CN114692898A - MEC federated learning method, device and computer-readable storage medium - Google Patents

Abstract

Embodiments of the present application provide an MEC federated learning method, device, and computer-readable storage medium. Including: when any multi-access edge computing MEC system in any federated learning group sends a federated learning request, based on the resource capacity utilization of each MEC system in the federated learning group, determine the MEC system coordinator from each MEC system; Receive model information sent by other MEC systems in the federated learning group through the MEC system coordinator, and obtain the resource capacity utilization of the MEC system coordinator in real time during the receiving process; The coordinator switches. After the MEC system coordinator aggregates the model information, the aggregated model information is sent to each MEC system in the federated learning group. This scheme realizes the federated learning of MEC, ensures the performance of the coordinator of the MEC system, and realizes the balanced use of resources and capacity of the MEC system within the federated learning group of the MEC system.

Description

MEC federated learning method, device and computer-readable storage medium

technical field

The present application relates to the technical field of 5G edge computing, and in particular, the present application relates to an MEC federated learning method, apparatus, and computer-readable storage medium.

Background technique

MEC (Multi-access Edge Computing) is a natural product of the evolution of mobile base stations and the convergence of IT and telecom networks. By deploying various services and caching content at the edge of the network, congestion in the mobile core network can be further eased , and can efficiently serve the local area. MEC provides a new ecosystem and value chain where operators can open up their Radio Access Network (RAN) edge to authenticated and certified third parties, enabling them to flexibly and rapidly expand their reach to mobile users, enterprises and verticals Industry deployments to provide innovative applications and services such as video analytics, location services, enhanced displays, local content distribution and data caching.

Based on the problems of data fragmentation, data loneliness, leakage of user privacy, and data shortage faced by machine learning, federated learning was born. Federated learning is a distributed machine learning framework, which can be used in user privacy protection, data security and government regulations. It allows multiple participants to protect their own data locally and privately, and at the same time to collaborate and securely establish a federated learning model. Its technology can effectively solve the problem of data silos and realize intelligent cooperation among participants.

Based on the real-time, agile, intelligent, security and other characteristics of MEC, mobile network operators, enterprises and vertical industries have joined the MEC queue and established MEC systems. is an essential requirement. The federation of the MEC system enables the shared use of MEC services and applications. Provide new functionality by collaborating with other services instead of developing all services, for example, speech recognition can be a key function of other services (such as navigation applications), in this case, the speech recognition service provider does not necessarily have to work with the navigation service The providers are the same, and each service can be deployed on different MEC systems in the MEC environment.

However, there is currently no mature MEC federated learning scheme, so it is urgent to provide a new MEC federated learning scheme.

SUMMARY OF THE INVENTION

The purpose of this application is to solve at least one of the above-mentioned technical defects, and the technical solutions provided by the embodiments of this application are as follows:

In the first aspect, an embodiment of the present application provides a federated learning method for MEC, including:

When any multi-access edge computing MEC system in any federated learning group issues a federated learning request, the MEC system coordinator is determined from each MEC system based on the resource capacity utilization of each MEC system in the federated learning group;

Receive model information sent by other MEC systems in the federated learning group through the MEC system coordinator, and obtain the resource capacity utilization of the MEC system coordinator in real time during the receiving process;

If the resource capacity utilization rate of the MEC system coordinator does not exceed the first preset threshold during the receiving process, the MEC system coordinator aggregates the model information, and then sends the aggregated model information to each of the federated learning groups. MEC system;

If the resource capacity utilization rate of the MEC system coordinator exceeds the first preset threshold during the receiving process, a new MEC system coordinator is determined from each MEC system and switched to the new MEC system coordinator to receive each model information, repeating The new MEC system coordinator is determined and switched until the resource capacity utilization rate of the new MEC system coordinator does not exceed the first preset threshold during the receiving process, then after the new MEC system coordinator aggregates the model information, the aggregated The latter model information is sent to each MEC system in the federated learning group.

In an optional embodiment of the present application, based on the resource capacity utilization rate of each MEC system in the federated learning group, the MEC system coordinator is determined from each MEC system, including:

Obtain the preset available MEC system coordinator set, and each MEC system in the preset available MEC system coordinator set carries the corresponding priority;

Among the MEC systems in the federated learning group that belong to the set of available MEC system coordinators, the resource capacity utilization rate does not exceed the second preset threshold and the MEC system with the highest priority is determined as the MEC system coordinator, and the second preset threshold value is not greater than the first preset threshold.

In an optional embodiment of the present application, the method further includes:

If the resource capacity utilization of all MEC systems belonging to the set of available MEC system coordinators in the federated learning group exceeds the second preset threshold, the resource capacity utilization of the MEC systems in the federated learning group that do not belong to the set of available MEC system coordinators The MEC system with the smallest rate is determined as the MEC system coordinator.

In an optional embodiment of the present application, acquiring a set of preset available MEC system coordinators includes:

Obtain multiple federated learning groups from a preset number of MEC systems;

The MEC systems that appear in any federated learning group are identified as elements of the set of available MEC coordinators, and the set of available MEC coordinators is constructed, and the priority of each MEC system in the set of MEC coordinators is repeated in each federated learning group. is proportional to the number of times.

In an optional embodiment of the present application, a plurality of federated learning groups are obtained from a preset number of MEC systems, including:

Identify MEC systems with the same data source among MEC systems as a federated learning group; and/or,

The MEC system whose model feature similarity in each MEC system meets the preset conditions is determined as a federated learning group.

