Supporting federated learning in a communication system

Abstract

Various aspects of the disclosure relate to a network entity for wireless communications that supports network data analytics, the network entity comprising instructions executable by at least one processor to cause the network entity to obtain a request message for training a model for supporting a network analytics operation identified by an analytics identifier, select a set of one or more model training functions to perform a federated learning training process of the model, select a weighting factor for each function of the selected set of one or more model training functions, and control the federated learning training process of the model and the one or more selected functions based on the selected weighting factor for each of the functions.

Description

Technical Field

[0001] This disclosure relates to wireless communications, and more specifically to network analysis in wireless communication networks. Background Technology

[0002] A wireless communication system may include one or more network communication devices, such as base stations, that support wireless communication for one or more user communication devices, which may also be referred to as user equipment (UE) or other suitable terms. The wireless communication system can support wireless communication with one or more user communication devices by utilizing the resources of the wireless communication system (e.g., time resources (e.g., symbols, time slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers, etc.)). Furthermore, the wireless communication system can support wireless communication across a variety of radio access technologies, including third-generation (3G), fourth-generation (4G), fifth-generation (5G), and other suitable radio access technologies beyond 5G (e.g., sixth-generation (6G)). Summary of the Invention

[0003] The article “a” preceding an element is unrestricted and should be understood to refer to “at least one” or “one or more” of those elements. The terms “a,” “at least one,” “one or more,” and “at least one of one or more” are interchangeable. As used herein, the word “or,” as used in a list of items (e.g., a list of items beginning with phrases such as “at least one,” “one or more,” or “one or two”) indicates an inclusive list, such that a list of at least one of, for example, A, B, or C means A or B or C or AB or AC or BC or ABC (i.e., A and B and C). Furthermore, as used herein, the phrase “based on” should not be construed as referring to a closed set of conditions. For example, without departing from the scope of this disclosure, an example step described as “based on condition A” may be based on both condition A and condition B. In other words, as used herein, the phrase “based on” should be interpreted in the same manner as the phrase “at least partially based on.” Furthermore, as used herein, the word “group” may comprise one or more elements.

[0004] Some embodiments of the methods and apparatus described herein may include a network entity for supporting wireless communication for network data analysis, the network entity including instructions executable by at least one processor to cause the network entity to: receive a request message for training a model to support (e.g., perform) network analysis operations identified by an analysis identifier; select a set of one or more model training functions to perform a federated learning training process for the model; select a weighting factor for each of the selected set of one or more model training functions; and control the federated learning training process of the model and the one or more selected functions based on the selected weighting factor for each of the functions.

[0005] The instruction causes the network entity to output a response message in response to the completion of the federated learning training process. The response message includes at least: an identifier for each model training function; the contribution weight of each model training function; and an identifier for the federated learning training process of the model. The identifier for each model training function may, for example, include the address of each model training function. The identifier for the model training process may be a process ID of the model training process, for example, a single process ID identifying the federated learning training process.

[0006] The response message may include one or more of the following: information identifying the network entity as a function controlling (and / or coordinating) the federated learning process; and information identifying the model training function as a participant in the federated learning process.

[0007] The instruction enables the network entity to determine the completion of the federated learning training process based on the stability of the model.

[0008] The response message may be a response to the request message.

[0009] The instruction causes the network entity to output a training completion confirmation message to at least one second network entity in the communication network in response to the completion of the federated learning training process. The training completion confirmation message includes at least: an identifier for each of the model training functions; the contribution weight for each of the model training functions; and an identifier for the federated learning training process of the model. Similar to what has been described above, the identifier for the model training process may be a process ID of the model training process, for example, a single process ID identifying the federated learning training process.

[0010] The at least one second network entity may include one or more of the following: a network repository function; and an analytics data repository function.

[0011] The instruction enables the network entity to receive the request message from a consumer entity within the communication network.

[0012] The consumer entity can provide consumer analysis logic functions.

[0013] The instruction enables the network entity to select a first contribution weight for a given function of the one or more model training functions based on at least one of the following: the amount of available data from the given function for the federated learning training process; and the quantified importance weight of the available data from the given function.

[0014] The instruction may cause the network entity to perform at least one of the following: selecting the contribution weight such that a given model training function with a relatively high amount of available data has a higher contribution weight than the contribution weight of a given model training function with a relatively low amount of data; and selecting the contribution weight such that a given model training function with a relatively high level of quantization importance has a higher contribution weight than the contribution weight of a given model training function with a relatively low level of quantization importance.

[0015] The instruction enables the network entity to select the contribution weight such that the given model training function has a corresponding contribution weight higher than that of the second given model training function, in response to the given model training function having at least one of the following: a relatively higher amount of data compared to the second given model training function; and a relatively higher level of quantization importance compared to the second given model training function.

[0016] The control of the federated learning training process may include: receiving intermediate values ​​of the model from each of the one or more model training functions; calculating at least one of one or more gradients and one or more losses based on the intermediate values; and outputting one or more gradient / loss messages to the one or more model training functions, including the one or more gradients and the at least one of the one or more losses.

[0017] The control of the federated learning training process may include updating the first contribution weight of each of the model training functions based on at least one of the following: the intermediate value; the one or more gradients; the one or more losses; and the number of times the intermediate value has been received from a given model training function within a preset time period.

