Communication method, communication node, medium, and program product

Abstract

The invention discloses a communication method, a communication node, a medium and a program product. Comprising the steps of receiving a first model subscription request sent by a second communication node; wherein the first model subscription request comprises model demand time; performing training duration estimation according to the first model subscription request, and determining an estimated training duration; and comparing the estimated training duration with the model demand time, and feeding back a first model training instruction to the second communication node according to a comparison result. Before the longitudinal federated learning training is carried out, the longitudinal federated learning server estimates the time required by the training, and timely communicates and matches the estimated training duration obtained by estimation with the network element demand of a consumer. Therefore, the first communication node can carry out longitudinal federated learning training only under the condition that the user demand can be satisfied, and invalid consumption of training resources is reduced.

Description

Communication methods, communication nodes, media and software products Technical Field

[0001] This application relates to the field of communication technology, and in particular to a communication method, communication node, medium, and program product. Background Technology

[0002] In the 5th Generation Mobile Communication (5G) system, the Network Data Analytics Function (NWDAF) is a 5G Core Network Function (5GC NF) located in the control plane, which can perform statistical data and machine learning related tasks in the 5G system.

[0003] Current 5G systems support multiple NWDAFs and / or Application Functions (AFs) as participants for federated learning training and / or inference, and also support multiple triggering processes for federated learning training. However, the training and / or inference time for vertical federated learning is relatively long, which may lead to training waste if consumer network elements are unwilling to wait for a long time. Summary of the Invention

[0004] This application provides a communication method, communication node, medium, and program product to solve the problem of untimely negotiation of waiting time between consumer network elements and federated learning server during federated learning. It enables timely communication of consumer network element demand time and federated learning training time during the process of triggering federated learning training, and conducts vertical federated learning training only when it can be determined that the user actually needs it, thereby reducing the ineffective consumption of training resources.

[0005] To achieve the above objectives, embodiments of this application provide a communication method applied to a first communication node, comprising:

[0006] Receive a first model subscription request sent by a second communication node; wherein the first model subscription request includes the model requirement time;

[0007] Based on the first model subscription request, the training duration is estimated to determine the predicted training duration.

[0008] The estimated training time is compared with the model's required time, and the first model training instruction is fed back to the second communication node based on the comparison result.

[0009] To achieve the above objectives, embodiments of this application provide a communication method applied to a second communication node, comprising:

[0010] Receive the analysis requirement time and send a first model subscription request to the first communication node according to the analysis requirement time; wherein, the first model subscription request includes the model requirement time;

[0011] Receive the first model training instruction from the first communication node.

[0012] To achieve the above objectives, embodiments of this application provide a communication method applied to a first communication node, comprising:

[0013] The time required to receive the analysis request sent by the second communication node;

[0014] Estimate the analysis duration and determine the predicted analysis time.

[0015] The estimated analysis time is compared with the required analysis time, and vertical federated learning is initiated based on the comparison results.

[0016] To achieve the above objectives, embodiments of this application provide a communication method applied to a second communication node, comprising:

[0017] Receive the analysis request time and send the analysis request time to the first communication node.

[0018] To achieve the above objectives, embodiments of this application provide a communication method applied to a second communication node, comprising:

[0019] Send a training duration request to the third communication node;

[0020] Receive the estimated training time from the feedback of the third communication node.

[0021] To achieve the above objectives, embodiments of this application provide a communication method applied to a third communication node, comprising:

[0022] Receive the vertical federated learning training instruction sent by the second communication node; the vertical federated learning training instruction includes the training time requirement.

[0023] Based on the longitudinal federated learning training instructions, estimate the training duration and determine the estimated training duration.

[0024] The estimated training duration is compared with the required training time, and vertical federated learning training is initiated based on the comparison results.

[0025] The communication method provided in this application embodiment receives a first model subscription request sent by a second communication node; wherein the first model subscription request includes model required time; the training duration is estimated based on the first model subscription request to determine the estimated training duration; the estimated training duration is compared with the model required time, and a first model training instruction is fed back to the second communication node based on the comparison result. By adopting the above technical solution, before conducting vertical federated learning training, the vertical federated learning server first estimates the training time required, and the estimated training duration is promptly communicated and matched with the consumer network element requirements, so that the first communication node can only conduct vertical federated learning training when it is determined that the user's needs can be met, reducing the ineffective consumption of training resources. Attached Figure Description

[0026] Figure 1 is a schematic diagram of a 5G core network architecture provided in the prior art;

[0027] Figure 2 is a timing diagram of a vertical federated learning triggering process provided in the prior art;

[0028] Figure 3 is a timing diagram of another vertical federated learning triggering process provided in the prior art;

[0029] Figure 4 is a flowchart of a communication method provided in an embodiment of this application;

[0030] Figure 5 is a flowchart of a communication method provided in an embodiment of this application;

[0031] Figure 6 is a flowchart of a communication method provided in an embodiment of this application;

[0032] Figure 7 is a flowchart of a communication method provided in an embodiment of this application;

[0033] Figure 8 is a flowchart of a communication method provided in an embodiment of this application;

[0034] Figure 9 is a flowchart of a communication method provided in an embodiment of this application;

[0035] Figure 10 is a flowchart of a communication method provided in an embodiment of this application;

[0036] Figure 11 is a flowchart of a communication method provided in an embodiment of this application;

[0037] Figure 12 is a timing example diagram of a vertical federated learning triggering process provided in an embodiment of this application;

[0038] Figure 13 is a timing example diagram of a vertical federated learning triggering process provided in an embodiment of this application;

[0039] Figure 14 is a timing example diagram of a vertical federated learning triggering process provided in an embodiment of this application;

[0040] Figure 15 is a timing example diagram of a vertical federated learning triggering process provided in an embodiment of this application;

[0041] Figure 16 is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0042] Figure 17 is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0043] Figure 18 is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0044] Figure 19 is a schematic diagram of the structure of a communication device provided in an embodiment of this application;

[0045] Figure 20 is a schematic diagram of a communication device provided in an embodiment of this application;

[0046] Figure 21 is a schematic diagram of a communication device provided in an embodiment of this application;

[0047] Figure 22 is a schematic diagram of a communication device provided in an embodiment of this application;

[0048] Figure 23 is a schematic diagram of a communication device provided in an embodiment of this application;

[0049] Figure 24 is a schematic diagram of the structure of a communication node provided in an embodiment of this application. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be arbitrarily combined with each other.

[0051] The steps illustrated in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases the steps shown or described may be performed in a different order than that presented here.

[0052] The communication method provided in this application can be used in 5G systems to trigger federated learning training and / or inference for multiple NWDAFs and / or AFs that support federated learning training and / or inference. To clearly describe federated learning in 5G systems, a brief introduction is given here of the 5G core network architecture and the NWDAFs that perform federated learning as 5GC NFs based on the 5G core network architecture.

[0053] Figure 1 is a schematic diagram of a 5G core network architecture provided in the prior art, which has the following functions:

[0054] 1) User Equipment (UE).

[0055] 2) Radio Access Network (RAN). The RAN manages radio resources, transmits user data received through the N3 interface to the UE, and transmits user data from the UE through the N3 interface. The RAN maps between Quality of Service (QoS) traffic in Dedicated Radio Bearer (DRB) and Protocol Data Unit (PDU) sessions.

[0056] 3) Access and Mobility Management Function (AMF). This function includes registration management, connection management, reachability management, and mobility management. It also performs access authentication and authorization. AMF is a Network Attached Storage (NAS) security endpoint used to forward SM NAS data between the UE and the Session Management Function (SMF).

[0057] 4) SMF. This function includes the following: session establishment, modification, and release; UE IP address allocation and management (including optional authorization functions); selection and control of User Plane (UP) functions and downlink data notification, etc. The SMF controls the User Plane Function (UPF) through the N4 interface. The SMF provides the UPF with Packet Detection Rules (PDR) to indicate how to detect user data traffic; provides Forwarding and Routing Control Association Rules (FAR); and provides QoS enforcement rules and Usage Reporting Rules (URR) to indicate how the UPF performs user data traffic forwarding, QoS processing, and usage reporting on user data traffic detected using PDR.

[0058] 5) UPF. This function includes the following: serving as an anchor point for intra / inter-radio access type mobility, packet routing and forwarding, traffic usage reporting, QoS processing for the UP, downlink packet buffering, and downlink data notification triggering. The General Packet Radio Service (GPRS) Tunneling Protocol for User Plane (GTP-U) tunnel is used for the N3 interface between the RAN and UPF. The GTP-U tunnel operates on a per-PDU session basis. For downlink traffic, the UPF binds the downlink traffic to the QoS traffic within the PDU session's GTP-U tunnel using the FAR received from the SMF. For uplink traffic, the RAN transmits user plane traffic to the QoS stream identified by the UE.

[0059] 6) Policy Control Function (PCF). The PCF provides QoS policy rules to control plane functions for enforcement. The PCF translates AF requests into PCC rules applicable to PDU sessions.

[0060] 7) Unified Data Management (UDM). The UDM performs 3GPP AKA authentication credential generation, access authorization based on subscription data, UE service NF registration management (e.g., storing AMF for UE storage services, SMF for UE PDU session storage services), and subscription management. The UDM accesses the UDR to retrieve UE subscription data and stores the UE context in the UDR. The UDM and UDR can be deployed together.