In an optional embodiment of the present application, the model information sent by other MEC systems in the federated learning group is received by the MEC system coordinator, including:

Determine whether an existing MEC system coordinator is included in the federated learning group;

If not included, directly receive model information sent by other MEC systems in the federated learning group through the MEC system coordinator;

If it is included, the model information received by the existing MEC system coordinator is obtained through the MEC system coordinator, and the model information of other MEC systems is continuously received through the MEC system.

In an optional embodiment of the present application, a new MEC system coordinator is determined from each MEC system and switched to the new MEC system coordinator to receive each model information, including:

Based on the resource capacity utilization of each MEC system in the federated learning group, determine the new MEC system coordinator from each MEC system;

Obtain the model information received by the MEC system coordinator through the new MEC system coordinator, and continue to receive model information from other MEC systems through the new MEC system.

In a second aspect, an embodiment of the present application provides an MEC federated learning device, including:

The MEC system coordinator determines the module, which is used to obtain a federated learning request from any multi-access edge computing MEC system in any federated learning group based on the resource capacity utilization of each MEC system in the federated learning group. Determine the MEC system coordinator;

The model information receiving module is used to receive the model information sent by other MEC systems in the federated learning group through the MEC system coordinator, and obtain the resource capacity utilization rate of the MEC system coordinator in real time during the receiving process;

The first model information aggregation module is used for, if the resource capacity utilization rate of the MEC system coordinator does not exceed the first preset threshold during the receiving process, after the MEC system coordinator aggregates each model information, the aggregated model The information is sent to each MEC system in the federated learning group;

In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor;

A computer program is stored in the memory;

A processor, configured to execute a computer program to implement the method provided in the embodiment of the first aspect or any optional embodiment of the first aspect.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, any optional embodiment of the first aspect or the first aspect is implemented methods provided in the examples.

In a fifth aspect, embodiments of the present application provide a computer program product or computer program, where the computer program product or computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device implements the first aspect embodiment or any optional embodiment of the first aspect when executed. method.

The beneficial effects brought by the technical solution provided by the application are:

The MEC system coordinator is determined based on the resource capacity utilization of each MEC system in the group that initiates the federated learning request, and in the subsequent process of receiving model information of other MEC systems through the MEC system coordinator, the MEC system coordinator in the receiving process is obtained in real time. The resource capacity utilization rate is determined, and again based on the resource capacity utilization rate, it is determined whether to perform the handover of the MEC system coordinator. This solution realizes the MEC federated learning, and considers the resource capacity utilization in the process of confirming the MEC system coordinator, which ensures the performance of the MEC system coordinator, and realizes the balanced use of resource capacity of the MEC system within the MEC system federated learning group.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present application more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments of the present application.

FIG. 1 is a schematic flowchart of a federated learning method for MEC provided by an embodiment of the present application;

2 is a flowchart of detecting that the requested service is not in the MEC system in an example of the embodiment of the application;

FIG. 3 is a flow chart of federated learning when no handover occurs between the MEC system coordinator in an example of an embodiment of the present application;

FIG. 4 is a federated learning flow chart in which the coordinator of the handover is located in the set of available MEC system coordinators in an example of the embodiment of the application;

FIG. 5 is a federated learning flowchart in an example of an embodiment of the present application in which the coordinator of the handover is not located in the set of available MEC system coordinators;

FIG. 6 is a handover interaction diagram of the MEC system coordinator in an example of an embodiment of the application;

FIG. 7 is an interaction diagram of federated learning when no handover occurs between the MEC system coordinators in an example of the embodiment of the present application;

FIG. 8 is an interaction diagram of federated learning when the MEC system coordinator switches over in an example of an embodiment of the present application

FIG. 9 is a schematic structural diagram of an MEC federated learning apparatus provided by an embodiment of the present application;

FIG. 10 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed ways

The following describes in detail the embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals refer to the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, but not to be construed as a limitation on the present application.

It will be understood by those skilled in the art that the singular forms "a", "an", "the" and "the" as used herein can include the plural forms as well, unless expressly stated otherwise. It should be further understood that the word "comprising" used in the specification of this application refers to the presence of stated features, integers, steps, operations, elements and/or components, but does not preclude the presence or addition of one or more other features, Integers, steps, operations, elements, components and/or groups thereof. It will be understood that when we refer to an element as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Furthermore, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combination of one or more of the associated listed items.

In order to make the objectives, technical solutions and advantages of the present application clearer, the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

FIG. 1 is a schematic flowchart of an MEC federated learning method provided by an embodiment of the present application. As shown in FIG. 1 , the method may include:

Step S101, when any multi-access edge computing MEC system in any federated learning group issues a federated learning request, based on the number of times and resource capacity utilization of the MEC system in each federated learning group, determine from each MEC system. MEC system coordinator.

Specifically, when any MEC system in any federated learning group issues a federated learning request, the MEC federated learning method provided by the embodiment of the present application is triggered. Then, the resource capacity utilization rate of each MEC system in the federated learning group is obtained, and the MEC system serving as the MEC system coordinator is determined according to the resource capacity utilization rate of each MEC system. The MEC system coordinator will act as the executive body for aggregating the federated learning participants in the federated learning, wherein the federated learning participants are the MEC systems in the federated learning group.

It can be understood that, in the solution of this application, when selecting the MEC system coordinator from the federated learning group, in order to ensure the performance of the MEC system coordinator, the resource capacity utilization of each MEC system in the federated learning group needs to be considered.

