WO2023093238A1

WO2023093238A1 - Method and apparatus for performing service processing by using learning model

Info

Publication number: WO2023093238A1
Application number: PCT/CN2022/119866
Authority: WO
Inventors: 崔琪楣; 梁盛源; 赵博睿; 任崇万; 陶小峰
Original assignee: 北京邮电大学
Priority date: 2021-11-29
Filing date: 2022-09-20
Publication date: 2023-06-01
Also published as: CN114302422A

Abstract

The present application discloses a method and apparatus for performing service processing by using a learning model. By applying the technical solution of the present application, a distributed unit (DU) and a centralized unit (CU) of a base station device can be utilized, and jointly form a communication network architecture with an edge node. In this way, the communication network can be subsequently used to aggregate the model parameters of device nodes, and the aggregated model parameter is used to construct a hierarchical federated learning model deployed at a user equipment end or a base station device end. Therefore, the purpose of performing service processing by using the hierarchical federated learning model is achieved.

Description

Method and device for business processing using learning model

technical field

This application relates to data processing technology, especially a method and device for business processing using a learning model.

Background technique

The architecture of the existing communication network refers to the "cloud-edge-end" three-layer intelligent architecture, where the edge intelligence usually refers to the edge server, which is used to handle tasks such as calculations on the user data plane. However, it does not consider the intelligentization of the network control plane and management plane at the edge. In addition, the existing network architecture does not fully reflect the intelligent characteristics of base station equipment.

Therefore, how to design a communication network architecture that can fully utilize various node devices to realize business processing has become a problem to be solved by those skilled in the art.

Contents of the invention

Embodiments of the present application provide a method and device for processing services using a learning model, wherein, according to an aspect of the embodiments of the present application, a method for processing services using a learning model is provided, which is applied to base station equipment, including:

Obtain a communication network architecture composed of a first intelligent layer, a second intelligent layer, and a third intelligent layer, wherein the first intelligent layer and the second intelligent layer are deployed in base station equipment, and the third intelligent layer is deployed in edge nodes;

Using a hierarchical federated learning algorithm and the communication network architecture to generate a hierarchical federated learning model deployed on the user equipment side or the base station equipment side;

Business processing is performed by using the layered federated learning model.

Optionally, in another embodiment based on the above method of the present application, after the acquisition of the communication network architecture composed of the first intelligent layer, the second intelligent layer and the third intelligent layer, further includes:

Obtain the fourth intelligent layer deployed in the cloud server;

performing functional configuration on the fourth intelligent layer according to a preset configuration strategy;

After detecting that the function configuration of the fourth intelligent layer is completed, it is determined to generate a communication network architecture composed of the first intelligent layer, the second intelligent layer, the third intelligent layer, and the fourth intelligent layer.

Optionally, in another embodiment based on the above method of the present application, the use of a hierarchical federated learning algorithm and the communication network architecture to generate a hierarchical federated learning model deployed on the user equipment side or the base station equipment side includes:

Using each of the first intelligent layers to acquire initial model parameters, where the initial model parameters are model parameters obtained by the user equipment or the first intelligent layer using local data for model training;

After using the second intelligent layer to receive the initial model parameters transmitted by each of the first intelligent layers, perform first-level aggregation on the initial model parameters to obtain first aggregated model parameters;

The second intelligent layer sends the first aggregation model parameters to the first intelligent layer, and determines that the aggregation at the first level is completed after reaching the first number.

Optionally, in another embodiment based on the above-mentioned method of the present application, after determining that the first-level aggregation is completed after the first number is reached, the method includes:

The first intelligent layer sends the first aggregated model parameters to the user equipment, so that the user equipment trains an initial learning model according to the first aggregated model parameters; or,

The first intelligent layer trains an initial learning model according to the first aggregated model parameters.

Optionally, in another embodiment based on the above-mentioned method of the present application, after the determination that the first-level aggregation is completed, the method includes:

sending the first aggregation model parameters to the third intelligent layer by each of the second intelligent layers;

performing second-level aggregation on each of the first aggregation model parameters by the third intelligent layer to obtain second aggregation model parameters;

The third intelligent layer sends the second aggregation model parameters to the second intelligent layer, and determines that the aggregation at the second level is completed after reaching a second number of times.