[0061] Based on the 5G core network architecture, the NWDAF is a 5GC NF located in the control plane, performing data statistics and machine learning-related tasks in the 5G system. The NWDAF can interact with different entities for various purposes:

[0062] Data is collected based on event subscriptions provided by AMF, SMF, UPF, PCF, UDM, NSCAF, AF (directly or through NEF) and Operations, Administration and Maintenance (OAM);

[0063] [Optional] Use the Data Collection Coordination Function (DCCF) for analysis and data collection;

[0064] Retrieve information from data repositories (e.g., retrieve UDRs related to users via UDM or retrieve PFD information via NEF (PFDF));

[0065] Collect location information data from the LCS system;

[0066] [Optional] Store and retrieve information from the Analytics Data Storage Function (ADRF);

[0067] [Optional] Analyze and collect data from the Messaging Framework Adaptor Function (MFAF);

[0068] Retrieve information about NF (e.g., retrieve NF-related information from NRF);

[0069] Provide analytics to consumers on demand.

[0070] Provides batch data related to the analysis ID.

[0071] Provides information on the accuracy of the analysis ID.

[0072] Provides information on the accuracy of machine learning (ML) models or indications of ML model accuracy degradation.

[0073] In some examples, a single instance or multiple instances of NWDAF can be deployed in a public terrestrial mobile network. NWDAF may contain the following logical functions:

[0074] Analysis Logic Function (AnLF): A logic function in NWDAF used to perform inference, derive analytical information (i.e., derive statistical data and / or predictions based on analytical consumer requests) and expose analytical services (i.e., Nnwdaf_AnalyticsSubscription or Nnwdaf_AnalyticsInfo).

[0075] Model Training Logic Function (MTLF): A logic function in NWDAF used to train ML models and expose new training services (such as providing pre-trained ML models).

[0076] An NWDAF can contain a Model Training Logical Function (MTLF) or an Analytics Logical Function (AnLF), or both.

[0077] The Data Collection Coordination and Function (DCCF) is also an NF on the 5G core network control plane. The DCCF is responsible for coordinating the collection and distribution of data requested by NF consumers. It prevents data sources from processing multiple subscriptions to the same data and prevents multiple notifications containing the same information from being sent due to incoordination of data consumer requests.

[0078] DCCF is applicable to:

[0079] NWDAF requests data from a data source (such as for computational analysis).

[0080] NF consumers are analyzed from the NWDAF data source.

[0081] An NF consumer that requests data from an ADRF data source.

[0082] ADRF that receives data from NF data sources.

[0083] To clearly describe the differences between the vertical federated learning triggering process in this application embodiment and the existing vertical federated learning triggering process, a brief introduction to the existing vertical federated learning triggering process is given first. Figure 2 is a timing diagram of a vertical federated learning triggering process provided in the prior art. Specifically, it may include the following steps:

[0084] 0) If the NWDAF containing AnLF receives an analysis request from a consumer network element.

[0085] 1) If an NWDAF containing AnLF wants to perform inference but has no model, it will discover an NWDAF containing MTLF from the NRF and send a model subscription request to the NWDAF containing MTLF using Nnwdaf_MLModelProvision_Subscribe.

[0086] 2) NWDAF containing MTLF (as a VFL server) returns the error code: No model because the model was trained by VFL.

[0087] 3) The above model subscription process is terminated.

[0088] 4) The NWDAF containing AnLF sends an analytics subscription request to MTLF.

[0089] 5) MTLF, acting as the VFL server, initiates VFL training.

[0090] 6) The MTLF, acting as the VFL server, initiates VFL inference.

[0091] 7) The VFL server returns the inference results (also known as the analysis results) to the NWDAF containing AnLF.

[0092] Figure 3 is a timing diagram of another vertical federated learning triggering process provided in the prior art, which may include the following steps:

[0093] 1) AnLF sends a model training request to AF, which acts as the VFL server.

[0094] 2) AF estimates the duration of VFL training, returns whether VFL training can be completed within the required time limit, and initiates VFL training.

[0095] 3) AF sends a VFL training completion indication to AnLF.

[0096] However, although current 5G systems support various triggering procedures for vertical federated learning training, including the aforementioned vertical federated learning triggering process, the long duration of vertical federated learning and the lack of negotiation on the waiting time between the vertical federated learning server and the AnLF before triggering training may lead to wasted training if the consumer network element is unwilling to wait for an extended period. To address this issue, this application provides a communication method applicable between a first communication node, a second communication node, and / or a third communication node. The first communication node can be an NWDAF containing an MTLF that acts as a VFL server among the vertical federated learning participants, as shown in Figure 2. The second communication node can be an NWDAF containing an AnLF that acts as an interaction node with the consumer network element among the vertical federated learning participants, as shown in Figure 2 or Figure 3. The third communication node can be an AF that acts as a VFL server among the vertical federated learning participants, as shown in Figure 3. The first, second, and third communication nodes are generally electronic devices with certain computing capabilities. In some possible implementations, the communication method can be implemented by a processor calling computer-executable instructions stored in memory.

[0097] In one exemplary embodiment, Figure 4 is a flowchart of a communication method provided in an embodiment of this application. This method can be applied to the longitudinal federated learning process in a 5G system, where longitudinal federated learning training and / or inference triggering is performed on communication nodes participating in the longitudinal federated learning. This method can be executed by a communication device, which can be implemented by software and / or hardware and integrated on the communication node. This method can be applied to a first communication node. In this embodiment of the application, the first communication node can be an NWDAF containing an MTLF that acts as a VFL server among the longitudinal federated learning participants as shown in Figure 2. This embodiment of the application does not limit this.

[0098] As shown in Figure 4, the communication method provided in this embodiment of the application specifically includes the following steps:

[0099] S101, Receive the first model subscription request sent by the second communication node.

[0100] The first model subscription request includes the model requirement time.

[0101] In this embodiment, the second communication node may be an NWDAF containing AnLF that serves as an interaction node with the consumer network element among the vertical federated learning participants, as shown in Figure 2 or Figure 3.

[0102] In this embodiment, the first model subscription request can be specifically understood as a request from a second communication node, which receives a request from a consumer network element to perform inference but finds that it does not have a model, to subscribe to the model required by the consumer network element from a first communication node discovered by the Network Repository Function (NRF).

[0103] In this embodiment, the model demand time can be specifically understood as the latest time that the consumer expects to receive the model.

[0104] In a specific example, after receiving a request from a consumer network element, the second communication node will send a first model subscription request to the first communication node, based on the consumer network element's requirements for the model. This request will include the latest time the consumer network element expects to receive the model. The first communication node can then receive the first model subscription request and extract the model requirement time from it.

[0105] S102. Estimate the training duration based on the first model subscription request and determine the estimated training duration.

[0106] In this embodiment, the estimated training duration can be understood as the time length obtained by the first communication node from the trajectory of the longitudinal federated learning training to be triggered, or the estimated time to complete the longitudinal federated learning training.

[0107] In a specific example, the first communication node responds to the first model subscription request to determine the type of model that needs to be trained by vertical federated learning, and estimates the time required to complete the training to obtain the estimated training duration.

[0108] S103. Compare the estimated training time with the model's required time, and based on the comparison result, send the first model training instruction back to the second communication node.

[0109] In this embodiment, the first model training instruction can be specifically understood as an instruction information used to indicate that the second communication node cannot directly request the model required by the consumer network element because the first communication node supports the VFL model.

[0110] In a specific example, the first communication node compares the estimated training time obtained with the model requirement time extracted from the first model subscription request to determine whether the longitudinal federated learning training initiated by the first communication node can be completed before the latest time that the consumer expects to receive the model. The determined result is then fed back to the second communication node in the first model training instruction, which indicates that the second communication node cannot directly request the model required by the consumer network element.

[0111] The communication method provided in this application embodiment receives a first model subscription request sent by a second communication node; wherein the first model subscription request includes model required time; the training duration is estimated based on the first model subscription request to determine the estimated training duration; the estimated training duration is compared with the model required time, and a first model training instruction is fed back to the second communication node based on the comparison result. By adopting the above technical solution, before conducting vertical federated learning training, the vertical federated learning server first estimates the training time required, and the estimated training duration is promptly communicated and matched with the consumer network element requirements, so that the first communication node can only conduct vertical federated learning training when it is determined that the user's needs can be met, reducing the ineffective consumption of training resources.

[0112] In one embodiment, estimating the training duration includes at least one of the following:

[0113] Training duration for vertical federated learning;

[0114] The time required to complete the longitudinal federated learning training.

[0115] In one embodiment, the first model subscription request also includes an analytics identifier.

[0116] In this embodiment, the analysis identifier can be specifically understood as an identifier used to distinguish the purpose of a model in 5G system federated learning.

[0117] In some examples, the analytics identifier included in the first model subscription request may be an identifier used to indicate the purpose for which the consumer network element wishes to request the model.

[0118] In one embodiment, estimating the training duration based on the first model subscription request includes:

[0119] Training time is estimated based on the analysis labels.