It should be noted that the MEC system consists of the MEC host and the MEC management. The MEC host is an entity that includes the MEC platform and virtualization infrastructure, and provides computing, storage, and network resources for running MEC applications. The MEC platform is a collection of basic functions required to run MEC applications on a specific virtualized infrastructure and enable it to provide and use MEC services, which can also provide services. The MEC application is instantiated on the virtualized infrastructure of the MEC host according to the configuration or request verified by the MEC management layer. MEC management includes MEC system-level management and MEC host-level management. MEC system-level management includes the multi-access edge orchestrator as its core component, which provides an overview of the entire MEC system. MEC host-level management includes the MEC platform manager and the virtualization infrastructure manager, which are responsible for managing the MEC-specific functions of a specific MEC host and the applications running on it. The Mp1 interface is the interface between the MEC platform and the MEC application program, the Mm5 interface is the interface between the MEC platform manager and the MEC platform, and the Mm3 interface is the interface between the MEC orchestrator and the MEC platform manager.

When the instantiated MEC application in a certain MEC system in the federated learning group initiates a service request, it is detected that the service does not exist in the MEC system, as shown in Figure 2. The specific detection process is as follows:

1. The service consumer (ie the MEC application program instantiated in the MEC system) uses its ID to send a request to the MEC platform through the Mp1 reference point to request the required service.

2. The corresponding MEC platform in the MEC system finds that the requested service is not available locally.

3. Forward the service request to the MEC platform manager of the MEC system through the Mm5 interface.

4. The MEC platform manager sequentially forwards the service request to the MEC orchestrator through the Mm3 interface.

5. The MEC orchestrator detects the topology and available services of the MEC system and finds that the requested service is not available in the MEC system.

Step S102, the model information sent by other MEC systems in the federated learning group is received by the MEC system coordinator, and the resource capacity utilization rate of the MEC system coordinator is acquired in real time during the receiving process.

Specifically, the MEC system coordinator determined in the previous step receives the model information sent by other MEC systems in the federated learning group. After the model information of each MEC system is received, the model information can be aggregated to obtain an aggregated All model information, complete federated learning. However, in order to further ensure the performance of the MEC system coordinator, it is necessary to monitor its resource capacity utilization in real time during the process of receiving its data model information, and deactivate the current MEC system when the resource capacity utilization exceeds the first preset threshold coordinator, and determine the new MEC system coordinator, and switch from the current MEC system coordinator to the new MEC system coordinator. The subsequent steps of the handover of the MEC system coordinator will be described in detail.

Step S103, if the resource capacity utilization rate of the MEC system coordinator does not exceed the first preset threshold during the receiving process, after the MEC system coordinator aggregates the model information, the aggregated model information is respectively sent to federated learning. Each MEC system in the group.

In one case, the resource capacity utilization rate of the current MEC system coordinator does not exceed the first preset threshold, that is, after the current MEC system coordinator receives all model information, its resource capacity utilization rate is not greater than the first preset threshold. By setting the threshold, its performance can be guaranteed. It is possible to further aggregate the model information through the MEC system coordinator, and then send the aggregated model information to each MEC system in the federated learning group. After receiving the aggregated model information, each MEC system updates the local model information. .

It should be noted that the first preset threshold can be set according to experience and actual needs, for example, set to 60%.

Step S104, if the resource capacity utilization rate of the MEC system coordinator exceeds the first preset threshold during the receiving process, then determine a new MEC system coordinator from each MEC system and switch to receive each MEC system coordinator through the new MEC system coordinator. Model information, repeat the determination and handover of the new MEC system coordinator until the resource capacity utilization rate of the new MEC system coordinator does not exceed the first preset threshold during the receiving process, then aggregate the model information through the new MEC system coordinator Then, the aggregated model information is sent to each MEC system in the federated learning group.

In another situation, the resource capacity utilization rate of the current MEC system coordinator exceeds the first preset threshold, that is, after the current MEC system coordinator receives all model information, its resource capacity utilization rate is greater than the first preset threshold threshold, its performance may not be guaranteed. Therefore, a new MEC coordinator needs to be re-determined. After the new MEC system coordinator aggregates the model information, the aggregated model information is sent to each MEC system in the federated learning group.

Specifically, the method of re-determining the new MEC system coordinator can be the same as the method in step S101, and also considers the resource capacity utilization of each MEC system in the federated learning group. This process is repeated, that is, after each time a new MEC system coordinator is determined, switch from the previous MEC system coordinator to the new MEC system coordinator, receive model information through the new MEC system coordinator, and the same docking process The resource capacity utilization rate of the new MEC system coordinator is monitored in real time. If it still exceeds the first preset threshold, the new MEC system coordinator is determined again, and the resource capacity utilization rate in the handover and receiving process is monitored. Repeat the above steps of "determining a new MEC system coordinator-handover-receiving process monitoring" several times, until the new MEC system coordinator finally determined has received all model information, and the resource capacity utilization rate does not exceed the first pre-set. Set the threshold.

In the solution provided by this application, the MEC system coordinator is determined based on the resource capacity utilization of each MEC system in the group that initiates the federated learning request, and in the subsequent process of receiving the model information of other MEC systems through the MEC system coordinator, real-time acquisition and reception In the process, the resource capacity utilization rate of the MEC system coordinator is determined, and again based on the resource capacity utilization rate, it is determined whether to switch the MEC system coordinator. This solution realizes the MEC federated learning, and considers the resource capacity utilization in the process of confirming the MEC system coordinator, which ensures the performance of the MEC system coordinator, and realizes the balanced use of resource capacity of the MEC system within the MEC system federated learning group.

In an optional embodiment of the present application, based on the resource capacity utilization rate of each MEC system in the federated learning group, the MEC system coordinator is determined from each MEC system, including:

Obtain the preset available MEC system coordinator set, and each MEC system in the preset available MEC system coordinator set carries the corresponding priority;

Among the MEC systems in the federated learning group that belong to the set of available MEC system coordinators, the resource capacity utilization rate does not exceed the second preset threshold and the MEC system with the highest priority is determined as the MEC system coordinator, and the second preset threshold value is not greater than the first preset threshold.