Optionally, in another embodiment based on the above-mentioned method of the present application, after determining that the second-level aggregation is completed after the second number of times is reached, the method includes:

the second intelligence layer sends the second aggregated model parameters to the first intelligence layer; and,

The first intelligent layer sends the second aggregation model parameters to the user equipment, so that the user equipment trains the initial learning model according to the second aggregation model parameters to obtain the hierarchical federated learning model; or ,

The first intelligent layer trains an initial learning model according to the second aggregation model parameters to obtain the hierarchical federated learning model.

Optionally, in another embodiment based on the above method of the present application, after the third intelligent layer sends the second aggregation model parameters to the second intelligent layer, it includes:

If it is determined that there is a fourth intelligent layer, each of the third intelligent layers sends the second aggregation model parameters to the fourth intelligent layer, so that the fourth intelligent layer performs a process on the second aggregation model parameters. After the third level of aggregation, the third aggregation model parameters are obtained;

The fourth intelligent layer sends the third aggregation model parameters to the first intelligent layer step by step, so that the first intelligent layer trains the initial learning model according to the first aggregation model parameters; or ,

The first intelligent layer sends the third aggregated model parameters to the user equipment, so that the user equipment trains an initial learning model according to the third aggregated model parameters to obtain the hierarchical federated learning model.

Optionally, in another embodiment based on the above-mentioned method of this application, the use of the hierarchical federated learning model for business processing includes:

The user equipment uses the hierarchical federated learning model to perform the first service processing; or,

The base station device uses the layered federated learning model to perform the second service processing.

Optionally, in another embodiment based on the above method of the present application, the first intelligent layer and the second intelligent layer are deployed in the base station equipment, including:

The first intelligent layer is deployed in a distributed unit DU of the base station device, and the second intelligent layer is deployed in a centralized unit CU of the base station device; or,

The first intelligent layer is deployed in small base station equipment, and the second intelligent layer is deployed in macro base station equipment.

Wherein, according to another aspect of the embodiment of the present application, a device for using a learning model for business processing is provided, including:

An acquisition module configured to acquire a communication network architecture composed of a first intelligent layer, a second intelligent layer, and a third intelligent layer, wherein the first intelligent layer and the second intelligent layer are deployed in base station equipment, and the third intelligent layer Deployed in edge nodes;

The generating module is configured to generate a hierarchical federated learning model deployed on the user equipment side or the base station equipment side by using the hierarchical federated learning algorithm and the communication network architecture;

The processing module is configured to use the hierarchical federated learning model to perform business processing.

According to still another aspect of the embodiments of the present application, an electronic device is provided, including:

memory for storing executable instructions; and

The display is used for displaying with the memory to execute the executable instruction so as to complete the operation of any one of the above-mentioned methods for business processing using a learning model.

According to still another aspect of the embodiments of the present application, there is provided a computer-readable storage medium for storing computer-readable instructions, and when the instructions are executed, any one of the above-mentioned methods for business processing using a learning model is performed operation.

In this application, the communication network architecture composed of the first intelligent layer, the second intelligent layer and the third intelligent layer can be obtained, wherein the first intelligent layer and the second intelligent layer are deployed in the base station equipment, and the third intelligent layer is deployed in the edge In the node; use the hierarchical federated learning algorithm and the communication network architecture to generate a hierarchical federated learning model deployed on the user equipment side or the base station equipment side; and use the hierarchical federated learning model for business processing. By applying the technical solution of the present application, the distributed unit DU and the centralized unit CU of the base station equipment can be used to form a communication network architecture together with the edge nodes. In the future, the communication network can be used to aggregate the model parameters of each device node in the network, and the aggregated model parameters can be used to construct a hierarchical federated learning model deployed on the user equipment side or the base station equipment side. Then realize the purpose of using the hierarchical federated learning model for business processing.

The technical solutions of the present application will be described in further detail below with reference to the drawings and embodiments.