[0120] In a specific example, upon receiving a first model subscription request, the first communication node can determine the vertical federated learning model functionality required by the consumer network element based on the analysis identifier contained in the first model subscription request, and thus determine the model type that needs to be trained using VFL. Knowing the model type that needs VFL training, the first communication node can estimate the VFL training duration for that model type and obtain the corresponding estimated training duration.

[0121] In one embodiment, feeding back a first model training instruction to a second communication node based on the comparison result includes:

[0122] If the comparison result indicates that the estimated training time is later than the model's required time, send at least one of the following to the second communication node:

[0123] Error indication;

[0124] Estimated training time;

[0125] Error indications and estimated training time.

[0126] In this embodiment, the error indication can be specifically understood as an indication used to inform the second communication node that the time taken by the first communication node to perform VFL training is insufficient to meet consumer demand.

[0127] In some examples, the first model training instruction may also include an error code, such as: No model, because the model was trained by VFL.

[0128] In one embodiment, the error indication includes a reason code;

[0129] The reason code includes:

[0130] The model cannot be trained within the required time or made available within the required time.

[0131] In this embodiment, the reason code can be specifically understood as information used to inform the second communication node why its subscription failed.

[0132] In one embodiment, the first communication node is the model training logic function; the second communication node is the analysis logic function.

[0133] In one exemplary embodiment, Figure 5 is a flowchart of a communication method provided in an embodiment of this application. This method can be applied to the longitudinal federated learning process of a 5G system, where longitudinal federated learning training and / or inference triggering is performed on communication nodes participating in the longitudinal federated learning. This method can be executed by a communication device, which can be implemented by software and / or hardware and integrated on the communication node. This method can be applied to a second communication node. In this embodiment of the application, the second communication node can be an NWDAF containing AnLF that serves as an interaction node with the consumer network element among the longitudinal federated learning participants, as shown in Figure 2 or Figure 3. This embodiment of the application does not limit this.

[0134] As shown in Figure 5, the communication method provided in this embodiment of the application specifically includes the following steps:

[0135] S201. Receive the analysis request time and send the first model subscription request to the first communication node according to the analysis request time.

[0136] The first model subscription request includes the model requirement time.

[0137] In this embodiment, the analysis requirement time can be specifically understood as the latest time that the consumer network element sends to the second communication node to indicate the latest time that the consumer expects to receive the analysis data provided by the first communication node, which can also be understood as the latest time that the first communication node completes inference.

[0138] In a specific example, the second communication node receives the analysis request time sent by the consumer network element, and determines the latest time that the consumer can accept for the model to complete training, i.e. the model request time, based on the analysis request time. Then, it sends the model request time along with the first model subscription request to the first communication node that can provide VFL model training and inference.

[0139] S202, Receive the first model training instruction fed back by the first communication node.

[0140] In a specific example, after estimating and comparing the model training time based on the first model subscription request, the first communication node will send a first model training instruction back to the second communication node. That is, the second communication node will wait after sending the first model subscription request until it receives the first model training instruction from the first communication node.

[0141] In one embodiment, the first model subscription request also includes an analytics identifier.

[0142] In one embodiment, receiving a first model training instruction fed back by a first communication node includes:

[0143] Receive at least one of the following feedbacks from the first communication node:

[0144] Error indication;

[0145] Estimated training time;

[0146] Error indications and estimated training time.

[0147] In one embodiment, the error indication includes a reason code;

[0148] Reason codes include:

[0149] The model cannot be trained within the required time or made available within the required time.

[0150] In one embodiment, after receiving the first model training instruction fed back by the first communication node, the method further includes:

[0151] If the estimated training time exceeds a preset time threshold, a new first communication node is discovered.

[0152] In a specific example, after the second communication node receives the first model training instruction from the first communication node, it can determine whether to continue using the first communication node for VFL training and / or inference based on the estimated training time in the first model training instruction. If the estimated training time exceeds a preset time threshold, it can be considered that the second communication node feels that the training time is too long and the first communication node is no longer suitable for providing analysis and inference services to consumer network elements. At this time, the second communication node can find a new first communication node through NRF and no longer trigger the VFL training or inference process for the current first communication node.

[0153] In one embodiment, estimating the training time includes at least one of the following:

[0154] Training duration for vertical federated learning;

[0155] The time required to complete the longitudinal federated learning training.

[0156] In one embodiment, the first communication node is a model training logic function; the second communication node is an analysis logic function.

[0157] In one exemplary embodiment, Figure 6 is a flowchart of a communication method provided in an embodiment of this application. This method can be applied to the longitudinal federated learning process in a 5G system, where longitudinal federated learning training and / or inference triggering is performed on communication nodes participating in the longitudinal federated learning. This method can be executed by a communication device, which can be implemented by software and / or hardware and integrated on the communication node. This method can be applied to a first communication node. In this embodiment of the application, the first communication node can be an NWDAF containing an MTLF that acts as a VFL server among the longitudinal federated learning participants as shown in Figure 2. This embodiment of the application does not limit this.

[0158] As shown in Figure 6, the communication method provided in this embodiment of the application specifically includes the following steps:

[0159] S301, Receive the analysis request time sent by the second communication node.

[0160] In a specific example, the second communication node can receive the analysis request time, which indicates the latest time that the consumer expects to receive the analysis data provided by the first communication node, from the consumer network element, and can directly forward the analysis request time to the first communication node.

[0161] S302. Estimate the analysis duration and determine the estimated analysis duration.

[0162] In this embodiment, the estimated analysis duration can be understood as the estimated time length obtained by the first communication node for the duration that the longitudinal federated learning training and / or inference to be triggered will consume, or the estimated time to complete the longitudinal federated learning training and / or inference.

[0163] In a specific example, after receiving the analysis request time, the first communication node determines the model type that needs to be trained and / or inferred by VFL corresponding to the analysis request time, and estimates the time required to complete the VFL training and / or inference to obtain the estimated analysis duration.

[0164] S303. Compare the estimated analysis duration with the required analysis time, and initiate vertical federated learning based on the comparison results.

[0165] In a specific example, after estimating the analysis duration and determining the estimated analysis duration, the first communication node can compare it with the received analysis request time to determine whether the VFL training and / or inference initiated by the first communication node can be completed before the latest time that the consumer expects to receive the analysis data. Then, based on the comparison result, it can determine whether the first communication node needs to perform federated learning training and / or inference, and if it is determined that the first communication node needs to perform VFL training and / or inference, the first communication node initiates VFL.

[0166] In one embodiment, the analysis requirement time also includes an analysis identifier.

[0167] In a specific example, when the consumer network element sends the analysis request time to the second communication node, it will carry the functionality of the model it requires, that is, it will carry the analysis identifier in the analysis request time.

[0168] In one embodiment, the analysis duration estimation includes:

[0169] Estimate the analysis duration based on the analysis identifier.

[0170] In a specific example, upon receiving an analysis request time, the first communication node can determine the VFL model functionality required by the consumer network element based on the analysis identifier included in the analysis request time. This allows it to identify the model type requiring VFL training and / or inference. Knowing the model type requiring VFL training and / or inference, the first communication node can estimate the VFL analysis duration for that model type and obtain the corresponding estimated analysis duration.

[0171] In one embodiment, the estimated analysis duration includes at least one of the following:

[0172] The duration of training and reasoning in vertical federated learning;

[0173] The time required to complete vertical federated learning training and reasoning;

[0174] Vertical federated learning inference duration;

[0175] Time required to complete longitudinal federated learning inference;

[0176] Vertical federated learning training duration;

[0177] Time required to complete longitudinal federated learning training;

[0178] The duration of the analysis;

[0179] The time required to generate the analysis.

[0180] In one embodiment, initiating longitudinal federated learning based on the comparison results includes:

[0181] If the comparison results show that the estimated analysis time is no later than the required analysis time, initiate longitudinal federated learning training and / or inference.

[0182] In a specific example, if the comparison result shows that the estimated analysis time is no later than the analysis requirement time, that is, if the latest time for the first communication node to complete VFL training and / or inference is no later than the analysis requirement time, it can be considered that the time required for the first communication node to perform VFL training and / or inference can meet the consumer's expected needs. In other words, it can be considered that the consumer can use the first communication node to perform VFL model training and / or inference. At this time, the first communication node can initiate VFL training and / or inference.

[0183] In one embodiment, initiating longitudinal federated learning based on the comparison results includes:

[0184] When the comparison results show that the estimated analysis time is later than the required analysis time, sending the required analysis time to the second communication node cannot meet the indication and the estimated analysis time.

[0185] Time to receive new analysis requests from the second communication node;

[0186] Compare the estimated analysis time with the new analysis requirement time, and if the comparison result shows that the estimated analysis time is no later than the new analysis requirement time, initiate vertical federated learning training and / or inference.

[0187] In this embodiment, the "analysis demand time cannot be met" indication can be specifically understood as an indication used to inform the second communication node that the time required for the VFL model corresponding to the first communication node to perform inference analysis is insufficient to meet the consumer's analysis time requirements.