[0018] The request message may include at least one analysis identifier that identifies the set of network analysis operations.

[0019] The one or more network analysis operations may include at least one statistical analysis operation to be performed on the network.

[0020] The one or more network analysis operations may include at least one prediction operation to be performed regarding the behavior of one or more entities within the communication network.

[0021] The prediction operation may include at least one of the following: prediction of parameters associated with at least one entity within the communication network, the parameters optionally being the load of the at least one entity, the at least one entity optionally including at least one of a network entity of the communication network and a UE; and prediction of mobility characteristics of the UE in the communication network, the mobility characteristics optionally including the location of the UE.

[0022] Each of the one or more model training functions may be a network function for providing analytical data associated with one or more UEs within the communication network.

[0023] The model can be a machine learning model.

[0024] The federated learning model can be a vertical federated learning model.

[0025] The network entity may include at least one of the following: model training function; and application function.

[0026] The network entity may be configured to store one or more data tags associated with the model.

[0027] In some embodiments of the methods and apparatus described herein, one method is performed by a network entity of a wireless communication network, the method comprising: receiving a request message for training a model to support network analysis operations identified by an analysis identifier; selecting one or more model training functions to perform a federated learning training process for the model; selecting a weighting factor for each of the selected one or more model training functions; and controlling the federated learning training process of the model and the one or more selected functions based on the selected weighting factor for each of the functions. Attached Figure Description

[0028] Figure 1 Examples of wireless communication systems according to aspects of this disclosure are described.

[0029] Figure 2 and 3 Describe the network architecture.

[0030] Figure 4A and 4B Explain federated learning technology.

[0031] Figure 5A and 5B Describe the communication flow of federated learning.

[0032] Figure 6 Describe the architecture of instance-based federated learning.

[0033] Figure 7 Describe the process flow of federated learning based on instances.

[0034] Figure 8 Examples of network equipment (NE) 800 according to aspects of this disclosure are described.

[0035] Figure 9 A flowchart illustrating the method performed by NE according to aspects of this disclosure. Detailed Implementation

[0036] Some wireless communication systems (comprising one or more network entities) may support network analysis operations (e.g., processes, actions) for performing statistical analysis of activities within these wireless communication systems. For example, analysis may be performed on parameters such as signal quality and traffic volume. In some other instances, network analysis operations may involve one or more network entities performing predictions (e.g., estimations, determinations, forecasts) of characteristics of the wireless communication system (e.g., a network). For example, one or more network entities may perform predictions of the location (e.g., positioning) of user equipment (UEs) at a given time, or predictions of the load (e.g., data traffic) of one or more network entities at a future given time.

[0037] This disclosure provides improvements in the performance (e.g., accuracy and / or processing) of such network analysis. Various aspects of this disclosure relate to performing network analysis using one or more learning techniques (e.g., longitudinal federated learning techniques). Federated learning techniques, described in more detail below, are techniques in which multiple entities collaborate to train a model, and their data is not centralized. It should be understood that other learning techniques (e.g., machine learning models, artificial intelligence models, etc.) can be used to support network analysis and enable one or more network entities to support network analysis. Some aspects of this disclosure relate to the control of federated learning techniques (e.g., processes) by one or more network entities to efficiently provide improved network analysis. This control, described in more detail below, may include sending messages to participants in the process to coordinate their training of the model. This use of learning techniques can provide improved statistical analysis and improved prediction quality, thereby improving the accuracy of network analysis. Improved network performance can thus be achieved.

[0038] Aspects of this disclosure are described in the context of wireless communication systems.

[0039] Figure 1This describes an example of a wireless communication system 100 according to aspects of this disclosure. The wireless communication system 100 may include one or more NEs 102, one or more UEs 104, and a core network (CN) 106. The wireless communication system 100 may support various radio access technologies. In some embodiments, the wireless communication system 100 may be a 4G network, such as an LTE network or an LTE-A network. In some other embodiments, the wireless communication system 100 may be an NR network, such as a 5G network, a 5G-A network, or a 5G Ultra Wideband (5G-UWB) network. In other embodiments, the wireless communication system 100 may be a combination of 4G and 5G networks, or other suitable radio access technologies, including IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20. The wireless communication system 100 may support radio access technologies beyond 5G, such as 6G. In addition, the wireless communication system 100 can support technologies such as Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), or Code Division Multiple Access (CDMA).

[0040] One or more NEs 102 may be distributed across a geographical area to form a wireless communication system 100. One or more of the NEs 102 described herein may be, include, or be referred to as a network node, base station, network element, network function, network entity, radio access network (RAN), NodeB, eNodeB (eNB), next-generation NodeB (gNB), or other suitable terms. NEs 102 and UEs 104 may communicate via a communication link, which may be a wireless or wired connection. For example, NEs 102 and UEs 104 may perform wireless communication (e.g., receive signaling, transmit signaling) via a Uu interface.

[0041] NE 102 can provide a geographic coverage area that supports services for one or more UEs 104 within that geographic coverage area. For example, NE 102 and UE 104 can support wireless communication of signals associated with services (e.g., voice, video, packet data, messaging, broadcasting, etc.) based on one or more radio access technologies. In some embodiments, NE 102 can be mobile, for example, a satellite associated with a non-terrestrial network (NTN). In some embodiments, different geographic coverage areas 112 associated with the same or different radio access technologies can overlap, but different geographic coverage areas can be associated with different NEs 102.