Among them, the set of preset available MEC system coordinators is obtained, including:

Obtain multiple federated learning groups from a preset number of MEC systems;

The MEC systems that appear in any federated learning group are identified as elements of the set of available MEC coordinators, and the set of available MEC coordinators is constructed, and the priority of each MEC system in the set of MEC coordinators is repeated in each federated learning group. is proportional to the number of times.

Further, from a preset number of MEC systems, multiple federated learning groups are obtained, including:

Identify MEC systems with the same data source among MEC systems as a federated learning group; and/or,

The MEC system whose model feature similarity in each MEC system meets the preset conditions is determined as a federated learning group.

Specifically, first, assuming that there are m MEC systems, p MEC system federated learning groups are formed based on the same data source of the business model or the method based on the feature similarity of the business model.

Among them, the same source of samples means that different MEC systems may have different model samples, and the data of different AI model samples come from the same physical device. For example, one MEC system has an image recognition model, and another MEC system has a speech recognition model. Although the models in different MEC systems are different, the data of these two models come from the same physical terminal device. Model feature similarity means that each MEC system has different models, but all models have the same feature attributes. For example, different MEC systems have image recognition models from different visual terminals, although the collected image/video samples The data sources are different, but the model samples have the same feature attributes.

Then, count the number N of repetitions of each MEC system in the p MEC system federated learning groups in the p MEC system federated learning groups, then calculate the weight of each MEC system in the p MEC system groups as R= N/p, the weight value represents the importance of the MEC system in p MEC system groups. The larger the weight value is, the more model information the MEC system has, and the federated learning model can be used for training to improve the efficiency of federated learning.

Finally, based on the weight value R of the MEC system calculated in the above steps, sort the weight value R from large to small, and select the corresponding MEC system whose weight value R is greater than 0. After the weight value R is sorted, the corresponding MEC system set For {MEC ₁ , MEC ₂ ......, MEC _n ; n<m}, the MEC system set is defined as the set of available MEC system coordinators, and the priority of the MEC systems included in it is also determined. The higher the priority of the MEC system.

As can be seen from the foregoing description, the MEC federated learning scheme of the present application can be divided into two situations: one is that the MEC system coordinator does not switch after the MEC system coordinator is determined. The other is that the MEC system coordinator switches over after the MEC system coordinator is determined. In this case, it also includes: if the resource capacity utilization of all the MEC systems belonging to the set of available MEC system coordinators in the federated learning group exceeds the second With the preset threshold, the MEC system with the smallest resource capacity utilization among the MEC systems in the federated learning group that does not belong to the set of available MEC system coordinators is determined as the MEC system coordinator. Next, the two cases will be described in detail respectively.

As shown in Figure 3, the solution for the MEC system coordinator without handover (no handover) may include the following steps:

1. The MEC system in the MEC federated learning group initiates a federated learning request.

2. Set the first and second preset thresholds of resource capacity utilization of the MEC system as C1 and C2, respectively, and select the MEC system from the federated learning group according to the priority of the set of available MEC system coordinators.

3. Determine whether the resource capacity utilization rate of the currently selected MEC system exceeds the second preset threshold C2.

If the resource capacity utilization rate of the currently selected MEC system exceeds the second preset threshold C2, then go back to step 2 and select another MEC system from the federated learning group again according to the priority of the available MEC system coordinator set.

If the currently selected MEC system does not exceed the second preset threshold C2, it is determined that the currently selected MEC system is the MEC system coordinator.

4. Each MEC system participant encrypts and sends the local computing model information to the MEC system coordinator, and the model information includes model features, model parameters and other information.

5. In the process of sending model information, the resource capacity utilization rate of the MEC system coordinator is monitored in real time. The resource capacity utilization rate of the MEC system coordinator is always less than the first preset threshold C1, and the MEC system coordinator does not switch.

6. The MEC system coordinator aggregates the received model information, and then sends the aggregated model information to each MEC system participant.

7. Each MEC participant receives the aggregated model information and updates the local model information.

As shown in Figure 4, the solution in which the MEC system coordinator is handed over, and the new MEC system coordinator to which the handover is located is in the set of available MEC system coordinators, may include the following steps:

1. The MEC system in the MEC federated learning group initiates a federated learning request.

2. Set the first preset threshold value and the second preset threshold value of MEC system resource capacity utilization as C1 and C2 respectively, and select the MEC system from the federated learning group according to the priority of the set of available MEC system coordinators.

3. Determine whether the resource capacity utilization rate of the currently selected MEC system exceeds the second preset threshold C2.

If the resource capacity utilization rate of the currently selected MEC system exceeds the second preset threshold C2, go back to step 2 and select the MEC system again from the federated learning group according to the priority of the available MEC system coordinator set.

If the resource capacity utilization rate of the currently selected MEC system does not exceed the second preset threshold C2, it is determined whether an existing MEC system coordinator is included in the federated learning group of the MEC system.

Further, if the MEC system federated learning group does not include an existing MEC system coordinator, the selected MEC system is determined to be the MEC system coordinator.

If the MEC system federated learning group includes an existing MEC system coordinator, the selected MEC system is determined as the MEC system handover coordinator, and the existing MEC system coordinator is switched to the determined MEC system coordinator

4. Each MEC system participant encrypts and sends the local computing model information to the MEC system coordinator, and the model information includes model features, model parameters and other information.

5. In the process of model information transmission, real-time monitoring of resource capacity utilization of the MEC system coordinator.

6. Determine whether the resource capacity utilization rate of the MEC system coordinator exceeds the first preset threshold C1.

If the resource capacity utilization rate of the MEC system coordinator exceeds the first preset threshold value C1, return to step 2 for execution.