[0188] In a specific example, if the comparison result shows that the estimated analysis time is later than the required analysis time (i.e., the latest time for the first communication node to complete VFL training and / or inference is later than the required analysis time), it can be considered that the time required for VFL training and / or inference by the first communication node cannot meet the consumer's expected needs. In this case, the first communication node will send an indication to the second communication node that the required analysis time cannot be met to inform the second communication node of this situation, and will also send the determined estimated analysis time to the second communication node. If the second communication node decides to continue using the model trained by the first communication node, the first communication node will receive a new required analysis time from the second communication node. The first communication node will then compare the estimated analysis time and the new required analysis time in the same way as described above. If the comparison result shows that the estimated analysis time is not later than the new required analysis time, it is determined that the model trained and / or inferred by the first communication node can meet the consumer's expected needs. At this point, the first communication node will initiate VFL training and / or inference.

[0189] In one embodiment, after initiating longitudinal federated learning training and / or inference, the method further includes:

[0190] The reasoning or analysis results are sent to the second communication node.

[0191] In one embodiment, initiating longitudinal federated learning based on the comparison results includes:

[0192] When the comparison results show that the estimated analysis time is later than the required analysis time, sending the required analysis time to the second communication node cannot meet the indication and the estimated analysis time.

[0193] If no new analysis request is received from the second communication node, longitudinal federated learning training and / or inference will not be initiated.

[0194] In a specific example, if the comparison result shows that the estimated analysis time is later than the required analysis time—that is, the latest time for the first communication node to complete VFL training and / or inference is later than the required analysis time—it can be assumed that the time required for VFL training and / or inference by the first communication node cannot meet the consumer's expected demand. In this case, the first communication node will send an indication to the second communication node that the required analysis time cannot be met to inform the second communication node of the situation, and will also send the determined estimated analysis time to the second communication node. However, if the second communication node decides that the model trained by the first communication node is no longer needed, the first communication node will not receive the new required analysis time sent by the second communication node, and the first communication node will not initiate subsequent VFL training and / or inference.

[0195] In one embodiment, the first communication node is the model training logic function; the second communication node is the analysis logic function.

[0196] In one exemplary embodiment, Figure 7 is a flowchart of a communication method provided in an embodiment of this application. This method can be applied to the longitudinal federated learning process of a 5G system, where longitudinal federated learning training and / or inference triggering is performed on communication nodes participating in the longitudinal federated learning. This method can be executed by a communication device, which can be implemented by software and / or hardware and integrated on the communication node. This method can be applied to a second communication node. In this embodiment of the application, the second communication node can be an NWDAF containing AnLF that serves as an interaction node with the consumer network element among the longitudinal federated learning participants, as shown in Figure 2 or Figure 3. This embodiment of the application does not limit this.

[0197] As shown in Figure 7, the communication method provided in this embodiment of the application specifically includes the following steps:

[0198] S401, Receive the analysis request time and send the analysis request time to the first communication node.

[0199] In one embodiment, after sending the analysis requirement time to the first communication node, the method further includes:

[0200] The analysis request time received from the first communication node cannot meet the indication and estimated analysis duration.

[0201] In one embodiment, the estimated analysis duration includes at least one of the following:

[0202] The duration of training and reasoning in vertical federated learning;

[0203] The time required to complete vertical federated learning training and reasoning;

[0204] Vertical federated learning inference duration;

[0205] Time required to complete longitudinal federated learning inference;

[0206] Vertical federated learning training duration;

[0207] Time required to complete longitudinal federated learning training;

[0208] The duration of the analysis;

[0209] The time required to generate the analysis.

[0210] In one embodiment, after receiving the analysis request time sent by the first communication node, which cannot meet the indication and estimated analysis duration, the method further includes:

[0211] Send the new analysis request time to the first communication node.

[0212] In a specific example, after the second communication node receives the indication and estimated analysis duration sent by the first communication node that the analysis request time cannot be met, it can determine whether the VFL model trained by the first communication node still needs to continue based on the estimated analysis duration. If it is determined that it still needs to continue, the second communication node can send a new analysis request time to the first communication node.

[0213] In one embodiment, after sending the analysis request time to the first communication node, or sending a new analysis request time to the first communication node, the method further includes:

[0214] Receive the reasoning or analysis results sent by the first communication node.

[0215] In one embodiment, the analysis requirement time also includes an analysis identifier.

[0216] In one embodiment, the first communication node is the model training logic function; the second communication node is the analysis logic function.

[0217] In one exemplary embodiment, Figure 8 is a flowchart of a communication method provided in an embodiment of this application. This method can be applied to the longitudinal federated learning process of a 5G system, where longitudinal federated learning training and / or inference triggering is performed on communication nodes participating in the longitudinal federated learning. This method can be executed by a communication device, which can be implemented by software and / or hardware and integrated on the communication node. This method can be applied to a second communication node. In this embodiment of the application, the second communication node can be an NWDAF containing AnLF that serves as an interaction node with the consumer network element among the longitudinal federated learning participants, as shown in Figure 2 or Figure 3. This embodiment of the application does not limit this.

[0218] As shown in Figure 8, the communication method provided in this embodiment of the application specifically includes the following steps:

[0219] S501, Send a training duration request to the third communication node.

[0220] In this embodiment, the third communication node can be the AF that acts as the VFL server among the participants in the vertical federated learning, as shown in Figure 3.

[0221] In this embodiment, the training duration request can be specifically understood as a request instructing the third communication node to estimate the time required for VFL training that it may need to perform.

[0222] S502, Receive the estimated training time fed back by the third communication node.

[0223] In a specific example, since the second communication node knows that it will request the VFL model from the third communication node, it can query the third communication node for training duration information before initiating the VFL training request. In other words, it can instruct the third communication node to estimate the time required for VFL training by sending a training duration request, and receive the estimated training duration from the third communication node.

[0224] In one embodiment, the training duration request includes an analysis identifier.

[0225] In some examples, a training duration request can be used to request the duration of VFL training or to estimate the time required to complete VFL training.

[0226] In one embodiment, estimating the training duration includes at least one of the following:

[0227] Training duration for vertical federated learning;

[0228] The time required to complete the longitudinal federated learning training.

[0229] In one embodiment, after receiving the estimated training duration from the third communication node, the method further includes:

[0230] If the estimated training duration meets the preset training duration requirement, a vertical federated learning training request is sent to the third communication node.

[0231] In this embodiment, the preset training duration requirement can be understood as the duration determined according to actual needs, which is the time that the second communication node needs to wait for the third communication node to complete the VFL model training when initiating a training request.

[0232] In a specific example, if the estimated training time meets the preset training time requirement, it can be assumed that the second communication node believes that the training time of the third communication node for the VFL model can meet the usage requirements. At this time, the second communication node can send a VFL training request to the third communication node to instruct the third communication node to initiate VFL training.

[0233] In one embodiment, after receiving the estimated training duration from the third communication node, the method further includes:

[0234] If the estimated training duration does not meet the preset training duration requirement, a vertical federated learning training request is initiated to other second communication nodes.

[0235] In a specific example, if the estimated training time does not meet the preset training time requirement, it can be assumed that the second communication node believes that the training time of the third communication node for the VFL model cannot meet its usage requirements. At this time, the second communication node can be rediscovered by NRF to a third communication node other than the current third communication node, and the rediscovered third communication node initiates a VFL training request.

[0236] In one embodiment, the second communication node is the model training logic function; the third communication node is the application function.

[0237] In one exemplary embodiment, Figure 9 is a flowchart of a communication method provided in an embodiment of this application. This method can be applied to the longitudinal federated learning process in a 5G system, where longitudinal federated learning training and / or inference triggering is performed on communication nodes participating in the longitudinal federated learning. This method can be executed by a communication device, which can be implemented by software and / or hardware and integrated on the communication node. This method can be applied to a third communication node. In this embodiment of the application, the third communication node can be the AF acting as the VFL server among the participants in the longitudinal federated learning, as shown in Figure 3. This embodiment of the application does not limit this.

[0238] As shown in Figure 9, the communication method provided in this embodiment of the application specifically includes the following steps:

[0239] S601, Receive the training duration request sent by the second communication node.

[0240] S602. Estimate the training duration based on the training duration request, determine the estimated training duration, and feed the estimated training duration back to the second communication node.

[0241] In one embodiment, the training duration request includes an analysis identifier.

[0242] In one embodiment, estimating the training duration based on the training duration request includes:

[0243] Training time is estimated based on the analysis labels.

[0244] In one embodiment, estimating the training duration includes at least one of the following:

[0245] Training duration for vertical federated learning;

[0246] The time required to complete the longitudinal federated learning training.

[0247] In one embodiment, after feeding back the estimated training duration to the second communication node, the method further includes:

[0248] Receive the vertical federated learning training request sent by the second communication node;

[0249] Initiate vertical federated learning training in response to a request for vertical federated learning training.

[0250] In one exemplary embodiment, Figure 10 is a flowchart of a communication method provided in an embodiment of this application. This method can be applied to the longitudinal federated learning process in a 5G system, where longitudinal federated learning training and / or inference triggering is performed on communication nodes participating in the longitudinal federated learning. This method can be executed by a communication device, which can be implemented by software and / or hardware and integrated on the communication node. This method can be applied to a third communication node. In this embodiment of the application, the third communication node can be the AF acting as the VFL server among the participants in the longitudinal federated learning, as shown in Figure 3. This embodiment of the application does not limit this.

[0251] As shown in Figure 10, the communication method provided in this embodiment of the application specifically includes the following steps:

[0252] S701, Receive the longitudinal federated learning training instruction sent by the second communication node.