Description of drawings

The accompanying drawings, which constitute a part of this specification, illustrate the embodiments of the application and, together with the description, serve to explain the principles of the application.

The present application can be more clearly understood from the following detailed description with reference to the accompanying drawings, in which:

FIG. 1 is a schematic diagram of a method of using a learning model for business processing proposed in this application;

Fig. 2 is a schematic diagram of the system architecture applied to the business processing method using the learning model proposed by the present application;

FIG. 3 is a schematic structural diagram of an electronic device using a learning model for business processing proposed in this application;

FIG. 4 is a schematic structural diagram of an electronic device using a learning model for business processing proposed in this application.

Detailed ways

Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangements of components and steps, numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present application unless specifically stated otherwise.

At the same time, it should be understood that, for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

The following description of at least one exemplary embodiment is merely illustrative in nature and not intended as any limitation of the application, its application or uses.

Techniques, methods and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, such techniques, methods and devices should be considered part of the description.

It should be noted that like numerals and letters denote like items in the following figures, therefore, once an item is defined in one figure, it does not require further discussion in subsequent figures.

In addition, the technical solutions of the various embodiments of the present application can be combined with each other, but it must be based on the realization of those skilled in the art. When the combination of technical solutions is contradictory or cannot be realized, it should be considered as a combination of technical solutions. Does not exist, nor is it within the scope of protection required by this application.

It should be noted that all directional indications (such as up, down, left, right, front, back...) in the embodiments of the present application are only used to explain If the specific posture changes, the directional indication will also change accordingly.

A method for performing service processing using a learning model according to an exemplary embodiment of the present application will be described below with reference to FIGS. 1-2 . It should be noted that the following application scenarios are only shown for easy understanding of the spirit and principle of the present application, and the implementation manners of the present application are not limited in this regard. On the contrary, the embodiments of the present application can be applied to any applicable scene.

The present application also proposes a method and device for business processing using a learning model.

Fig. 1 schematically shows a schematic flowchart of a method for using a learning model to process business according to an embodiment of the present application. As shown in Figure 1, the method includes:

S101. Obtain a communication network architecture composed of a first intelligent layer, a second intelligent layer, and a third intelligent layer, where the first intelligent layer and the second intelligent layer are deployed in base station devices, and the third intelligent layer is deployed in edge nodes.

As shown in FIG. 2 , it is a system schematic diagram of a communication network architecture proposed in this application. It includes the first intelligent layer deployed in the distributed unit DU of the base station equipment, the second intelligent layer deployed in the centralized unit CU of the base station equipment, and the third intelligent layer deployed in the edge node. In one manner, the fourth intelligent layer deployed in the cloud server is also included.

Among them, the fourth intelligent layer is a high-level management intelligent component, which is responsible for the management between various sub-networks. The third intelligent layer is the network intelligent orchestration component above the base station, which is responsible for the functional orchestration management between each base station. The second intelligent layer is a centralized intelligent component inside the base station, which is responsible for intelligent enhancement and realization of traditional wireless resource management (RRM). The first intelligent layer is a distributed intelligent component inside the base station, which is responsible for further optimizing the parameters with short scheduling cycles.

Further, the first intelligent layer and the second intelligent layer mentioned in this application can be deployed in various ways, for example, the first intelligent layer can be deployed in the distributed unit DU of the base station equipment, and the second intelligent layer can be deployed in In the centralized unit CU of the base station equipment.

In another manner, the first intelligent layer may be deployed on small base station equipment. And the second intelligent layer can be deployed on the macro base station equipment.

In related technologies, the process of a traditional distributed machine learning model usually includes steps:

1. The central (centralized) server collects various distributed scattered data;

2. After the convergence, the central server will distribute learning tasks (and training data) to each distributed node;

3. Each distributed node receives the assigned learning tasks (and training data) and starts learning;

4. After the learning of each distributed node is completed, the learning results will be returned to the central server;

5. The central server merges the learning results of each node;

6. Repeat processes 3 to 5 until the combined learning results reach the preset training conditions, wherein the preset conditions include one of training until the model converges, the number of training times reaching the maximum number of iterations, and the training duration reaching the maximum training time.