[0042] One or more UEs 104 may be distributed across a geographical area of ​​the wireless communication system 100. UE 104 may include or be referred to as a remote unit, mobile device, wireless device, remote device, subscriber device, transmitter device, receiver device, or some other suitable term. In some implementations, UE 104 may be referred to as a unit, station, terminal, or client, and other instances thereof. Additionally or alternatively, UE 104 may be referred to as an Internet of Things (IoT) device, an Internet of Everything (IoE) device, or a Machine Type Communication (MTC) device, and other instances thereof.

[0043] UE 104 may be able to support direct wireless communication with other UE 104 via a communication link. For example, UE 104 may support direct wireless communication with another UE 104 via a device-to-device (D2D) communication link. In some implementations, such as vehicle-to-vehicle (V2V) deployments, vehicle-to-everything (V2X) deployments, or cellular-V2X deployments, communication link 114 may be referred to as a side link. For example, UE 104 may support direct wireless communication with another UE 104 via a PC5 interface.

[0044] NE 102 may support communication with CN 106 or with another NE 102 or both. For example, NE 102 may interface with other NE 102 or CN 106 via one or more backhaul links (e.g., S1, N2, N2, or network interfaces). In some embodiments, NE 102 may communicate directly with each other. In some other embodiments, NE 102 may communicate with each other or indirectly (e.g., via CN 106). In some embodiments, one or more NE 102 may include sub-components, such as access network entities, which may be instances of access node controllers (ANCs). The ANC may communicate with one or more UE 104s via one or more other access network transmitting entities (which may be referred to as radio heads, smart radio heads, or transmit-receive points (TRPs)).

[0045] CN 106 can support user authentication, access authorization, tracking, connectivity, and other access, routing, or mobility functions. CN 106 can be an evolved packet core (EPC) or a 5G core (5GC), which may include control plane entities (e.g., Mobility Management Entity (MME), Access and Mobility Management Function (AMF)) that manage access and mobility, and user plane entities (e.g., Serving Gateway (S-GW), Packet Data Network (PDN) Gateway (P-GW), or User Plane Function (UPF)) that route packets or interconnect to external networks. In some implementations, the control plane entities may manage non-access stratum (NAS) functions of one or more UEs 104 served by one or more NEs 102 associated with CN 106, such as mobility, authentication, and bearer management (e.g., data bearers, signaling bearers, etc.).

[0046] CN 106 can communicate with the packet data network via one or more backhaul links (e.g., via S1, N2, N2, or another network interface). The packet data network may contain an application server. In some implementations, one or more UEs 104 can communicate with the application server. UE 104 can establish a session (e.g., a Protocol Data Unit (PDU) session, etc.) with CN 106 via NE 102. CN 106 can use the established session (e.g., an established PDU session) to route services (e.g., control information, data, etc.) between UE 104 and the application server. A PDU session may be an instance of a logical connection between UE 104 and CN 106 (e.g., one or more network functions of CN 106).

[0047] In the wireless communication system 100, NE 102 and UE 104 can use the resources of the wireless communication system 100 (e.g., time resources (e.g., symbols, time slots, subframes, frames, etc.) or frequency resources (e.g., subcarriers, carriers)) to perform various operations (e.g., wireless communication). In some embodiments, NE 102 and UE 104 may support different resource structures. For example, NE 102 and UE 104 may support different frame structures. In some embodiments, such as in 4G, NE 102 and UE 104 may support a single frame structure. In some other embodiments, such as in 5G and other suitable radio access technologies, NE 102 and UE 104 may support various frame structures (i.e., multiple frame structures). NE 102 and UE 104 may support various frame structures based on one or more parameter sets.

[0048] The wireless communication system 100 may support one or more parameter sets, and the parameter sets may include subcarrier spacing and cyclic prefixes. A first parameter set (e.g., μ=0) may be associated with a first subcarrier spacing (e.g., 15 kHz) and a regular cyclic prefix. In some embodiments, the first parameter set (e.g., μ=0) associated with the first subcarrier spacing (e.g., 15 kHz) may utilize one time slot per subframe. A second parameter set (e.g., μ=1) may be associated with a second subcarrier spacing (e.g., 30 kHz) and a regular cyclic prefix. A third parameter set (e.g., μ=2) may be associated with a third subcarrier spacing (e.g., 60 kHz) and a regular cyclic prefix or an extended cyclic prefix. A fourth parameter set (e.g., μ=3) may be associated with a fourth subcarrier spacing (e.g., 120 kHz) and a regular cyclic prefix. A fifth parameter set (e.g., μ=4) may be associated with a fifth subcarrier spacing (e.g., 240 kHz) and a regular cyclic prefix.

[0049] Time intervals for resources (e.g., communication resources) can be organized according to frames (also known as radio frames). Each frame may have a duration, for example, 10 milliseconds (ms). In some embodiments, each frame may contain multiple subframes. For example, each frame may contain 10 subframes, and each subframe may have a duration, for example, 1 ms. In some embodiments, each frame may have the same duration. In some embodiments, each subframe of a frame may have the same duration.