If the resource capacity utilization rate of the MEC system coordinator does not exceed the first preset threshold C1, the MEC system coordinator aggregates the received model information, and then sends the aggregated model information to each MEC system participant.

7. Each MEC participant receives the aggregated model information and updates the local model information.

As shown in Figure 5, the MEC system coordinator is switched, and the new MEC system coordinator is not located in the set of available MEC system coordinators, which may include the following steps:

1. The MEC system in the MEC federated learning group initiates a federated learning request.

2. Set the first preset threshold value and the second preset threshold value of MEC system resource capacity utilization as C1 and C2 respectively, and select the MEC system from the federated learning group according to the priority of the available MEC system coordinators.

3. Determine whether the resource capacity utilization rate of the currently selected MEC system exceeds the second preset rate threshold C2.

If the resource capacity utilization rate of the currently selected MEC system exceeds the second preset threshold C2, it is determined whether the resource capacity utilization rates of the MEC systems belonging to the preset available MEC coordinators in the federated learning group are all greater than C2.

If the resource capacity utilization rate of the above-mentioned MEC systems in the federated learning group is all greater than C2, select the MEC system with the minimum resource capacity utilization rate from the coordination set of available MEC systems, and then execute step 4 to determine whether the federated learning group contains the MEC system with the minimum resource capacity utilization rate. There is a MEC system coordinator.

If the resource capacity utilization of the above-mentioned MEC systems in the federated learning group is not all greater than C2, go back to step 2 to execute, and select the MEC system again from the federated learning group according to the priority of the set of available MEC system coordinators.

4. If the resource capacity utilization rate of the currently selected MEC system does not exceed the second preset threshold C2, determine whether the federated learning group includes an existing MEC system coordinator.

If no existing MEC system coordinator is included in the federated learning group, the currently selected MEC system is determined as the MEC system coordinator.

If the federated learning group includes an existing MEC system coordinator, the selected MEC system is determined as the MEC system switching coordinator, and the existing MEC system coordinator switches to the currently selected MEC system coordinator.

5. Each MEC system participant encrypts and sends the local computing model information to the MEC system coordinator, and the model information includes model features, model parameters and other information.

6. In the process of model information transmission, monitor the resource capacity utilization of the MEC system coordinator in real time.

7. Determine whether the resource capacity utilization rate of the MEC system coordinator exceeds the first preset threshold C1.

If the resource capacity utilization rate of the MEC system coordinator exceeds the first preset threshold C1, return to step 2 for execution, and select the MEC system again from the federated learning group according to the priority of the set of available MEC system coordinators.

If the resource capacity utilization rate of the MEC system coordinator does not exceed the first preset threshold C1, the MEC system coordinator aggregates the received model information, and then sends the aggregated model information to each MEC system participant.

8. Each MEC participant receives the aggregated model information and updates the local model information.

In an optional embodiment of the present application, a new MEC system coordinator is determined from each MEC system and switched to receive each model information through the new MEC system coordinator, including:

Based on the resource capacity utilization of each MEC system in the federated learning group, determine the new MEC system coordinator from each MEC system;

Obtain the model information received by the MEC system coordinator through the new MEC system coordinator, and continue to receive model information from other MEC systems through the new MEC system.

Among them, when the MEC system in the federated learning group initiates a federated learning request, the MEC system coordinator 1 has been determined, and other participants in the federated learning group transmit the model to the MEC system coordinator 1. At this time, it is determined that the handover of the MEC system coordinator needs to be performed, as shown in Figure 6, the process may include the following steps:

1. During the model information transmission process, the resource capacity utilization rate of the MEC system coordinator 1 reaches the first preset threshold C1.

2. The MEC system coordinator 1 will notify each participant to stop sending model information.

3. The MEC system coordinator 1 determines the MEC system coordinator 2 (ie, the new MEC system coordinator) according to the method for determining the MEC system coordinator.

4. The MEC system coordinator 1 sends a handover request to the MEC system coordinator 2 .

5. The MEC system coordinator 2 sends a response to the MEC system coordinator 1, agreeing to the handover request.

6. The MEC system coordinator 1 sends the received model information and its own model information to the MEC system coordinator 2.

7. The MEC system coordinator 1 will notify each participant that the coordinator has changed at this time.

8. Each participant goes to the coordinator 2 of the MEC system handover, and continues to send the model.

The solution of the present application is further described below through two examples. As shown in FIG. 7 , the MEC system coordinator does not switch over, and the federated learning method of the MEC system may include:

1. The MEC application initiates a service request and detects that the service does not exist in the MEC system where the MEC is located, which will trigger federated learning.

2. In the established federated learning group, determine the MEC system coordinator 1 in the MEC system according to the method for determining the MEC system coordinator.

3. The MEC system coordinator 1 sends a federated learning request to each participant in the federated learning group.

4. The participants of each MEC system send model information to the MEC system coordinator 1, and the model information includes information such as model features and model parameters.

5. The MEC system coordinator 1 aggregates the model information sent by each participant.

6. The MEC system coordinator 1 sends the aggregated model information to each participant.

7. Each MEC system participant updates the local model information.

As shown in Figure 7, the coordinator of the MEC system is switched, and the federated learning method of the MEC system can include:

1. The MEC application initiates a service request and detects that the service does not exist in the MEC system where the MEC is located, which will trigger federated learning.

2. In the established federated learning group, determine the MEC system coordinator 1 in the MEC system based on determining the MEC system coordinator.

3. The MEC system coordinator 1 sends a federated learning request to each participant in the federated learning group.

4. Each MEC system participant sends model information to the MEC system coordinator 1, where the model information includes model type and model parameter information.

5. During the model information sending process, the handover of the MEC system coordinator occurs, and the MEC system coordinator 2 is determined according to the method 2 in determining the MEC system coordinator.