However, the traditional distributed machine learning model in related technologies does not consider the huge transmission pressure brought by a large amount of data transmission on the wireless link, and it does not consider the data privacy protection brought by the direct data transmission of distributed nodes. question. Therefore, this application can use the communication network architecture constructed by multiple intelligent layers to realize the aggregation of the model parameters uploaded by each client device node, and obtain the aggregated model parameters, so that the subsequent use of the aggregated model parameters for the initial learning model Perform hierarchical learning training to obtain a hierarchical federated learning model for service processing on the user equipment side or base station side.

It should be noted that the edge node in the embodiment of the present application may be an edge server, or may be an edge device such as an edge network element.

S102, using a hierarchical federated learning algorithm and a communication network architecture to generate a hierarchical federated learning model deployed on a user equipment side or a base station equipment side.

Among them, the communication network architecture proposed in this application may have the function of aggregating parameters, so that the hierarchical federated learning algorithm is used to aggregate the model parameters uploaded by each node in the network architecture and send them to the user equipment. The user equipment uses the aggregated model parameters to train the initial learning model deployed on itself, and then obtains a hierarchical federated learning model. In order to enable subsequent business processing through this model.

S103, use the layered federated learning model to perform business processing.

It should be noted that in this application, the user equipment may use the hierarchical federated learning model to perform the first service processing; and/or, the base station device may use the hierarchical federated learning model to perform the second service processing.

The first business processing may include driving route planning, face recognition, keyboard input prediction, and so on. It can be understood that in this way, the trained hierarchical federated learning model is handed over to the user for business processing.

Optionally, the second service processing may include traditional RRM AI enhanced services performed by the base station, such as mobility management, load balancing, dynamic resource allocation, interference coordination, MAC real-time scheduling, beam management and so on. Among them, the purpose of RRM is to improve the utilization rate of radio resources and meet the demand of mobile services for radio resources. It can be understood that in this way, the trained hierarchical federated learning model is handed over to the base station for business processing.

Obtain the fourth intelligent layer deployed in the cloud server;

The first intelligent layer trains the initial learning model according to the second aggregation model parameters to obtain the hierarchical federated learning model.

Further, in this application, the first intelligent layer deployed in the distributed unit DU of the base station equipment and the second intelligent layer deployed in the centralized unit CU of the base station equipment can be obtained; in this way, the base station equipment can There are one or more CUs and one or more DUs. Among them, one CU can be connected to one or more DUs. or,

Obtain the first intelligent layer deployed on small base station equipment, and the second intelligent layer deployed on macro base station equipment.

Among them, the small base station equipment SBS (small base station) is a base station with a small signal transmission coverage radius and suitable for small-scale accurate coverage. It can provide users with high-speed data services. As for the macro base station equipment (MBS, macro base station), it is a base station that has a wide communication coverage, but the capacity that a single user can share is small, and can only provide low-speed data services and communication services.

In this way, either MSB or SBS includes one or more CUs and DUs. In addition, one MBS can manage one or more SBSs.

In one approach, the embodiment of the present application is illustrated by taking the communication network architecture including three intelligent layers as an example:

Step 1: Among them, the first intelligent layer obtains the user equipment and uses local data to perform model training and learning, so as to generate initial model parameters. It should be noted that, in the embodiment of the present application, the high-level aggregator and the low-level aggregator can be defined first before this And build a digital twin network.

Step 2: The first intelligent layer uploads the update of the initial model parameters to the second intelligent layer, and the second intelligent layer updates all the received model parameters based on the aggregation criteria for first-level aggregation to obtain the first aggregated model parameters. Wherein, the aggregation criterion includes: a layered federated average algorithm, etc., which can be used for aggregation algorithms or criteria.

Step 3: The second intelligent layer sends the first aggregated model parameters after aggregation to the first intelligent layer that manages the connection, and completes a process of low-level federated learning.