[0050] Alternatively, the time intervals of resources (e.g., communication resources) can be organized according to time slots. For example, a subframe may contain a certain number (e.g., quantity) of time slots. The number of time slots in each subframe may also depend on one or more parameter sets supported in the wireless communication system 100. For example, the first, second, third, fourth, and fifth parameter sets (i.e., μ=0, μ=1, μ=2, μ=3, μ=4) associated with corresponding subcarrier intervals of 15 kHz, 30 kHz, 60 kHz, 120 kHz, and 240 kHz may utilize a single time slot per subframe, two time slots per subframe, four time slots per subframe, eight time slots per subframe, and 16 time slots per subframe, respectively. Each time slot may contain a certain number (e.g., quantity) of symbols (e.g., OFDM symbols). In some embodiments, the number (e.g., quantity) of time slots in a subframe may depend on the parameter set. For a conventional cyclic prefix, a time slot may contain 14 symbols. For an extended cyclic prefix (e.g., applicable to a 60 kHz subcarrier spacing), a time slot may contain 12 symbols. The relationship between the number of symbols per time slot for the regular cyclic prefix and the extended cyclic prefix, the number of time slots per subframe, and the number of time slots per frame may depend on the parameter set. It should be understood that references to the first parameter set (e.g., μ=0) associated with the first subcarrier spacing (e.g., 15 kHz) may be used interchangeably between subframes and time slots.

[0051] In the wireless communication system 100, the electromagnetic (EM) spectrum can be divided into various categories, frequency bands, channels, etc., based on frequency or wavelength. For example, the wireless communication system 100 may support one or more operating frequency bands, such as frequency range names FR1 (410 MHz to 7.125 GHz), FR2 (24.25 GHz to 52.6 GHz), FR3 (7.125 GHz to 24.25 GHz), FR4 (52.6 GHz to 114.25 GHz), FR4a or FR4-1 (52.6 GHz to 71 GHz), and FR5 (114.25 GHz to 300 GHz). In some embodiments, NE 102 and UE 104 may perform wireless communication on one or more of the operating frequency bands. In some embodiments, FR1 may be used by NE 102 and UE 104, as well as other equipment or devices, for cellular communication services (e.g., control information, data). In some implementations, FR2 can be used by NE 102 and UE 104, as well as other equipment or devices, for short-range, high data rate capabilities.

[0052] FR1 may be associated with one or more parameter sets (e.g., at least three parameter sets). For example, FR1 may be associated with a first parameter set containing a 15 kHz subcarrier spacing (e.g., μ=0); a second parameter set containing a 30 kHz subcarrier spacing (e.g., μ=1); and a third parameter set containing a 60 kHz subcarrier spacing (e.g., μ=2). FR2 may be associated with one or more parameter sets (e.g., at least two parameter sets). For example, FR2 may be associated with a third parameter set containing a 60 kHz subcarrier spacing (e.g., μ=2); and a fourth parameter set containing a 120 kHz subcarrier spacing (e.g., μ=3).

[0053] In the wireless communication system 100, NE 102 can be configured (as described in more detail below) to control the federated learning process to perform network analysis.

[0054] Figure 2 This describes the network architecture that allows data collection related to the use of the Network Data Analysis Function (NWDAF). In some cases, the Analysis Logic Function (AnLF) trains a machine learning (ML) model from a Model Training Logic Function (MTLF) request by including the analysis ID corresponding to the analysis ID requested by the analysis consumer in the request. Information about the available analysis IDs supported by AnLF is provided in 3GPP TS 23.288. MTLF training corresponds to the ML model requested by collecting data from one or more data sources (NF, OAM, or UE).

[0055] The functionality of NWDAF (e.g.) Figure 1The functionality of NWDAF can be divided between AnLF and MTLF. Figure 3 This illustrates an example of this functional division.

[0056] exist Figure 3 In this framework, each MTLF is trained to perform a corresponding ML model for network analysis. Each MTLF can be associated with an analysis ID. Each analysis ID identifies a specific network analysis. A given ML model can be configured by the ML model designer.

[0057] Each MTLF can train its model based on data collected from a data producer network function (NF). A data producer NF can be any network function that generates data. For example, a data producer NF could be an Access and Mobility Function (AMF) that generates location information about the UE. Alternatively or additionally, MTLFs can train their models based on historical data received from a Data Collection Coordination Function (DCCF) or an Analytics Data Repository Function (ADRF).

[0058] AnLF then receives the trained ML model (and / or data indicating the trained ML model) for the specific analysis ID from the MTLF corresponding to the analysis ID. AnLF uses the model, possibly combined with new data from the data producer NF and / or DCCF, to derive the analysis requested by the consumer NF.

[0059] AnLF can subscribe to MTLF to receive trained ML models with analytics IDs using a specific API (existing technology). AnLF subscriptions are based on analytics requests from consumer NFs. NWDAF supports both analytics inference and ML model training.

[0060] In some cases, direct data exchange between the data producer (NF) and the NWDAF (i.e., MTLF or AnLF) may not be possible. For example, this could be due to privacy concerns: operators of edge networks or private slices may not want to share data with external networks. As another example, this could be due to high signaling load: signaling load can increase significantly when data is provided to centralized training functions, such as those offered by MTLF. Receiving trained models from the network at the edge, rather than collecting data to derive analytics, may be more efficient.