6. According to the switching method of the MEC system coordinator, the MEC system coordinator switches from the MEC system coordinator 1 to the MEC system coordinator 2.

7. Each MEC system participant sends model information to the MEC system coordinator 2, where the model information includes model type and model parameter information.

8. The MEC system coordinator 2 aggregates the received model information.

9. The MEC system coordinator 2 sends the aggregated model information to each MEC system participant.

10. Each MEC participant updates the local model information.

FIG. 9 is a schematic structural diagram of an MEC federated learning apparatus provided by an embodiment of the present application. As shown in FIG. 9 , the apparatus 900 may include: a MEC system coordinator determining module 901 , a model information receiving module 902 , and a first model information aggregation Module 903 and second model information aggregation module 904, wherein:

The MEC system coordinator determination module 901 is configured to, when any multi-access edge computing MEC system in any federated learning group sends a federated learning request, based on the resource capacity utilization rate of each MEC system in the federated learning group, from each MEC system Determine the MEC system coordinator;

The model information receiving module 902 is configured to receive model information sent by other MEC systems in the federated learning group through the MEC system coordinator, and obtain the resource capacity utilization rate of the MEC system coordinator in real time during the receiving process;

The first model information aggregation module 083 is configured to, if the resource capacity utilization rate of the MEC system coordinator does not exceed the preset threshold during the receiving process, after the MEC system coordinator aggregates each model information, the aggregated model information is divided into two parts. Sent to each MEC system in the federated learning group;

The second model information aggregation module 904 is configured to determine a new MEC system coordinator from each MEC system and switch to the new MEC system if the resource capacity utilization rate of the MEC system coordinator exceeds a preset threshold during the receiving process The coordinator receives the information of each model, and repeats the determination and switching of the new MEC system coordinator until the resource capacity utilization rate of the new MEC system coordinator does not exceed the preset threshold during the receiving process, then the new MEC system coordinator checks the information of each model through the new MEC system coordinator. After aggregation, the aggregated model information is sent to each MEC system in the federated learning group.

In the solution provided by this application, the MEC system coordinator is determined based on the resource capacity utilization of each MEC system in the group that initiates the federated learning request, and in the subsequent process of receiving the model information of other MEC systems through the MEC system coordinator, real-time acquisition and reception In the process, the resource capacity utilization rate of the MEC system coordinator is determined, and again based on the resource capacity utilization rate, it is determined whether to switch the MEC system coordinator. This solution realizes the MEC federated learning, and considers the resource capacity utilization in the process of confirming the MEC system coordinator, which ensures the performance of the MEC system coordinator, and realizes the balanced use of resource capacity of the MEC system within the MEC system federated learning group.

In an optional embodiment of the present application, the MEC system coordinator determination module is specifically used for:

Obtain the preset available MEC system coordinator set, and each MEC system in the preset available MEC system coordinator set carries the corresponding priority;

Among the MEC systems in the federated learning group that belong to the set of available MEC system coordinators, the MEC system whose resource capacity utilization does not exceed the preset threshold and has the highest priority is determined as the MEC system coordinator.

In an optional embodiment of the present application, in an optional embodiment of the present application, the MEC system coordinator determination module is further configured to:

If the resource capacity utilization of all MEC systems belonging to the set of available MEC system coordinators in the federated learning group exceeds the preset threshold, the resource capacity utilization rate of the MEC systems that do not belong to the set of available MEC system coordinators in the federated learning group will be the smallest. The MEC system is determined as the MEC system coordinator.

In an optional embodiment of the present application, in an optional embodiment of the present application, the MEC system coordinator determination module is further configured to:

Obtain multiple federated learning groups from a preset number of MEC systems;

The MEC systems that appear in any federated learning group are identified as elements of the set of available MEC coordinators, and the set of available MEC coordinators is constructed, and the priority of each MEC system in the set of MEC coordinators is repeated in each federated learning group. is proportional to the number of times.

In an optional embodiment of the present application, in an optional embodiment of the present application, the MEC system coordinator determination module is further configured to:

Identify MEC systems with the same data source among MEC systems as a federated learning group; and/or,

The MEC system whose model feature similarity in each MEC system meets the preset conditions is determined as a federated learning group.

In an optional embodiment of the present application, the model information receiving module is specifically used for:

Determine whether an existing MEC system coordinator is included in the federated learning group;

If not included, directly receive model information sent by other MEC systems in the federated learning group through the MEC system coordinator;

If it is included, the model information received by the existing MEC system coordinator is obtained through the MEC system coordinator, and the model information of other MEC systems is continuously received through the MEC system.

In an optional embodiment of the present application, the second model information aggregation module is specifically used for:

Based on the resource capacity utilization of each MEC system in the federated learning group, determine the new MEC system coordinator from each MEC system;

Obtain the model information received by the MEC system coordinator through the new MEC system coordinator, and continue to receive model information from other MEC systems through the new MEC system.

Referring next to FIG. 10 , it shows a schematic structural diagram of an electronic device (eg, a terminal device or a server that executes the method shown in FIG. 1 ) 1000 suitable for implementing an embodiment of the present application. The electronic devices in the embodiments of the present application may include, but are not limited to, such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablets), PMPs (portable multimedia players), in-vehicle terminals (such as mobile terminals such as in-vehicle navigation terminals), wearable devices, etc., and stationary terminals such as digital TVs, desktop computers, and the like. The electronic device shown in FIG. 10 is only an example, and should not impose any limitations on the functions and scope of use of the embodiments of the present application.