Step 4: Repeat the above steps to the first number until it is determined that the aggregation of the first level is completed, the second intelligent layer uploads the parameters of the first aggregation model after aggregation to the third intelligent layer, and the third intelligent layer will receive all the parameters of the first aggregation model An aggregation model parameter update performs second-level aggregation based on aggregation criteria to obtain second aggregation model parameters. Similarly, the aggregation criterion includes: a layered federated average algorithm, etc., which can be used for aggregation algorithms or criteria.

Step 5: The third intelligent layer sends the aggregated parameters of the second aggregation model to the second intelligent layer of its management connection, and the second intelligent layer sends the aggregated parameters of the second aggregation model to the first layer of its management connection. The intelligent layer, and selectively sends the aggregated model parameters to the user equipment. Complete a process of high-level federated learning.

Step 6: After receiving the aggregated model parameters, the user equipment uses the aggregated model parameters to perform initial learning model training until it is determined that the trained business network model meets the preset conditions, and then determines to generate a hierarchical federated learning model , where the preset conditions include one of training until the model converges, the number of training times reaching the maximum number of iterations, and the training duration reaching the maximum training time.

In an optional manner, the embodiment of the present application uses a case where the communication network architecture includes four intelligent layers, and the user equipment uses a hierarchical federated learning model to perform the first service processing as an example:

Step 4: Repeat the above steps to the first number until it is determined that the aggregation of the first level is completed, and the second intelligent layer uploads the first aggregation model parameters after aggregation to the third intelligent layer, and the third intelligent layer will receive all first-level The aggregation model parameter update performs second-level aggregation based on the aggregation criterion to obtain second aggregation model parameters. Similarly, the aggregation criterion includes: a layered federated average algorithm, etc., which can be used for aggregation algorithms or criteria.

Step 5: The third intelligent layer sends the aggregated parameters of the second aggregation model to the second intelligent layer of its management connection, and the second intelligent layer sends the aggregated parameters of the second aggregation model to the first layer of its management connection. intelligence layer, and subsequently to the user device. Complete a process of high-level federated learning.

Step 6: Repeat the above steps to the second number of times, the third intelligent layer uploads the aggregated second aggregation model parameters to the fourth intelligent layer, and the fourth intelligent layer performs third-level aggregation on the second aggregation model parameters to obtain the first Three aggregation model parameters.

Step 7: The fourth intelligent layer delivers the aggregated third aggregation model parameters to the first intelligent layer level by level; and, the first intelligent layer selectively sends the third aggregation model parameters to the user equipment.

Step 8: After receiving the aggregated model parameters, the user equipment uses the aggregated model parameters to perform initial learning model training until it is determined that the trained service network model meets the preset conditions, and then determines to generate a hierarchical federated learning model , where the preset conditions include one of training to model convergence, the number of training times reaching the maximum number of iterations, and the training duration reaching the maximum training time. It should be noted that the model training and subsequent reasoning process can be performed locally or on The digital twin network is performed in vivo.

In another optional manner, the embodiment of the present application uses a case where the communication network architecture includes four intelligent layers, and the base station equipment uses a hierarchical federated learning model to perform the second service processing as an example:

Step 1: Wherein, the first intelligent layer (i.e., the base station equipment) uses local data for model training and learning to generate initial model parameters. It should be noted that in the embodiment of the present application, the high-level aggregator and the low-level aggregator can be first defined before this. Aggregators and building digital twin networks.

Step 5: The third intelligent layer sends the aggregated parameters of the second aggregation model to the second intelligent layer of its management connection, and the second intelligent layer sends the aggregated parameters of the second aggregation model to the first layer of its management connection. smart layer. Complete a process of high-level federated learning.

Step 7: The fourth intelligent layer sends the aggregated parameters of the third aggregation model to the first intelligent layer level by level.

Step 8: After receiving the aggregated model parameters, the first intelligent layer uses the aggregated model parameters to perform initial learning model training until it is determined that the trained business network model meets the preset conditions, and then determines to generate a hierarchical federation Learning model, where the preset conditions include one of the training until the model converges, the number of training times reaches the maximum number of iterations, and the training time reaches the maximum training time. It should be noted that the model training and subsequent reasoning process can be performed locally or Can be done within the digital twin network body.