[0253] The training instructions for vertical federated learning include the time required for training.

[0254] In this embodiment, the vertical federated learning training instruction can be specifically understood as information instructing the third communication node to initiate VFL training.

[0255] In this embodiment, the training requirement time can be specifically understood as the latest time that the second communication node needs the third communication node to complete VFL training.

[0256] S702. Estimate the training duration based on the longitudinal federated learning training instructions and determine the estimated training duration.

[0257] In a specific example, the third communication node responds to the VFL training instruction to estimate the training duration of the model that needs to initiate VFL training, and obtains the estimated training duration.

[0258] S703. Compare the estimated training duration with the required training time, and initiate vertical federated learning training based on the comparison results.

[0259] In a specific example, after determining the estimated training duration, the third communication node compares the estimated training duration with the training requirement time received by the second communication node to determine whether the VFL training initiated by the third communication node can be completed before the latest time when the second communication node requires the third communication node to complete the VFL training. Based on the comparison result, it can be determined whether the third communication node needs to perform VFL training. If it is determined that the third communication node needs to perform VFL training, the third communication node directly initiates VFL training.

[0260] In one embodiment, the longitudinal federated learning training instruction also includes an analysis identifier.

[0261] In one embodiment, estimating training duration based on longitudinal federated learning training instructions includes:

[0262] Training time is estimated based on the analysis labels.

[0263] In one embodiment, the training time requirement includes at least one of the following:

[0264] Training completion time requirements;

[0265] Training duration requirements.

[0266] In one embodiment, estimating the training duration includes at least one of the following:

[0267] Training duration for vertical federated learning;

[0268] The time required to complete the longitudinal federated learning training.

[0269] In one embodiment, initiating longitudinal federated learning training based on the comparison results includes:

[0270] If the comparison results show that the estimated training time is no later than the required training time, then initiate vertical federated learning training.

[0271] In one embodiment, initiating longitudinal federated learning training based on the comparison results includes:

[0272] If the comparison result shows that the estimated training time is later than the required training time, send an error code and the estimated training time to the second communication node.

[0273] Time to receive new training requirements sent by the second communication node;

[0274] The estimated training duration is compared with the new training requirement time, and if the comparison result shows that the estimated training duration is no later than the new training requirement time, vertical federated learning training is initiated.

[0275] In a specific example, if the comparison result shows that the estimated training time is later than the required training time (i.e., the latest time for the third communication node to complete VFL training is later than the latest time required by the second communication node), the third communication node can send an error code to the second communication node to inform it that it cannot complete training on time, and simultaneously send the estimated training time to the second communication node. If the second communication node decides that the model still needs to be trained by the third communication node, the third communication node will receive a new deliberation time from the second communication node. The third communication node will then compare the estimated training time with the new required training time in the same way as described above. If the comparison result shows that the estimated training time is not later than the new required training time, it is determined that the model trained by the third communication node can meet the requirements of the second communication node, and at this point, the third communication node directly initiates VFL training.

[0276] In one embodiment, the second communication node is the model training logic function; the third communication node is the application function.

[0277] In one exemplary embodiment, Figure 11 is a flowchart of a communication method provided in an embodiment of this application. This method can be applied to the longitudinal federated learning process of a 5G system, where longitudinal federated learning training and / or inference triggering is performed on communication nodes participating in the longitudinal federated learning. This method can be executed by a communication device, which can be implemented by software and / or hardware and integrated on the communication node. This method can be applied to a second communication node. In this embodiment of the application, the second communication node can be an NWDAF containing AnLF that serves as an interaction node with the consumer network element among the longitudinal federated learning participants, as shown in Figure 2 or Figure 3. This embodiment of the application does not limit this.

[0278] As shown in Figure 11, the communication method provided in this embodiment of the application specifically includes the following steps:

[0279] S801, Send a vertical federated learning training instruction to the third communication node.

[0280] The training instructions for vertical federated learning include the time required for training.

[0281] In one embodiment, the longitudinal federated learning training instruction also includes an analysis identifier.

[0282] In one embodiment, the training time requirement includes at least one of the following:

[0283] Training completion time requirements;

[0284] Training duration requirements.

[0285] In one embodiment, after sending the longitudinal federated learning training instruction to the third communication node, the method further includes:

[0286] Receive error codes and estimated training duration sent by the third communication node;

[0287] Send the new training request time to the third communication node.

[0288] In one embodiment, estimating the training duration includes at least one of the following:

[0289] Training duration for vertical federated learning;

[0290] The time required to complete the longitudinal federated learning training.

[0291] In one embodiment, the second communication node is the model training logic function; the third communication node is the application function.

[0292] The communication method of this application is illustrated below through some exemplary schemes. In the following schemes, NWDAF (VFL server) containing MTLF refers to the first communication node, NWDAF containing AnLF refers to the second communication node, and AF (VFL server) refers to the third communication node.

[0293] Solution 1: This solution provides a triggering example for training and inference when the vertical federated learning process only includes AnLF and MTLF, and MTLF does not directly trigger analysis. Figure 12 is a timing example diagram of a vertical federated learning triggering process provided by an embodiment of this application. As shown in Figure 12, it may specifically include the following steps:

[0294] 0) AnLF receives an analysis request that includes: Analysis Requirement Time: indicating the latest time the consumer expects to receive the analysis data provided by NWDAF.

[0295] 1) If an NWDAF containing AnLF wants to perform inference but has no model, it will discover an NWDAF containing MTLF from the NRF and send a model subscription request to the NWDAF containing MTLF using Nnwdaf_MLModelProvision_Subscribe. AnLF may carry the following two parameters in the Nnwdaf_MLModelProvision_Subscribe service:

[0296] Analytics ID: Used to distinguish the purpose of the model;

[0297] Model Requirement Time: Indicates the latest time that consumers expect to receive the model.

[0298] 2) The NWDAF containing MTLF acts as a VFL server to estimate the VFL training duration or the estimated time to complete VFL training based on the received Analytics ID.

[0299] If the VFL training completes later than the time AnLF requires for the model, return an error message to AnLF indicating that the VFL model training cannot be completed within the required time. It can also return the estimated time to complete VFL training or the duration of VFL training.

[0300] MTLF will also return the error code: No model, because the model was trained by VFL.

[0301] 3) This subscription process terminates. If AnLF deems the training time too long, it may seek other MTLF models. Afterward, the VFL training or inference process will not be triggered.

[0302] 4) If AnLF deems the training duration sufficient, it sends an analytics subscription request to MTLF.

[0303] 5) MTLF, acting as the VFL server, initiates VFL training.

[0304] 6) The MTLF, acting as the VFL server, initiates VFL inference.

[0305] 7) The VFL server returns the inference results (also known as the analysis results) to the NWDAF containing AnLF.

[0306] Solution 2: A triggering example is provided for training and inference when the vertical federated learning triggering process only includes AnLF and MTLF, and MTLF can directly trigger analysis. Figure 13 is a timing example diagram of a vertical federated learning triggering process provided by an embodiment of this application. As shown in Figure 13, it may specifically include the following steps:

[0307] 0) AnLF receives an analysis request that includes: Analysis Requirement Time: indicating the latest time the consumer expects to receive the analysis data provided by NWDAF.

[0308] 1) If an NWDAF containing AnLF wants to perform inference but has no model, it will discover an NWDAF containing MTLF from the NRF and send a model subscription request to the NWDAF containing MTLF using Nnwdaf_MLModelProvision_Subscribe.

[0309] 2) NWDAF containing MTLF returns the error code as a VFL server: No model because the model was trained by VFL.

[0310] 3) The subscription process for this model is terminated.

[0311] 4) AnLF sends an analysis request to the VFL server, which includes: Analysis Requirement Time: indicating the latest time AnLF expects to receive the analysis data provided by NWDAF.

[0312] 5) Include MTLF in the NWDAF as the VFL server to estimate the total duration of VFL training + VFL inference or the (latest) completion time of VFL training and / or inference.

[0313] If the (latest) completion time information for VFL training and / or inference is later than the analysis requirement time, MTLF returns an indication to AnLF that analysis cannot be returned on time.

[0314] If the (latest) completion time of VFL training and / or inference is no later than the analysis requirement time, then MTLF executes step 7) and its subsequent steps.

[0315] 6) AnLF sends a new analysis request to the VFL server, which includes: Adjusted analysis request time: indicating the latest time AnLF expects to receive the analysis data provided by NWDAF.

[0316] 7) The VFL server initiates VFL training.

[0317] 8) The VFL server initiates VFL inference.

[0318] 9) The VFL server returns the inference results (also known as the analysis results) to the NWDAF containing AnLF.

[0319] Solution 3: A triggering example is provided for a scenario where the vertical federated learning triggering process only includes AnLF and AF, and AF cannot directly trigger training. Figure 14 is a timing example diagram of a vertical federated learning triggering process provided by an embodiment of this application. As shown in Figure 14, it may specifically include the following steps:

[0320] 1) Before sending a training request to the AF, AnLF queries the AF for (VFL) training duration information. The query includes the analysis ID.

[0321] The training duration information may include the VFL training duration or the estimated time to complete VFL training.

[0322] 2) The NWDAF containing MTLF acts as the VFL server, estimating the VFL training duration or the estimated time to complete VFL training based on the received Analytics ID, and returns it to AnLF.