[0061] One way to address situations where data won't be directly exchanged (e.g., due to privacy concerns or signaling load efficiency) is through federated learning. Federated learning allows local model training functions to exchange model parameters and aggregate trained models and / or intermediate training results, rather than centrally training the model by collecting raw data. Therefore, federated learning is a distributed machine learning framework that allows models to be trained collectively from data distributed across different data owners. This offers the advantages that AI / ML models can be trained closer to the data source, rather than sending raw data to a centralized training node. Instead of sending raw data, the parameters, weights, or intermediate results of the ML model can be sent back to the centralized node to aid in the training of a general model.

[0062] It should be understood that federated learning can be conceptually divided into horizontal and vertical federated learning. These are respectively in... Figure 4A and 4B The image is depicted schematically.

[0063] Lateral federated learning, or sample-based federated learning, can be applied to scenarios where datasets share the same feature space (e.g., each dataset contains the same features) but have different samples (e.g., collected from different users). This is applicable in... Figure 4A The explanation is as follows.

[0064] Vertical federated learning, or feature-based federated learning, can be applied to scenarios with datasets that share the same sample space (e.g., each applied to a different set of users) but differ in their feature spaces (e.g., collecting different features in each dataset). This is in... Figure 4B The explanation is as follows.

[0065] In both horizontal and vertical federated learning models, parameters from each local model training function are sent to a model aggregator to compute an aggregated model. The model aggregator provides each MTLF with updated model parameters, and each MTLF uses these parameters to retrain its own model, thus allowing each local model training function to have a trained model that uses data from multiple sources.

[0066] There are different approaches to training models using Vertical Federated Learning (VFL), including both "split" and "non-split" methods. Examples of these approaches are now described in *Vertical Federated Learning: Taxonomies, Threats, and Prospects* by Qun Li et al. Specifically, its... Figure 2 Describe the classification of the VFL algorithm, including those that split and those that do not. Figure 3 The non-split VFL is depicted in more detail, while Figure 4 depicts the split VFL.

[0067] In both split and non-split VFLs, one party is the "label owner" or "active participant." This party has the ability to classify the input data. The other parties are "passive participants" or "workers" participating in the VFL training process.

[0068] In non-split VFLs, passive participants use their own models to send intermediate results to active participants based on locally collected data, while active participants use labels and their own ML models as a basis to compute gradients / loss. Some scenarios also involve a coordinator (such as Li) who ensures that message exchanges between VFL parties are encrypted. Figure 3 (As depicted in the text).

[0069] In split VFLs, the model is split among several parties. One party owns the top model (label owner / VLF server), while the other parties own one or more bottom models (passive participants). This is illustrated in Lie Figure 4. The label owner can also be an active participant. Compared to non-split VFLs, the VFL server knows the labels and can compute the gradients / losses shared with the passive participants.

[0070] Figure 5A and 5B Describes the communication flows of federated learning in different NWDAFs. These communication flows are described in 3GPP TS 23.288.

[0071] Figure 5A This describes the registration, discovery, and selection of clients used for federated learning. Each MTLF registers its FL capabilities (FL client or FL server) in its NWDAF profile within the NRF. The profile additionally includes information about the analytics IDs supported for the FL and / or location region and / or event ID that supports data collection therein. An NWDAF MTLF with FL server capabilities discovers candidate FL clients for training models using federated learning by receiving a list from the NRF of FL clients that support the analytics IDs and / or location regions and / or event IDs required to train the model. The FL server sends a readiness request to all candidates to check their readiness to join the FL process, i.e., in terms of data availability and required timeline (steps 7 through 9), and each FL client responds to whether it is ready to join the process.

[0072] 5B describes the FL procedure once the FL customer has been selected and joined the FL process.

[0073] The general procedure of FL is that the FL server sends a subscription request related to federated learning for model training to each selected FL client, including local model parameters, and each FL client uses data from the local NF to train the model.

[0074] This disclosure provides examples of solutions to support longitudinal federated learning, enabling 5GS to assist collaborative AI / ML operations involving 5GC / NWDAF or AF. Specifically, to support VFL, this disclosure provides methods for VFL entities (servers, active participants, or passive participants) to participate in the VFL training process. The following description also illustrates actions that servers and active participants can take after the VFL training process is completed (e.g., model information storage).

[0075] Figure 6 Various architectures for executing VFL according to examples of this disclosure are described. These architectures describe the communication between VFL server 610, active participant 615, and passive participant 625. In this example, VFL server 710 (which may also be referred to as VFL coordinator or controller) manages the VFL program, selects VFL participants 615 and 625 for model training, and assigns functions to act as either active participant 615 or passive participant 625.

[0076] In this example, VFL active participant 615 is a VFL function that possesses a portion of the ML model for a given analysis ID and is aware of the ML model's labels. Active participant 615 is the primary function used to train the ML model for the given analysis ID.

[0077] In this example, VFL passive participant 625 is a VFL function that has a portion of an ML model with a given analysis ID, is unaware of the labels in the ML model, but is able to collect local data on one or more features.

[0078] A given VFL function in a 3GPP network can support a combination of the functions mentioned above, for example, supporting both VFL server and active participant functions.

