises Nnwdaf MLModelTraining Notify service operation information.

42. The method of any one of Claims 25 to 35, wherein: the first node comprises a Network Data Analysis Function, NWDAF, comprising an Analytics Logic Function, AnLF, and the second node comprises a consumer (402).P111793WO01 (017997.4195) PATENT APPLICATION7143. The method of Claim 42, wherein the information comprises Nnwdaf AnalyticsSubscription Subscribe operation information.

44. The method of any one of Claims 25 to 35, wherein: the first node comprises a consumer (402), and the second node comprises a Network Data Analysis Function, NWDAF, comprising an Analytics Logic Function, AnLF.

45. The method of Claim 44, wherein the information comprises Nnwdaf AnalyticsSubscription Notify operation information.

46. The method of any one of Claims 25 to 35, wherein the first node or second node is operating as a VFL client and comprises a User Equipment, UE (1312, 1400), or a gNodeB, gNB (1310, 1500).

47. The method of any one of Claims 25 to 46, wherein the information is transmitted during a registration procedure for registering the first node with the second node.

48. The method of any one of Claims 25 to 47, wherein the information is transmitted in response to or with a Federated Learning, FL, preparation request.

49. The method of Claim 48, comprising receiving, from the first node, a FL response indicating the one or more FL parameters that the first node supports.

50. A first node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500) using Vertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, the first node configured to: receive, from a second node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500), information comprising and / or indicating at least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result.P111793WO01 (017997.4195) PATENT APPLICATION7251. The first node of Claim 50, configured to perform any of the steps of any of Claims 2 to 24.

52. A second node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500) usingVertical Federated Learning, VFL, to train an Artificial Intelligence, Al, and / or Machine Learning, ML, model, the second node configured to: transmit, to a first node (105, 110, 205, 210, 305, 310, 402, 405, 410, 1310, 1312, 1400, 1500), information comprising and / or indicating at least one of: a type of the Al and / or ML model, interoperability information, and at least one intermediate result.

53. The second node of Claim 52, configured to perform any of the steps of any of the Claims 26 to 49.

Images