The electronic device includes: a memory and a processor, where the memory is used to store a program for executing the methods described in the above method embodiments; the processor is configured to execute the program stored in the memory. The processor here may be referred to as the processing device 1001 described below, and the memory may include at least one of a read-only memory (ROM) 1002, a random access memory (RAM) 1003, and a storage device 1008 described below, and the details are as follows shown:

As shown in FIG. 10, an electronic device 1000 may include a processing device (eg, a central processing unit, a graphics processor, etc.) 1001, which may be loaded into random access according to a program stored in a read only memory (ROM) 1002 or from a storage device 1008 Various appropriate operations and processes are executed by the programs in the memory (RAM) 1003 . In the RAM 1003, various programs and data necessary for the operation of the electronic device 1000 are also stored. The processing device 1001 , the ROM 1002 , and the RAM 1003 are connected to each other through a bus 1004 . An input/output (I/O) interface 1005 is also connected to the bus 1004 .

Typically, the following devices can be connected to the I/O interface 1005: input devices 1006 including, for example, a touch screen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; including, for example, a liquid crystal display (LCD), speakers, vibration An output device 1007 such as a computer; a storage device 1008 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009 . The communication means 1009 may allow the electronic device 1000 to communicate wirelessly or by wire with other devices to exchange data. While FIG. 10 illustrates an electronic device having various means, it should be understood that not all of the illustrated means are required to be implemented or available. More or fewer devices may alternatively be implemented or provided.

In particular, according to embodiments of the present application, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present application include a computer program product comprising a computer program carried on a non-transitory computer readable medium, the computer program containing program code for performing the method illustrated in the flowchart. In such an embodiment, the computer program may be downloaded and installed from the network via the communication device 1009 , or from the storage device 1008 , or from the ROM 1002 . When the computer program is executed by the processing apparatus 1001, the above-mentioned functions defined in the methods of the embodiments of the present application are executed.

It should be noted that, the computer-readable storage medium mentioned above in the present application may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or a combination of any of the above. More specific examples of computer readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable Programmable read only memory (EPROM or flash memory), fiber optics, portable compact disk read only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, however, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code therein. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device . Program code embodied on a computer readable medium may be transmitted using any suitable medium including, but not limited to, electrical wire, optical fiber cable, RF (radio frequency), etc., or any suitable combination of the foregoing.

In some embodiments, the client and server can communicate using any currently known or future developed network protocol such as HTTP (HyperText Transfer Protocol), and can communicate with digital data in any form or medium (eg, a communications network) interconnected. Examples of communication networks include local area networks ("LAN"), wide area networks ("WAN"), the Internet (eg, the Internet), and peer-to-peer networks (eg, ad hoc peer-to-peer networks), as well as any currently known or future development network of.

The above-mentioned computer-readable medium may be included in the above-mentioned electronic device; or may exist alone without being assembled into the electronic device.

The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the electronic device, the electronic device:

When any multi-access edge computing MEC system in any federated learning group issues a federated learning request, based on the resource capacity utilization of each MEC system in the federated learning group, the MEC system coordinator is determined from each MEC system; The system coordinator receives the model information sent by other MEC systems in the federated learning group, and obtains the resource capacity utilization rate of the MEC system coordinator in real time during the receiving process; if the resource capacity utilization rate of the MEC system coordinator during the receiving process does not exceed the expected Set the threshold, after the information of each model is aggregated by the MEC system coordinator, the aggregated model information is sent to each MEC system in the federated learning group; if the resource capacity utilization rate of the MEC system coordinator exceeds the predetermined value during the receiving process. Set the threshold, determine the new MEC system coordinator from each MEC system and switch to receive each model information through the new MEC system coordinator, repeat the determination and switching of the new MEC system coordinator until the new MEC system coordinator is in During the receiving process, if the resource capacity utilization does not exceed the preset threshold, the new MEC system coordinator aggregates the model information, and then sends the aggregated model information to each MEC system in the federated learning group.

Computer program code for performing the operations of the present application may be written in one or more programming languages, including but not limited to object-oriented programming languages—such as Java, Smalltalk, C++, and This includes conventional procedural programming languages - such as the "C" language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (eg, using an Internet service provider through Internet connection).

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code that contains one or more logical functions for implementing the specified functions executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It is also noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented in dedicated hardware-based systems that perform the specified functions or operations , or can be implemented in a combination of dedicated hardware and computer instructions.

The modules or units involved in the embodiments of the present application may be implemented in a software manner, and may also be implemented in a hardware manner. Wherein, the name of the module or unit does not constitute a limitation of the unit itself under certain circumstances, for example, the first program switching module may also be described as "a module for switching the first program".

The functions described herein above may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), Systems on Chips (SOCs), Complex Programmable Logical Devices (CPLDs) and more.

In the context of this application, a machine-readable medium may be a tangible medium that may contain or store the program for use by or in connection with the instruction execution system, apparatus or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media may include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media would include one or more wire-based electrical connections, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), fiber optics, compact disk read only memory (CD-ROM), optical storage, magnetic storage, or any suitable combination of the foregoing.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, for the specific method implemented when the computer-readable medium described above is executed by an electronic device, reference may be made to the corresponding process in the foregoing method embodiment, which is not repeated here. Repeat.

Embodiments of the present application provide a computer program product or computer program, where the computer program product or computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that when the computer device executes, the following conditions are realized:

When any multi-access edge computing MEC system in any federated learning group issues a federated learning request, based on the resource capacity utilization of each MEC system in the federated learning group, the MEC system coordinator is determined from each MEC system; The system coordinator receives the model information sent by other MEC systems in the federated learning group, and obtains the resource capacity utilization rate of the MEC system coordinator in real time during the receiving process; if the resource capacity utilization rate of the MEC system coordinator during the receiving process does not exceed the expected Set the threshold, after the information of each model is aggregated by the MEC system coordinator, the aggregated model information is sent to each MEC system in the federated learning group; if the resource capacity utilization rate of the MEC system coordinator exceeds the predetermined value during the receiving process. Set the threshold, determine the new MEC system coordinator from each MEC system and switch to receive each model information through the new MEC system coordinator, repeat the determination and switching of the new MEC system coordinator until the new MEC system coordinator is in During the receiving process, if the resource capacity utilization does not exceed the preset threshold, the new MEC system coordinator aggregates the model information, and then sends the aggregated model information to each MEC system in the federated learning group.