[0079] In this example, VFL server 610 or active participant 615 identifies how each VFL passive participant 625 contributes during model training and assigns weights subsequently used for inference. VFL server 610 or active participant 615 stores ML model information in a function so that if the ML model needs to be retrained, the VFL service can subsequently retrieve the details.

[0080] In this example, a given VFL function (i.e., VFL server 610, active participant 615, and / or passive participant 625) may be part of a function in NWDAF AnLF, NWDAF MTLF, or AF. When interactions occur between a VFL function from NWDAF and a VFL function from AF, message exchange may be performed via Network Exposure Function (NEF) 628.

[0081] Figure 6 This paper describes four architectural scenarios, including VFL server 610, active participant 615, and passive participant 625, involving NWDAF AnLF, MTLF, and AF. Specifically, Figure 6 The top line depicts instances where VFL is executed within the core network, and the bottom line depicts instances where VFL is started outside the core network. Within those lines, Figure 6 The left side depicts an instance containing VFL server 610, and the right side depicts an instance in which VFL server 610 is not used.

[0082] Figure 7 A process flow according to aspects of this disclosure is described. The process flow can be implemented as referenced. Figure 1 The described aspects of the wireless communication system 100, and more specifically, in... Figure 6 The process flow can be implemented within one or more of the architectures described herein. The process flow may involve one or more of the following: Consumer AnLF 705, VFL Server 710, VFL Active Participant 715, Data Producer 720, VFL Passive Participant 725, Data Producer 730, VFL Passive Participant 735, Data Producer 740, NRF 745, and ADRF 750. Therefore, it can be implemented within one or more of these components. Figure 6 The architecture described herein implements a process flow in which entities with the same names have the same functionality as described above. In the following description, operations performed by one or more of entities 705 to 750 may be performed in a different order than shown, or at different times. Some operations may also be omitted from the process flow, and other operations may be added.

[0083] exist Figure 7 In this instance, VFL server 710 controls (i.e., coordinates) the VFL operations performed together with active VFL participants 715 and passive participants 725 and 735. This is based on data from data producers 720, 730, and 740. In other instances, smaller or larger numbers of participants and / or data producers can participate in the VFL process. Therefore, Figure 7 The process flow provides an efficient way to perform VFL procedures in wireless communication networks. As explained above, this improves the accuracy and performance of network analysis, thereby improving system performance.

[0084] At point 1, consumer 705 (in this instance, an NWDAF supporting AnLF) requests a derived trained ML model to support an analysis (e.g., DN performance analysis). This request includes the analysis ID and / or model ID of the required trained model. A list of analyses is outlined in 3GPP TS 23.288, one or more of which can be supported by an NWDAF. The request is launched to VFL server 710. In an alternative instance, the request is launched to VFL active participant 715 or a server supporting a combination of the functionalities of VFL server 710 and VFL active participant 715.

[0085] At two points, VFL server 710 (and / or VFL active participant 715) determines that model training using federated learning will be performed. This may include one or more of the following: identifying which VFL features acting as passive participants (i.e., those not owning labels) can participate in the VFL process; identifying what data (features or samples) and the data size available in each participant; performing feature / sample alignment (so that all VFL participants use the same samples to perform VFL); and / or selecting VFL passive participants 725, 735 that can participate in the VFL process for model training.

[0086] At three points, based on the selected passive VFL participants 725 and 735, VFL server 710 and / or active participant 715 identify the initial contribution weight from each passive participant 725 and 735. The contribution weight may be based on the importance of the feature and / or the amount of data available to each active participant. A VFL process ID is also assigned to the training process, causing all VFL participants 715, 725, and 735 to associate the exchange of VFL-related messages (including the VFL process ID) with a single VFL process initiated by VFL server 710 and / or active participant 715.

[0087] At four points, each participant in the VFL process, 715, 725, and 735, trains their model using the available data (features) and derives intermediate values ​​(4a, 4b, and 4c).

[0088] At point 5, intermediate values ​​from each passive participant 725, 735 are sent to active participant 715 and / or directly to VFL server 710. If VFL server 710 is used, active participant 715 can provide its derived intermediate values ​​to VFL server 710.

[0089] At point 6, VFL server 710 or active participant 715 calculates the gradient / loss based on the received intermediate values. The gradient / loss can be an aggregate value based on all received intermediate values ​​or a gradient / loss value for each VFL passive participant 725, 735.

[0090] At point 7, the gradient / loss is sent to each passive participant 725, 735 or active participant 715 (if using VFL server 710).

[0091] At point 8, each VFL active / passive participant 715, 725, 735 uses gradients / losses and other input data (8a, 8b, 8c) to update their ML models.

[0092] At point 9, one or more iterations from 4 to 8 occur until the VFL server 710 or the VFL active participant 715 determines that the ML model has been trained with the confidence level requested by the ML model consumer.

[0093] At point 10, after the VFL iteration is complete, the VFL server / active participants 710, 715 update the contribution weights from each passive participant 725, 730, for example, based on the number of times feedback (i.e., intermediate values) has been provided by each passive participant 725, 735 or the number of times feedback has been provided within the time limit, and the gradient / loss estimate for each passive participant.

[0094] At point 11, once the ML model is trained, the VFL server / active participant 710, 715 indicates to the ML model consumer 705 that the trained ML model is available. This response may include one or more of the following: the address and / or NF ID of each VFL participant 715, 725, 735 and their respective contribution weights; the address of the VFL server 710 (if used); the address of the VFL active participant 715; and / or the process ID of the VFL process.