Optionally, in another implementation manner of the present application, as shown in FIG. 3 , the present application further provides an apparatus for performing service processing by using a learning model. Among them, including:

In another embodiment of the present application, the acquiring module 201 further includes:

The obtaining module 201 is configured to obtain the fourth intelligent layer deployed in the cloud server;

The acquisition module 201 is configured to perform functional configuration on the fourth intelligent layer according to a preset configuration policy;

The acquiring module 201 is configured to, when it is detected that the function configuration of the fourth intelligent layer is completed, determine to generate the first intelligent layer, the second intelligent layer, the third intelligent layer and the fourth intelligent layer Composed of communication network architecture.

The obtaining module 201 is configured to use each of the first intelligent layers to obtain initial model parameters, where the initial model parameters are model parameters obtained by the user equipment or the first intelligent layer using local data for model training;

The acquisition module 201 is configured to use the second intelligent layer to receive the initial model parameters transmitted by each of the first intelligent layers, and perform first-level aggregation on the initial model parameters to obtain first aggregated model parameters;

The obtaining module 201 is configured so that the second intelligent layer sends the first aggregation model parameter to the first intelligent layer, and determines that the aggregation of the first layer is completed after reaching the first number.

The obtaining module 201 is configured such that the first intelligent layer sends the first aggregated model parameters to the user equipment, so that the user equipment trains an initial learning model according to the first aggregated model parameters; or,

The obtaining module 201 is configured such that the first intelligent layer trains an initial learning model according to the first aggregation model parameters.

The acquisition module 201 is configured to send the first aggregation model parameters to the third intelligent layer by each of the second intelligent layers;

The acquisition module 201 is configured to perform second-level aggregation on each of the first aggregation model parameters by the third intelligent layer to obtain second aggregation model parameters;

The obtaining module 201 is configured so that the third intelligent layer sends the second aggregation model parameters to the second intelligent layer, and determines that the second layer aggregation is completed after reaching a second number of times.

The acquisition module 201 is configured to send the second aggregation model parameters to the first intelligent layer by the second intelligent layer; and,

The obtaining module 201 is configured such that the first intelligent layer sends the second aggregation model parameters to the user equipment, so that the user equipment trains an initial learning model according to the second aggregation model parameters to obtain the A hierarchical federated learning model; or,

The acquisition module 201 is configured such that the first intelligent layer trains an initial learning model according to the second aggregation model parameters to obtain the hierarchical federated learning model.

The obtaining module 201 is configured to, if it is determined that there is a fourth intelligent layer, each of the third intelligent layers sends the second aggregation model parameters to the fourth intelligent layer, so that the fourth intelligent layer can After the second aggregation model parameters are aggregated at the third level, the third aggregation model parameters are obtained;

The obtaining module 201 is configured such that the fourth intelligent layer sends the third aggregation model parameters to the first intelligent layer step by step, so that the first intelligent layer The initial learning model is trained; or,

The obtaining module 201 is configured such that the first intelligent layer sends the third aggregated model parameters to the user equipment, so that the user equipment trains an initial learning model according to the third aggregated model parameters to obtain the described Hierarchical federated learning model.

The obtaining module 201 is configured to use the hierarchical federated learning model to perform the first service processing by the user equipment; or,

The obtaining module 201 is configured to use the hierarchical federated learning model to perform second service processing by the base station device.

The acquiring module 201 is configured such that the first intelligent layer is deployed in the distributed unit DU of the base station device, and the second intelligent layer is deployed in the centralized unit CU of the base station device; or,

The obtaining module 201 is configured such that the first intelligent layer is deployed in small base station equipment, and the second intelligent layer is deployed in macro base station equipment.

Fig. 4 is a logical structural block diagram of an electronic device according to an exemplary embodiment. For example, the electronic device 300 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, and the like.