[0323] 3) If AnLF deems the VFL training duration sufficient, it sends a VFL training request to the AF. If AnLF deems the VFL training duration too long, it sends VFL training requests to other AFs.

[0324] 4) AF initiates VFL training.

[0325] Solution 4: A triggering example is provided for a vertical federated learning process that only includes AnLF and AF, and where AF can directly trigger training. Figure 15 is a timing example diagram of a vertical federated learning triggering process provided by an embodiment of this application. As shown in Figure 15, it may specifically include the following steps:

[0326] 1) AnLF directs AF (VFL) model training, which includes the latest time to complete training or the training duration requirement, and includes the analytics ID.

[0327] 2) The AF performs an evaluation. If the estimated (VFL) training completion time is no later than the latest training completion time or the training duration is less than or equal to the requirement, then (VFL) training is initiated. If the estimated (VFL) training completion time is later than the required latest training completion time, an error code is returned: Training cannot be completed on time, including the (VFL) training duration or the latest time that (VFL) training can be completed.

[0328] 3) AnLF directs AF (VFL) model training, which includes an adjusted latest time to complete training or an adjusted training duration requirement, which includes the analytics ID.

[0329] 4) AF performs the evaluation as in step 2) again. If it passes, VFL training is initiated.

[0330] In one exemplary embodiment, FIG16 is a schematic diagram of a communication device provided in an embodiment of the present application. The communication device is applied to a first communication node. As shown in FIG16, the device includes:

[0331] The first request receiving module 910 is configured to receive a first model subscription request sent by the second communication node. The first model subscription request includes the model's required time.

[0332] The first duration determination module 920 is configured to estimate the training duration based on the first model subscription request and determine the estimated training duration.

[0333] The first instruction feedback module 930 is configured to compare the estimated training time with the model's required time, and to feed back the first model training instruction to the second communication node based on the comparison result.

[0334] In one embodiment, estimating the training duration includes at least one of the following:

[0335] Training duration for vertical federated learning;

[0336] The time required to complete the longitudinal federated learning training.

[0337] In one embodiment, the first model subscription request also includes an analytics identifier.

[0338] In one embodiment, estimating the training duration based on the first model subscription request includes:

[0339] Training time is estimated based on the analysis labels.

[0340] In one embodiment, feeding back a first model training instruction to a second communication node based on the comparison result includes:

[0341] If the comparison result indicates that the estimated training time is later than the model's required time, send at least one of the following to the second communication node:

[0342] Error indication;

[0343] Estimated training time;

[0344] Error indications and estimated training time.

[0345] In one embodiment, the error indication includes a reason code;

[0346] The reason code includes:

[0347] The model cannot be trained within the required time or made available within the required time.

[0348] In one embodiment, the first communication node is the model training logic function; the second communication node is the analysis logic function.

[0349] In one exemplary embodiment, FIG17 is a schematic diagram of a communication device provided in an embodiment of this application. The communication device is applied to a second communication node. As shown in FIG17, the device includes:

[0350] The first request sending module 1010 is configured to receive analysis request time and send a first model subscription request to the first communication node according to the analysis request time. The first model subscription request includes the model request time.

[0351] The first instruction receiving module 1020 is configured to receive the first model training instruction fed back by the first communication node.

[0352] In one embodiment, the first model subscription request also includes an analytics identifier.

[0353] In one embodiment, receiving a first model training instruction fed back by a first communication node includes:

[0354] Receive at least one of the following feedbacks from the first communication node:

[0355] Error indication;

[0356] Estimated training time;

[0357] Error indications and estimated training time.

[0358] In one embodiment, the error indication includes a reason code;

[0359] Reason codes include:

[0360] The model cannot be trained within the required time or made available within the required time.

[0361] In one embodiment, after receiving the first model training instruction fed back by the first communication node, the method further includes:

[0362] If the estimated training time exceeds a preset time threshold, a new first communication node is discovered.

[0363] In one embodiment, estimating the training time includes at least one of the following:

[0364] Training duration for vertical federated learning;

[0365] The time required to complete the longitudinal federated learning training.

[0366] In one embodiment, the first communication node is a model training logic function; the second communication node is an analysis logic function.

[0367] In one exemplary embodiment, FIG18 is a schematic diagram of a communication device provided in an embodiment of the present application. The communication device is applied to a first communication node. As shown in FIG18, the device includes:

[0368] The second time receiving module 1110 is configured to receive the analysis requirement time sent by the second communication node.

[0369] The second duration determination module 1120 is configured to perform analysis duration estimation and determine the estimated analysis duration.

[0370] The first learning initiation module 1130 is configured to compare the estimated analysis time with the analysis requirement time, and initiate vertical federated learning based on the comparison results.

[0371] In one embodiment, the analysis requirement time also includes an analysis identifier.

[0372] In one embodiment, the analysis duration estimation includes:

[0373] Estimate the analysis duration based on the analysis identifier.

[0374] In one embodiment, the estimated analysis duration includes at least one of the following:

[0375] The duration of training and reasoning in vertical federated learning;

[0376] The time required to complete vertical federated learning training and reasoning;

[0377] Vertical federated learning inference duration;

[0378] Time required to complete longitudinal federated learning inference;

[0379] Vertical federated learning training duration;

[0380] Time required to complete longitudinal federated learning training;

[0381] The duration of the analysis;

[0382] The time required to generate the analysis.

[0383] In one embodiment, initiating longitudinal federated learning based on the comparison results includes:

[0384] If the comparison results show that the estimated analysis time is no later than the required analysis time, initiate longitudinal federated learning training and / or inference.

[0385] In one embodiment, initiating longitudinal federated learning based on the comparison results includes:

[0386] When the comparison results show that the estimated analysis time is later than the required analysis time, sending the required analysis time to the second communication node cannot meet the indication and the estimated analysis time.

[0387] Time to receive new analysis requests from the second communication node;

[0388] Compare the estimated analysis time with the new analysis requirement time, and if the comparison result shows that the estimated analysis time is no later than the new analysis requirement time, initiate vertical federated learning training and / or inference.

[0389] In one embodiment, after initiating longitudinal federated learning training and / or inference, the method further includes:

[0390] The reasoning or analysis results are sent to the second communication node.

[0391] In one embodiment, initiating longitudinal federated learning based on the comparison results includes:

[0392] When the comparison results show that the estimated analysis time is later than the required analysis time, sending the required analysis time to the second communication node cannot meet the indication and the estimated analysis time.

[0393] If no new analysis request is received from the second communication node, longitudinal federated learning training and / or inference will not be initiated.

[0394] In one embodiment, the first communication node is the model training logic function; the second communication node is the analysis logic function.

[0395] In one exemplary embodiment, FIG19 is a schematic diagram of a communication device provided in an embodiment of this application. The communication device is applied to a second communication node. As shown in FIG19, the device includes:

[0396] The second time sending module 1210 is configured to receive the analysis request time and send the analysis request time to the first communication node.

[0397] In one embodiment, after sending the analysis requirement time to the first communication node, the method further includes:

[0398] The analysis request time received from the first communication node cannot meet the indication and estimated analysis duration.

[0399] In one embodiment, the estimated analysis duration includes at least one of the following:

[0400] The duration of training and reasoning in vertical federated learning;

[0401] The time required to complete vertical federated learning training and reasoning;

[0402] Vertical federated learning inference duration;

[0403] Time required to complete longitudinal federated learning inference;

[0404] Vertical federated learning training duration;

[0405] Time required to complete longitudinal federated learning training;

[0406] The duration of the analysis;

[0407] The time required to generate the analysis.

[0408] In one embodiment, after receiving the analysis request time sent by the first communication node, which cannot meet the indication and estimated analysis duration, the method further includes:

[0409] Send the new analysis request time to the first communication node.

[0410] In one embodiment, after sending the analysis request time to the first communication node, or sending a new analysis request time to the first communication node, the method further includes:

[0411] Receive the reasoning or analysis results sent by the first communication node.

[0412] In one embodiment, the analysis requirement time also includes an analysis identifier.

[0413] In one embodiment, the first communication node is the model training logic function; the second communication node is the analysis logic function.

[0414] In one exemplary embodiment, FIG20 is a schematic diagram of a communication device provided in an embodiment of the present application. The communication device is applied to a second communication node. As shown in FIG20, the device includes:

[0415] The third request sending module 1310 is configured to send a training duration request to the third communication node.

[0416] The third duration receiving module 1320 is configured to receive the estimated training duration fed back by the third communication node.

[0417] In one embodiment, the training duration request includes an analysis identifier.

[0418] In some examples, a training duration request can be used to request the duration of VFL training or to estimate the time required to complete VFL training.

[0419] In one embodiment, estimating the training duration includes at least one of the following:

[0420] Training duration for vertical federated learning;

[0421] The time required to complete the longitudinal federated learning training.

[0422] In one embodiment, after receiving the estimated training duration from the third communication node, the method further includes:

[0423] If the estimated training duration meets the preset training duration requirement, a vertical federated learning training request is sent to the third communication node.

[0424] In one embodiment, after receiving the estimated training duration from the third communication node, the method further includes:

[0425] If the estimated training duration does not meet the preset training duration requirement, a vertical federated learning training request is initiated to other second communication nodes.