[0095] At point 12, VFL servers / active participants 710 and 715 can also update NRF 745 with an indication that the model for a given analytics ID was trained using VFL. This indication may include the address / NF ID of one or more of each of the VFL participants 715, 725, 735, VFL server 710, and / or VFL active participant 715, and / or the VFL process ID.

[0096] At point 13, VFL servers / active participants 710 and 715 can also indicate to ADRF 750 that the model for a given analytics ID was trained using VFL. This indication may include the address / NF ID of one or more of each of the VFL participants 715, 725, 735, VFL server 710, and / or VFL active participant 715. The address storing the global model may also be included in the request. The VFL process ID may also be stored.

[0097] Figure 6 architecture combination Figure 7The process flow thus provides an efficient way to perform VFL within a wireless communication network.

[0098] Figure 8 This describes an example of an NE 800 according to aspects of this disclosure. The NE 800 may include a processor 802, a memory 804, a controller 806, and a transceiver 808. The processor 802, memory 804, controller 806, or transceiver 808, or various combinations thereof, or various components thereof, may be examples of components for performing the various aspects of this disclosure as described herein. These components may be coupled via one or more interfaces (e.g., operatively ground, communicatively ground, functional ground, electronic ground, electrical ground).

[0099] Processor 802, memory 804, controller 806, or transceiver 808, or various combinations or components thereof, may be implemented in hardware (e.g., a circuit system). The hardware may be a processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), or other programmable logic device, or any combination thereof configured or otherwise supporting components for performing the functions described in this disclosure.

[0100] Processor 802 may include intelligent hardware devices (e.g., a general-purpose processor, DSP, CPU, ASIC, FPGA, or any combination thereof). In some embodiments, processor 802 may be configured to operate memory 804. In some other embodiments, memory 804 may be integrated into processor 802. Processor 802 may be configured to execute computer-readable instructions stored in memory 804 to cause NE 800 to perform various functions of this disclosure.

[0101] Memory 804 may comprise volatile or non-volatile memory. Memory 804 may store computer-readable, computer-executable code containing instructions that, when executed by processor 802, cause NE 800 to perform the various functions described herein. The code may be stored in a non-transitory computer-readable medium, such as memory 804 or another type of memory. Computer-readable medium includes both non-transitory computer storage media and communication media, wherein the communication media includes any media that facilitates the transfer of a computer program from one place to another. Non-transitory storage media may be any available media accessible by a general-purpose or special-purpose computer.

[0102] In some implementations, processor 802 and memory 804 coupled to processor 802 may be configured to cause NE 800 to perform one or more of the functions described herein (e.g., instructions stored in memory 804 are executed by processor 802). For example, processor 802 may support wireless communication at NE 800 according to examples disclosed herein. NE 800 may be configured to support components for: receiving a request message for training a model to support network analysis operations identified by an analysis identifier; selecting a set of one or more model training functions to perform a federated learning training process for the model; selecting a weighting factor for each of the selected set of one or more model training functions; and controlling the federated learning training process of the model and one or more of the selected functions based on the selected weighting factor for each of the functions.

[0103] Controller 806 manages the input and output signals of NE 800. Controller 806 can also manage peripheral devices not integrated into NE 800. In some embodiments, controller 806 may utilize an operating system such as iOS®, ANDROID®, WINDOWS®, or other operating systems. In some embodiments, controller 806 may be implemented as part of processor 802.

[0104] In some embodiments, NE 800 may include at least one transceiver 808. In other embodiments, NE 800 may have more than one transceiver 808. Transceiver 808 may represent a wireless transceiver. Transceiver 808 may include one or more receiver chains 810, one or more transmitter chains 812, or a combination thereof.

[0105] Receiver chain 810 may be configured to receive signals (e.g., control information, data, packets) over a wireless medium. For example, receiver chain 810 may include one or more antennas for receiving signals over the air or over a wireless medium. Receiver chain 810 may include at least one amplifier (e.g., a low-noise amplifier (LNA)) configured to amplify the received signal. Receiver chain 810 may include at least one demodulator configured to demodulate the received signal and obtain the transmitted data by reversing the modulation technique applied during signal transmission. Receiver chain 810 may include at least one decoder for decoding the demodulated signal to receive the transmitted data.

[0106] Transmitter chain 812 can be configured to generate and transmit signals (e.g., control information, data, packets). Transmitter chain 812 may include at least one modulator for modulating data onto a carrier signal, preparing the signal for transmission over a wireless medium. The at least one modulator may be configured to support one or more techniques, such as amplitude modulation (AM), frequency modulation (FM), or digital modulation schemes like phase shift keying (PSK) or quadrature amplitude modulation (QAM). Transmitter chain 812 may also include at least one power amplifier configured to amplify the modulated signal to an appropriate power level suitable for transmission over a wireless medium. Transmitter chain 812 may also include one or more antennas for transmitting the amplified signal into the air or wireless medium.