In an exemplary embodiment, there is also provided a non-transitory computer-readable storage medium including instructions, such as a memory including instructions, the instructions can be executed by the electronic device processor to complete the above-mentioned method for business processing using a learning model, The method includes: obtaining a communication network architecture composed of a first intelligent layer, a second intelligent layer, and a third intelligent layer, wherein the first intelligent layer and the second intelligent layer are deployed in base station equipment, and the third intelligent layer is deployed in In the edge node: using the layered federated learning algorithm and the communication network architecture to generate a layered federated learning model deployed on the user equipment side or the base station device side; using the layered federated learning model to perform business processing. Optionally, the above instructions may also be executed by a processor of the electronic device to complete other steps involved in the above exemplary embodiments. For example, the non-transitory computer readable storage medium may be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, optical data storage device, and the like.

In an exemplary embodiment, an application program/computer program product is also provided, including one or more instructions, which can be executed by a processor of an electronic device to complete the above-mentioned business processing using a learning model The method includes: obtaining a communication network architecture composed of a first intelligent layer, a second intelligent layer, and a third intelligent layer, wherein the first intelligent layer and the second intelligent layer are deployed in base station equipment, and the third intelligent layer The layer is deployed in the edge node; using the layered federated learning algorithm and the communication network architecture to generate a layered federated learning model deployed on the user equipment side or the base station equipment side; using the layered federated learning model for business processing. Optionally, the above instructions may also be executed by a processor of the electronic device to complete other steps involved in the above exemplary embodiments.

FIG. 4 is an example diagram of a computer device 30 . Those skilled in the art can understand that the schematic diagram 4 is only an example of the computer device 30, and does not constitute a limitation to the computer device 30, and may include more or less components than those shown in the figure, or combine certain components, or different components For example, the computer device 30 may also include an input and output device, a network access device, a bus, and the like.

The so-called processor 302 may be a central processing unit (Central Processing Unit, CPU), and may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), Field-Programmable Gate Array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor 302 can also be any conventional processor, etc. The processor 302 is the control center of the computer device 30 and uses various interfaces and lines to connect various parts of the entire computer device 30 .

The memory 301 can be used to store computer-readable instructions 303 , and the processor 302 implements various functions of the computer device 30 by running or executing computer-readable instructions or modules stored in the memory 301 and calling data stored in the memory 301 . The memory 301 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playback function, an image playback function, etc.); Data created using the computer device 30 and the like. In addition, the memory 301 may include a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash memory card (Flash Card), at least one magnetic disk storage device, a flash memory device, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), or other non-volatile/volatile storage devices.

If the integrated modules of the computer device 30 are realized in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the methods of the above embodiments, and can also use computer-readable instructions to instruct related hardware to complete computer-readable instructions that can be stored in a computer-readable storage medium. When the readable instructions are executed by the processor, the steps in the above-mentioned various method embodiments can be realized.

Other embodiments of the present application will be readily apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the application, these modifications, uses or adaptations follow the general principles of the application and include common knowledge or conventional technical means in the technical field not disclosed in the application . The specification and examples are to be considered exemplary only, with a true scope and spirit of the application indicated by the following claims.

It should be understood that the present application is not limited to the precise constructions which have been described above and shown in the accompanying drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims

A method for business processing using a learning model, wherein:

Obtain a communication network architecture composed of a first intelligent layer, a second intelligent layer, and a third intelligent layer, wherein the first intelligent layer and the second intelligent layer are deployed in base station equipment, and the third intelligent layer is deployed in edge nodes;

Using a hierarchical federated learning algorithm and the communication network architecture to generate a hierarchical federated learning model deployed on the user equipment side or the base station equipment side;

Business processing is performed by using the layered federated learning model.
The method according to claim 1, further comprising:

Obtain the fourth intelligent layer deployed in the cloud server;

performing functional configuration on the fourth intelligent layer according to a preset configuration strategy;

After detecting that the function configuration of the fourth intelligent layer is completed, it is determined to generate a communication network architecture composed of the first intelligent layer, the second intelligent layer, the third intelligent layer, and the fourth intelligent layer.
The method according to claim 1 or 2, characterized in that, using a hierarchical federated learning algorithm and the communication network architecture to generate a hierarchical federated learning model deployed on the user equipment side or the base station equipment side includes:

Using each of the first intelligent layers to acquire initial model parameters, where the initial model parameters are model parameters obtained by the user equipment or the first intelligent layer using local data for model training;

After using the second intelligent layer to receive the initial model parameters transmitted by each of the first intelligent layers, perform first-level aggregation on the initial model parameters to obtain first aggregated model parameters;

The second intelligent layer sends the first aggregation model parameters to the first intelligent layer, and determines that the aggregation at the first level is completed after reaching the first number.
The method according to claim 3, characterized in that, after determining that the first-level aggregation is completed until the first number is reached, comprising:

The first intelligent layer sends the first aggregated model parameters to the user equipment, so that the user equipment trains an initial learning model according to the first aggregated model parameters; or,

The first intelligent layer trains an initial learning model according to the first aggregated model parameters.
The method according to claim 3, after said determining that said first-level aggregation is completed, comprising:

sending the first aggregation model parameters to the third intelligent layer by each of the second intelligent layers;

performing second-level aggregation on each of the first aggregation model parameters by the third intelligent layer to obtain second aggregation model parameters;

The third intelligent layer sends the second aggregation model parameters to the second intelligent layer, and determines that the aggregation at the second level is completed after reaching a second number of times.
The method according to claim 4, wherein after the second number of times is reached and it is determined that the aggregation of the second level is completed, it comprises:

the second intelligence layer sends the second aggregated model parameters to the first intelligence layer; and,

The first intelligent layer sends the second aggregation model parameters to the user equipment, so that the user equipment trains the initial learning model according to the second aggregation model parameters to obtain the hierarchical federated learning model; or ,

The first intelligent layer trains an initial learning model according to the second aggregation model parameters to obtain the hierarchical federated learning model.
The method according to claim 5, characterized in that, after the third intelligent layer sends the second aggregation model parameters to the second intelligent layer, comprising:

If it is determined that there is a fourth intelligent layer, each of the third intelligent layers sends the second aggregation model parameters to the fourth intelligent layer, so that the fourth intelligent layer performs a process on the second aggregation model parameters. After the third level of aggregation, the third aggregation model parameters are obtained;

The fourth intelligent layer sends the third aggregation model parameters to the first intelligent layer step by step, so that the first intelligent layer trains the initial learning model according to the first aggregation model parameters; or ,

The first intelligent layer sends the third aggregated model parameters to the user equipment, so that the user equipment trains an initial learning model according to the third aggregated model parameters to obtain the hierarchical federated learning model.
The method according to claim 1, wherein said utilizing said hierarchical federated learning model for business processing comprises:

The user equipment uses the hierarchical federated learning model to perform the first service processing; or,

The base station device uses the layered federated learning model to perform the second service processing.
The method according to claim 1, wherein the first intelligent layer and the second intelligent layer are deployed in base station equipment, comprising:

The first intelligent layer is deployed in a distributed unit DU of the base station device, and the second intelligent layer is deployed in a centralized unit CU of the base station device; or,

The first intelligent layer is deployed in small base station equipment, and the second intelligent layer is deployed in macro base station equipment.
A device for business processing using a learning model, characterized in that it is applied to base station equipment, including:

An acquisition module configured to acquire a communication network architecture composed of a first intelligent layer, a second intelligent layer, and a third intelligent layer, wherein the first intelligent layer and the second intelligent layer are deployed in base station equipment, and the third intelligent layer Deployed in edge nodes;

The generating module is configured to generate a hierarchical federated learning model deployed on the user equipment side or the base station equipment side by using the hierarchical federated learning algorithm and the communication network architecture;

The processing module is configured to use the hierarchical federated learning model to perform business processing.
An electronic device, characterized in that it comprises:

memory for storing executable instructions; and,

A processor, configured to execute the executable instructions with the memory so as to complete the operations of the method for business processing using a learning model in any one of claims 1-9.
A computer-readable storage medium for storing computer-readable instructions, wherein when the instructions are executed, the operations of the method for business processing by using a learning model in any one of claims 1-9 are performed.