[0426] In one embodiment, the second communication node is the model training logic function; the third communication node is the application function.

[0427] In one exemplary embodiment, FIG21 is a schematic diagram of a communication device provided in an embodiment of the present application. The communication device is applied to a third communication node. As shown in FIG21, the device includes:

[0428] The third request receiving module 1410 is configured to receive training duration requests sent by the second communication node.

[0429] The third duration feedback module 1420 is configured to estimate the training duration based on the training duration request, determine the estimated training duration, and feed back the estimated training duration to the second communication node.

[0430] In one embodiment, the training duration request includes an analysis identifier.

[0431] In one embodiment, estimating the training duration based on the training duration request includes:

[0432] Training time is estimated based on the analysis labels.

[0433] In one embodiment, estimating the training duration includes at least one of the following:

[0434] Training duration for vertical federated learning;

[0435] The time required to complete the longitudinal federated learning training.

[0436] In one embodiment, after feeding back the estimated training duration to the second communication node, the method further includes:

[0437] Receive the vertical federated learning training request sent by the second communication node;

[0438] Initiate vertical federated learning training in response to a request for vertical federated learning training.

[0439] In one exemplary embodiment, FIG22 is a schematic diagram of a communication device provided in an embodiment of the present application. The communication device is applied to a third communication node. As shown in FIG22, the device includes:

[0440] The fourth instruction receiving module 1510 is configured to receive the longitudinal federated learning training instruction sent by the second communication node. The longitudinal federated learning training instruction includes the required training time.

[0441] The fourth duration determination module 1520 is configured to estimate the training duration based on the longitudinal federated learning training instructions and determine the estimated training duration.

[0442] The fourth training initiation module 1530 is configured to compare the estimated training duration with the required training time and initiate vertical federated learning training based on the comparison results.

[0443] In one embodiment, the longitudinal federated learning training instruction also includes an analysis identifier.

[0444] In one embodiment, estimating training duration based on longitudinal federated learning training instructions includes:

[0445] Training time is estimated based on the analysis labels.

[0446] In one embodiment, the training time requirement includes at least one of the following:

[0447] Training completion time requirements;

[0448] Training duration requirements.

[0449] In one embodiment, estimating the training duration includes at least one of the following:

[0450] Training duration for vertical federated learning;

[0451] The time required to complete the longitudinal federated learning training.

[0452] In one embodiment, initiating longitudinal federated learning training based on the comparison results includes:

[0453] If the comparison results show that the estimated training time is no later than the required training time, then initiate vertical federated learning training.

[0454] In one embodiment, initiating longitudinal federated learning training based on the comparison results includes:

[0455] If the comparison result shows that the estimated training time is later than the required training time, send an error code and the estimated training time to the second communication node.

[0456] Time to receive new training requirements sent by the second communication node;

[0457] The estimated training duration is compared with the new training requirement time, and if the comparison result shows that the estimated training duration is no later than the new training requirement time, vertical federated learning training is initiated.

[0458] In one embodiment, the second communication node is the model training logic function; the third communication node is the application function.

[0459] In one exemplary embodiment, FIG23 is a schematic diagram of a communication device provided in an embodiment of the present application. The communication device is applied to a second communication node. As shown in FIG23, the device includes:

[0460] The fourth instruction sending module 1610 is configured to send a longitudinal federated learning training instruction to the third communication node. The longitudinal federated learning training instruction includes the required training time.

[0461] In one embodiment, the longitudinal federated learning training instruction also includes an analysis identifier.

[0462] In one embodiment, the training time requirement includes at least one of the following:

[0463] Training completion time requirements;

[0464] Training duration requirements.

[0465] In one embodiment, after sending the longitudinal federated learning training instruction to the third communication node, the method further includes:

[0466] Receive error codes and estimated training duration sent by the third communication node;

[0467] Send the new training request time to the third communication node.

[0468] In one embodiment, estimating the training duration includes at least one of the following:

[0469] Training duration for vertical federated learning;

[0470] The time required to complete the longitudinal federated learning training.

[0471] In one embodiment, the second communication node is the model training logic function; the third communication node is the application function.

[0472] This application embodiment also provides a communication node. Figure 24 is a structural schematic diagram of a communication node provided in this application embodiment. As shown in Figure 24, the communication node provided in this application embodiment includes a memory 1720, a processor 1710, and a computer program stored in the memory and executable on the processor. When the processor 1710 executes the program, it implements the above-mentioned communication method.

[0473] The communication node may also include a memory 1720; the processor 1710 in the communication node may be one or more, with one processor 1710 as an example in FIG24; the memory 1720 is used to store one or more programs; the one or more programs are executed by the one or more processors 1710, so that the one or more processors 1710 implement the communication method as described in the embodiments of this application.

[0474] The communication node also includes: a communication device 1730, an input device 1740, and an output device 1750.

[0475] The processor 1710, memory 1720, communication device 1730, input device 1740 and output device 1750 in the communication node can be connected by a bus or other means. Figure 24 shows an example of connection by bus.

[0476] Input device 1740 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the communication node. Output device 1750 may include display devices such as a display screen.

[0477] The communication device 1730 may include a receiver and a transmitter. The communication device 1730 is configured to perform information transmission and reception communication under the control of the processor 1710.

[0478] The memory 1720, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, and modules, such as program instructions / modules corresponding to the communication method described in the embodiments of this application (e.g., a first request receiving module 910, a first duration determining module 920, a first indication feedback module 930; or a first request sending module 1010, a first indication receiving module 1020; or a second time receiving module 1110, a second duration determining module 1120, a first learning initiation module 1130; or a second time sending module 1210; or a third request sending module 1310, a third duration receiving module 1320; or a third request receiving module 1410, a third duration feedback module 1420; or a fourth indication receiving module 1510, a fourth duration determining module 1520, a fourth training initiation module 1530; or a fourth indication sending module 1610). The memory 1720 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and application programs required for at least one function; the data storage area may store data created according to the use of the communication node, etc. Furthermore, memory 1720 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, memory 1720 may further include memory remotely located relative to processor 1710, which can be connected to a communication node via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0479] This application also provides a storage medium storing a computer program, which, when executed by a processor, implements any of the communication methods described in this application.

[0480] Optionally, the communication method, applied to the first communication node, includes: receiving a training duration request sent by the second communication node; estimating the training duration based on the training duration request, determining the estimated training duration, and feeding back the estimated training duration to the second communication node.

[0481] Optionally, the communication method, applied to the second communication node, includes: receiving a first model subscription request sent by the third communication node, and sending a training duration request to the first communication node according to the first model subscription request; wherein the first model subscription request includes the model required time; receiving an estimated training duration sent by the first communication node; comparing the estimated training duration with the model required time, and sending a longitudinal federated learning training instruction to the first communication node according to the comparison result.

[0482] Optionally, the communication method, applied to the first communication node, includes: receiving a first required time sent by the second communication node; estimating the training duration to determine the estimated training duration; comparing the estimated training duration with the first required time, and initiating longitudinal federated learning training based on the comparison result.

[0483] Optionally, the communication method, applied to a second communication node, includes: receiving a first model subscription request sent by a third communication node; wherein the first model subscription request includes a model demand time; and sending the first demand time to the first communication node.

[0484] Optionally, the communication method, applied to a third communication node, includes: receiving a first instruction; wherein the first instruction includes a first identifier; and sending a first analysis request to a first communication node; wherein the first analysis request includes the first identifier.

[0485] The computer storage medium in this application embodiment can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable CD-ROM, optical storage device, magnetic storage device, or any suitable combination thereof. The computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0486] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device.

[0487] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency (RF), etc., or any suitable combination thereof.

[0488] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0489] Optionally, embodiments of the present invention also provide a computer program product, including a computer program that, when executed by a processor, implements the communication method provided in any embodiment of the present invention.

[0490] The above description is merely an exemplary embodiment of this application and is not intended to limit the scope of protection of this application.

[0491] Those skilled in the art will understand that the term user terminal encompasses any suitable type of wireless user equipment, such as mobile phones, portable data processing devices, portable web browsers, or vehicle-mounted mobile stations.

[0492] Generally, the various embodiments of this application can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. For example, some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device, although this application is not limited thereto.

[0493] Embodiments of this application can be implemented by executing computer program instructions through the data processor of a mobile device, for example, in a processor entity, or through hardware, or through a combination of software and hardware. The computer program instructions can be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages.

[0494] Any block diagram of logical flow in the accompanying drawings of this application may represent program steps, or may represent interconnected logic circuits, modules, and functions, or may represent a combination of program steps and logic circuits, modules, and functions. The computer program may be stored on memory. Memory may be of any type suitable to the local technical environment and may be implemented using any suitable data storage technology, such as, but not limited to, read-only memory (ROM), random access memory (RAM), optical storage devices and systems (Digital Video Disc (DVD) or Compact Disk (CD), etc.). Computer-readable media may include non-transitory storage media. Data processors may be of any type suitable to the local technical environment, such as, but not limited to, general-purpose computers, special-purpose computers, microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and processors based on multi-core processor architectures.

[0495] A detailed description of exemplary embodiments of this application has been provided above through exemplary and non-limiting examples. However, various modifications and adjustments to the above embodiments will be apparent to those skilled in the art when considered in conjunction with the accompanying drawings and claims, without departing from the scope of this application. Therefore, the proper scope of this application will be determined by the claims.