The above are only some embodiments of the present invention. It should be pointed out that for those skilled in the art, without departing from the principles of the present invention, several improvements and modifications can be made. It should be regarded as the protection scope of the present invention.

Claims (10)

1. A MEC federated learning method, characterized in that, comprising: When any multi-access edge computing MEC system in any federated learning group sends a federated learning request, the MEC system coordinator is determined from each MEC system based on the resource capacity utilization of each MEC system in the federated learning group; Receive model information sent by other MEC systems in the federated learning group through the MEC system coordinator, and obtain the resource capacity utilization rate of the MEC system coordinator in real time during the receiving process; If the resource capacity utilization rate of the MEC system coordinator does not exceed the first preset threshold during the receiving process, after the MEC system coordinator aggregates the model information, the aggregated model information is sent separately to each MEC system in the federated learning group; If the resource capacity utilization rate of the MEC system coordinator exceeds the first preset threshold during the receiving process, a new MEC system coordinator is determined from each MEC system and switched to the new MEC system coordinator Receive each model information, repeat the determination and handover of the new MEC system coordinator until the resource capacity utilization rate of the new MEC system coordinator does not exceed the first preset threshold during the receiving process, then the new MEC system coordinator After each model information is aggregated, the aggregated model information is respectively sent to each MEC system in the federated learning group. 2. The method according to claim 1, wherein, determining the MEC system coordinator from each MEC system based on the resource capacity utilization rate of each MEC system in the federated learning group, comprising: Obtain a preset available MEC system coordinator set, where each MEC system in the preset available MEC system coordinator set carries a corresponding priority; Among the MEC systems in the federated learning group that belong to the set of available MEC system coordinators, the resource capacity utilization rate does not exceed the second preset threshold and the MEC system with the highest priority is determined as the MEC system coordinator, so The second preset threshold is not greater than the first preset threshold. 3. The method according to claim 2, wherein the method further comprises: If the resource capacity utilization of all the MEC systems belonging to the available MEC system coordinator set in the federated learning group exceeds the second preset threshold, the federated learning group does not belong to the available MEC systems The MEC system with the smallest resource capacity utilization rate among the MEC systems in the set of coordinators is determined as the MEC system coordinator. 4. The method according to claim 2, wherein the acquiring a preset available MEC system coordinator set comprises: Obtain multiple federated learning groups from a preset number of MEC systems; The MEC systems that appear in any federated learning group are determined as elements of the set of available MEC coordinators, and the set of available MEC coordinators is constructed, and the priority of each MEC system in the set of MEC coordinators is the same as that of each MEC coordinator set. The number of repetitions in the federated learning group is proportional. 5. The method according to claim 4, wherein, obtaining a plurality of federated learning groups from a preset number of MEC systems, comprising: Identify MEC systems with the same data source among MEC systems as a federated learning group; and/or, The MEC system whose model feature similarity in each MEC system meets the preset conditions is determined as a federated learning group. 6. The method according to claim 1, wherein the receiving, through the MEC system coordinator, model information sent by other MEC systems in the federated learning group, comprises: determining whether the federated learning group includes an existing MEC system coordinator; If not included, directly receive the model information sent by other MEC systems in the federated learning group through the MEC system coordinator; If it is included, the MEC system coordinator obtains the model information that has been received by the existing MEC system coordinator, and continues to receive model information of other MEC systems through the MEC system. 7. method according to claim 1, is characterized in that, described from each MEC system, determine new MEC system coordinator and switch to new MEC system coordinator to receive each model information, comprising: determining the new MEC system coordinator from each MEC system based on the resource capacity utilization of each MEC system in the federated learning group; The model information received by the MEC system coordinator is acquired through the new MEC system coordinator, and the model information of other MEC systems is continuously received through the new MEC system. 8. A MEC federated learning device, comprising: The MEC system coordinator determines the module, which is used for, when a federated learning request is sent by any multi-access edge computing MEC system in any federated learning group, based on the resource capacity utilization of each MEC system in the federated learning group, from each MEC system The MEC system coordinator is determined in the system; a model information receiving module, configured to receive model information sent by other MEC systems in the federated learning group through the MEC system coordinator, and obtain the resource capacity utilization rate of the MEC system coordinator in real time during the receiving process; A first model information aggregation module, configured to aggregate each model information through the MEC system coordinator if the resource capacity utilization rate of the MEC system coordinator does not exceed a first preset threshold during the receiving process. , send the aggregated model information to each MEC system in the federated learning group; A second model information aggregation module, configured to determine a new MEC system coordinator from each MEC system if the resource capacity utilization rate of the MEC system coordinator exceeds the first preset threshold during the receiving process And switch to the new MEC system coordinator to receive each model information, repeat the determination and switching of the new MEC system coordinator until the resource capacity utilization rate of the new MEC system coordinator does not exceed the first preset threshold during the receiving process, then After the model information is aggregated by the new MEC system coordinator, the aggregated model information is respectively sent to each MEC system in the federated learning group. 9. An electronic device, comprising a memory and a processor; A computer program is stored in the memory; The processor for executing the computer program to implement the method of any one of claims 1 to 7. 10. A computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented .

Images