[0107] Figure 9 A flowchart illustrating a method according to an aspect of this disclosure is provided. The operation of the method can be implemented by an NE as described herein. In some embodiments, the NE can execute a set of instructions to control the functional elements of the NE to perform the described function. The operation of the method can be performed according to examples as described herein. In some embodiments, it can be implemented by... (refer to...) Figure 8 The NE described herein is the aspect that performs the operation. One or more of the operations may be omitted, or performed in a different order than in the example shown.

[0108] At 902, the method may include obtaining a request message for a model trained to support network analysis operations identified by an analysis identifier.

[0109] At 904, the method may include selecting a set of one or more model training functions to perform a federated learning training process for the model.

[0110] At 906, the method may include selecting a weighting factor for each feature in a selected set of one or more model training features.

[0111] At 908, the method may include, based on the selected weighting factor for each of the functions, controlling the federated learning training process of the model and one or more selected functions.

[0112] It should be noted that the methods described herein describe possible implementations, and the operations and steps may be rearranged or otherwise modified, and other implementations are possible.

[0113] The description herein is provided to enable those skilled in the art to make or use this disclosure. Various modifications to this disclosure will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the scope of this disclosure. Therefore, this disclosure is not limited to the examples and designs described herein, but should be given the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A network entity for supporting wireless communication for network data analysis, the network entity comprising instructions executable by at least one processor to cause the network entity to perform the following operations: Obtain a request message for the model trained to support network analysis operations identified by the analysis identifier; Select one or more model training functions to perform the federated learning training process for the model; Select a weighting factor for each of the selected set of one or more model training functions; and The federated learning training process of the model and the one or more selected functions are controlled based on the selected weighting factor for each of the functions.

2. The network entity of claim 1, wherein the instruction causes the network entity to output a response message in response to the completion of the federated learning training process, the response message comprising at least: Identifier for each of the model training functions described; The contribution weight of each of the model training functions; and An identifier that identifies the federated learning training process of the model.

3. The network entity according to claim 2, wherein the response message includes at least one of the following: Information identifying the network entities as control functions of the federated learning process; and The model training function is identified as information of the participants in the federated learning process.

4. The network entity according to claim 2 or claim 3, wherein the instructions cause the network entity to determine the completion of the federated learning training process based on the stability of the model.

5. The network entity according to any one of claims 2 to 4, wherein the response message is a response to the request message.

6. The network entity according to any of the preceding claims, wherein the instruction causes the network entity to output a training completion confirmation message to at least one second network entity of the communication network in response to the completion of the federated learning training process, the training completion confirmation message comprising at least: Identifiers for each of the model training functions; and The contribution weights for each of the model training functions. An identifier that identifies the federated learning training process of the model.

7. The network entity of claim 6, wherein the at least one second network entity comprises one or more of the following: Network repository functionality; and Analyze data repository functionality.

8. The network entity according to any of the preceding claims, wherein the instructions cause the network entity to receive the request message from a consumer entity within the communication network.

9. The network entity according to claim 8, wherein the consumer entity is a consumer analysis logic function.

10. A network entity according to any of the preceding claims, wherein the instructions cause the network entity to select a first contribution weight for a given function of the one or more model training functions based on at least one of the following: The amount of available data from the given function for the federated learning training process; and Quantified importance weights from available data of the given function.

11. The network entity of claim 10, wherein the instructions cause the network entity to perform at least one of the following: The contribution weights are selected such that the training performance of a given model with a relatively high amount of available data has a higher contribution weight than the training performance of a given model with a relatively low amount of data; and The contribution weights are selected such that a given model training function with a relatively high level of quantization importance has a higher contribution weight than the contribution weights of a given model training function with a relatively low level of quantization importance.

12. The network entity of claim 10 or claim 11, wherein the instruction causes the network entity to select the contribution weight such that the given model training function has a corresponding contribution weight higher than the contribution weight of the second given model training function, in response to the given model training function having at least one of the following: Compared to the training capabilities of the second given model, the amount of data is relatively high; and Compared to the training function of the second given model, it has a relatively high level of quantization importance.

13. The network entity according to any of the preceding claims, wherein controlling the federated learning training process includes: Receive intermediate values ​​of the model from each of the one or more model training functions; Based on the intermediate value, calculate at least one of one or more gradients and one or more losses; and The training function outputs one or more gradient / loss messages, including one or more gradients and at least one of the one or more losses, to the one or more models.

14. The network entity of claim 13, wherein controlling the federated learning training process includes updating the first contribution weight of each of the model training functions based on at least one of the following: The intermediate value; The one or more gradients; The one or more losses; and The number of times the intermediate value has been received from the given model training function within a preset time period.

15. The network entity according to any of the preceding claims, wherein the request message includes at least one analysis identifier identifying the set of network analysis operations.

16. The network entity according to any of the preceding claims, wherein the federated learning model is a longitudinal federated learning model.

17. The network entity according to any of the preceding claims, wherein the network entity comprises at least one of the following: Model training function; and Application functions.

18. The network entity according to any of the preceding claims, wherein the network entity is configured to store one or more data tags associated with the model.

19. A method performed by a network entity of a wireless communication network, the method comprising: Obtain a request message for the model trained to support network analysis operations identified by the analysis identifier; Select one or more model training functions to perform the federated learning training process for the model; Select a weighting factor for each of the selected set of one or more model training functions; and The federated learning training process of the model and the one or more selected functions are controlled based on the selected weighting factor for each of the functions.

Images