Claims

1. A communication method, characterized in that, The method is applied to a first communication node and includes: receiving a first model subscription request sent by a second communication node; wherein the first model subscription request includes a model requirement time; estimating the training duration based on the first model subscription request to determine an estimated training duration; comparing the estimated training duration with the model requirement time, and feeding back a first model training instruction to the second communication node based on the comparison result.

2. The communication method according to claim 1, characterized in that, The estimated training duration includes at least one of the following: the training duration of longitudinal federated learning; the time required to complete the training of longitudinal federated learning.

3. The communication method according to claim 1, characterized in that, The first model subscription request also includes an analysis identifier.

4. The communication method according to claim 3, characterized in that, The step of estimating the training duration based on the first model subscription request includes: estimating the training duration based on the analysis identifier.

5. The communication method according to claim 1, characterized in that, The step of feeding back the first model training instruction to the second communication node based on the comparison result includes: when the comparison result indicates that the estimated training time is later than the model's required time, sending at least one of the following to the second communication node: an error indication; the estimated training time; or an error indication and the estimated training time.

6. The communication method according to claim 5, characterized in that, The error indication includes a reason code; the reason code includes: the model cannot be trained within the required time, or the model cannot be made available within the required time.

7. The communication method according to any one of claims 1-6, characterized in that, The first communication node is for model training logic; the second communication node is for analysis logic.

8. A communication method, characterized in that, The method is applied to a second communication node and includes: receiving an analysis request time and sending a first model subscription request to a first communication node based on the analysis request time; wherein the first model subscription request includes a model request time; and receiving a first model training instruction fed back by the first communication node.

9. The communication method according to claim 8, characterized in that, The first model subscription request also includes an analysis identifier.

10. The communication method according to claim 9, characterized in that, The receipt of the first model training instruction fed back by the first communication node includes: receiving at least one of the following fed back by the first communication node: error indication; estimated training time; error indication and estimated training time.

11. The communication method according to claim 10, characterized in that, The error indication includes a reason code; the reason code includes: the model cannot be trained within the required time, or the model cannot be made available within the required time.

12. The communication method according to claim 10, characterized in that, After receiving the first model training instruction from the first communication node, the method further includes: if the estimated training time exceeds a preset time threshold, discovering a new first communication node.

13. The communication method according to claim 10, characterized in that, The estimated training time includes at least one of the following: the training duration of longitudinal federated learning; the time required to complete the training of longitudinal federated learning.

14. The communication method according to any one of claims 8-13, characterized in that, The first communication node is for model training logic; the second communication node is for analysis logic.

15. A communication method, characterized in that, The method is applied to the first communication node and includes: receiving the analysis request time sent by the second communication node; estimating the analysis duration to determine the estimated analysis duration; comparing the estimated analysis duration with the analysis request time, and initiating vertical federated learning based on the comparison result.

16. The communication method according to claim 15, characterized in that, The analysis requirement time also includes analysis identifiers.

17. The communication method according to claim 16, characterized in that, The process of estimating the analysis duration includes estimating the analysis duration based on the analysis identifier.

18. The communication method according to claim 15, characterized in that, The estimated analysis duration includes at least one of the following: the duration of vertical federated learning training and inference; the time to complete vertical federated learning training and inference; the duration of vertical federated learning inference; the time to complete vertical federated learning inference; the duration of vertical federated learning training; the time to complete vertical federated learning training; the duration of generating analysis; and the time to generate analysis.

19. The communication method according to claim 15, characterized in that, The step of initiating longitudinal federated learning based on the comparison results includes: initiating longitudinal federated learning training and / or inference when the comparison results indicate that the estimated analysis time is not later than the required analysis time.

20. The communication method according to claim 15, characterized in that, The step of initiating longitudinal federated learning based on the comparison result includes: if the comparison result indicates that the estimated analysis duration is later than the analysis requirement time, sending an indication that the analysis requirement time cannot be met and the estimated analysis duration to the second communication node; receiving a new analysis requirement time sent by the second communication node; comparing the estimated analysis duration and the new analysis requirement time, and if the comparison result indicates that the estimated analysis duration is not later than the new analysis requirement time, initiating longitudinal federated learning training and / or inference.

21. The communication method according to claim 19 or 20, characterized in that, After initiating the vertical federated learning training and / or inference, the method further includes: sending the inference results or analysis results to the second communication node.

22. The communication method according to claim 15, characterized in that, The step of initiating longitudinal federated learning based on the comparison result includes: if the comparison result indicates that the estimated analysis duration is later than the analysis requirement time, sending an indication that the analysis requirement time cannot be met and the estimated analysis duration to the second communication node; and not initiating longitudinal federated learning training and / or inference if no new analysis requirement time is received from the second communication node.

23. The communication method according to any one of claims 15-20 and 22, characterized in that, The first communication node is for model training logic; the second communication node is for analysis logic.

24. A communication method, characterized in that, The method applied to the second communication node includes: receiving the analysis request time and sending the analysis request time to the first communication node.

25. The communication method according to claim 24, characterized in that, After sending the analysis request time to the first communication node, the method further includes: receiving an indication and estimated analysis duration from the first communication node that the analysis request time cannot be met.

26. The communication method according to claim 25, characterized in that, The estimated analysis duration includes at least one of the following: the duration of vertical federated learning training and inference; the time to complete vertical federated learning training and inference; the duration of vertical federated learning inference; the time to complete vertical federated learning inference; the duration of vertical federated learning training; the time to complete vertical federated learning training; the duration of generating analysis; and the time to generate analysis.

27. The communication method according to claim 25, characterized in that, After receiving the analysis request time sent by the first communication node, which cannot meet the indication and estimated analysis duration, the method further includes: sending a new analysis request time to the first communication node.

28. The communication method according to any one of claims 24-27, characterized in that, After sending the analysis request time to the first communication node, or sending a new analysis request time to the first communication node, the method further includes: receiving the inference result or analysis result sent by the first communication node.

29. The communication method according to any one of claims 24-27, characterized in that, The analysis requirement time also includes analysis identifiers.

30. The communication method according to any one of claims 24-27, characterized in that, The first communication node is for model training logic; the second communication node is for analysis logic.

31. A communication method, characterized in that, The method is applied to the second communication node and includes: sending a training duration request to the third communication node; and receiving the estimated training duration fed back by the third communication node.

32. The communication method according to claim 31, characterized in that, The training duration request includes an analysis identifier.

33. The communication method according to claim 31, characterized in that, The estimated training duration includes at least one of the following: the training duration of longitudinal federated learning; the time required to complete the training of longitudinal federated learning.

34. The communication method according to claim 31, characterized in that, After receiving the estimated training duration from the third communication node, the method further includes: if the estimated training duration meets the preset training duration requirement, sending a vertical federated learning training request to the third communication node.

35. The communication method according to claim 31, characterized in that, After receiving the estimated training duration from the third communication node, the method further includes: if the estimated training duration does not meet the preset training duration requirement, initiating a vertical federated learning training request to other third communication nodes.

36. The communication method according to any one of claims 31-35, characterized in that, The second communication node is for model training logic; the third communication node is for application functionality.

37. A communication method, characterized in that, The method is applied to a third communication node and includes: receiving a vertical federated learning training instruction sent by a second communication node; the vertical federated learning training instruction includes a training requirement time; estimating the training duration based on the vertical federated learning training instruction to determine the estimated training duration; comparing the estimated training duration with the training requirement time, and initiating vertical federated learning training based on the comparison result.

38. The communication method according to claim 37, characterized in that, The longitudinal federated learning training instructions also include analysis identifiers.

39. The communication method according to claim 38, characterized in that, The step of estimating training duration based on the longitudinal federated learning training instruction includes: estimating training duration based on the analysis identifier.

40. The communication method according to claim 37, characterized in that, The training time requirement includes at least one of the following: training completion time requirement; training duration requirement.

41. The communication method according to claim 37, characterized in that, The estimated training duration includes at least one of the following: the training duration of longitudinal federated learning; the time required to complete the training of longitudinal federated learning.

42. The communication method according to claim 37, characterized in that, The step of initiating vertical federated learning training based on the comparison results includes: initiating vertical federated learning training when the comparison results indicate that the estimated training duration is not later than the required training time.

43. The communication method according to claim 37, characterized in that, The step of initiating vertical federated learning training based on the comparison result includes: sending an error code and the estimated training duration to the second communication node when the comparison result indicates that the estimated training duration is later than the required training time; receiving a new required training time from the second communication node; comparing the estimated training duration with the new required training time; and initiating vertical federated learning training when the comparison result indicates that the estimated training duration is not later than the new required training time.

44. The communication method according to any one of claims 37-43, characterized in that, The second communication node is for model training logic; the third communication node is for application functionality.

45. A communication node, characterized in that, include: The program includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for implementing communication between the processor and the memory, wherein the program, when executed by the processor, implements the steps of the communication method as described in any one of claims 1-44.

46. ​​A storage medium for computer-readable storage, characterized in that, The storage medium stores one or more programs, which can be executed by one or more processors to implement the steps of the communication method as described in any one of claims 1-44.

47. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the communication method as described in any one of claims 1-44.

Images