CN112990276B - Federal learning method, device, equipment and storage medium based on self-organizing cluster - Google Patents
Federal learning method, device, equipment and storage medium based on self-organizing cluster Download PDFInfo
- Publication number
- CN112990276B CN112990276B CN202110193253.5A CN202110193253A CN112990276B CN 112990276 B CN112990276 B CN 112990276B CN 202110193253 A CN202110193253 A CN 202110193253A CN 112990276 B CN112990276 B CN 112990276B
- Authority
- CN
- China
- Prior art keywords
- user equipment
- cluster
- model
- aggregation
- determining
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Mobile Radio Communication Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to the technical field of artificial intelligence and discloses a federal learning method, a federal learning device, computer equipment and a computer readable storage medium based on self-organizing clusters, wherein the method comprises the following steps: acquiring broadcast signals sent by each user equipment and generating a corresponding cluster; according to the cluster, determining target user equipment in the cluster, and taking the target user equipment as a central node; receiving model parameters sent by each user equipment, and sending each model parameter to the central node; acquiring aggregate model parameters after the central node performs aggregate federation learning on each model parameter; and sending the aggregate model parameters to each user equipment, and updating the model parameters of the preset model in each user equipment, so that joint FL model training can be provided without using a predetermined centralized cloud server, and the problem of single-point failure of the predetermined centralized server is effectively avoided.
Description
Technical Field
The present application relates to the field of artificial intelligence technologies, and in particular, to a federal learning method, apparatus, computer device, and computer readable storage medium based on self-organizing clusters.
Background
In conventional machine learning training, data is typically stored centrally in a central entity that must first collect data from various data sources for later learning, which can present a series of security and privacy concerns. The Federal Learning (FL) provides a distributed learning mode, and iterative exchange of parameters is continuously carried out based on a learning model between a terminal device and a centralized server until a global FL model converges to a certain precision level, so that data is not required to be transferred from the terminal device to the centralized server in the whole course, and the method is a very promising training paradigm of machine learning.
While FL shows great advantages in protecting data privacy while achieving collaborative machine learning, it still faces some problems. Since FL requires the use of a centralized server for iterative parameter modeling and parameter aggregation with participating clients in the training process, failure of the server can result in failure of the FL process if physically damaged or attacked.
Disclosure of Invention
The main purpose of the present application is to provide a federal learning method, a federal learning device, a federal learning computer device, and a federal learning computer readable storage medium based on self-organizing clusters, so as to solve the technical problem that in the existing centralized server used as a training process, the centralized server is physically damaged or attacked, and the failure of the FL process can be caused by the failure of the centralized server.
In a first aspect, the present application provides a federal learning method based on an ad hoc cluster, the federal learning method based on an ad hoc cluster including the steps of:
acquiring broadcast signals sent by each user equipment and generating a corresponding cluster;
according to the cluster, determining target user equipment in the cluster, and taking the target user equipment as a central node;
receiving model parameters sent by each user equipment, and sending each model parameter to the central node;
acquiring aggregate model parameters after the central node performs aggregate federation learning on each model parameter;
and sending the aggregate model parameters to each user equipment, and updating the model parameters of a preset model in each user equipment.
In a second aspect, the present application further provides a federal device based on ad hoc clusters, the federal device based on ad hoc clusters comprising:
the generating module is used for acquiring broadcast signals sent by each user equipment and generating corresponding clusters;
the determining module is used for determining target user equipment in the cluster according to the cluster, and taking the target user equipment as a central node;
the receiving and transmitting module is used for receiving the model parameters transmitted by each piece of user equipment and transmitting each piece of model parameters to the central node;
the acquisition module is used for acquiring the aggregate model parameters returned after the central node performs aggregate federal learning on the model parameters;
and the updating module is used for sending the aggregate model parameters to each user equipment and updating the model parameters of a preset model in each user equipment.
In a third aspect, the present application also provides a computer device comprising a processor, a memory, and a computer program stored on the memory and executable by the processor, wherein the computer program when executed by the processor implements the steps of the self-organizing cluster-based federal learning method as described above.
In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the self-organizing cluster-based federal learning method as described above.
The application provides a federal learning method, a federal learning device, computer equipment and a computer readable storage medium based on self-organizing clusters, which are used for generating corresponding clusters by acquiring broadcast signals sent by all user equipment; according to the cluster, determining target user equipment in the cluster, and taking the target user equipment as a central node; receiving model parameters sent by each user equipment, and sending each model parameter to the central node; acquiring aggregate model parameters after the central node performs aggregate federation learning on each model parameter; and sending the aggregate model parameters to each user equipment, and updating the model parameters of the preset model in each user equipment, so that joint FL model training can be provided without using a predetermined centralized cloud server, and the problem of single-point failure of the predetermined centralized server is effectively avoided.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a federal learning method based on self-organizing clusters according to an embodiment of the present application;
FIG. 2 is a flow chart of sub-steps of the federation learning method based on the ad hoc cluster of FIG. 1;
FIG. 3 is a flow chart illustrating sub-steps of the federation learning method based on the ad hoc cluster of FIG. 1;
FIG. 4 is a schematic flow chart of another federal learning method based on ad hoc clusters according to an embodiment of the present disclosure;
FIG. 5 is a schematic block diagram of a federal appliance based on ad hoc clusters provided in an embodiment of the present application;
fig. 6 is a schematic block diagram of a computer device according to an embodiment of the present application.
The realization, functional characteristics and advantages of the present application will be further described with reference to the embodiments, referring to the attached drawings.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all, of the embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are within the scope of the present disclosure.
The flow diagrams depicted in the figures are merely illustrative and not necessarily all of the elements and operations/steps are included or performed in the order described. For example, some operations/steps may be further divided, combined, or partially combined, so that the order of actual execution may be changed according to actual situations.
The embodiment of the application provides a federal learning method, a federal learning device, computer equipment and a computer readable storage medium based on self-organizing clusters. The federal learning method based on the self-organizing cluster can be applied to computer equipment, and the computer equipment can be electronic equipment such as a notebook computer, a desktop computer and the like.
Some embodiments of the present application are described in detail below with reference to the accompanying drawings. The following embodiments and features of the embodiments may be combined with each other without conflict.
Referring to fig. 1, fig. 1 is a flow chart of a federal learning method based on self-organizing clusters according to an embodiment of the present application.
As shown in fig. 1, the federation learning method based on the ad hoc cluster includes steps S101 to S105.
Step S101, obtaining broadcast signals sent by each user equipment, and generating corresponding clusters.
Exemplary, the broadcast signals sent by the user equipment are acquired, and the user equipment of the acquired broadcast signals is used as a cluster. For example, broadcast signals of 10 user equipments are acquired in a preset period, and the acquired 10 user equipments are used as a cluster. The broadcast signal includes identification information. For example, the identifier carried in the broadcast signal sent by each user equipment is obtained, the ID number of each user equipment is obtained through the identifier, and the user equipment corresponding to the ID number is used as a cluster.
In an embodiment, referring specifically to fig. 2, step S101 includes: substep S1011 to substep S1013.
Substep S1011, generating a corresponding group of graphs according to the broadcast signals transmitted by the respective user equipments.
Exemplary, the broadcast signals sent by the user devices are obtained, and the corresponding graph group is generated through the broadcast signals sent by the user devices. The graph community is a subset of vertices, the vertices in each subset being more closely connected relative to the other vertices of the network. For example, a broadcast signal sent by each user device is acquired, and the user devices that send the broadcast signals to each other are connected to generate a corresponding graph group, where each point is each user device.
In an embodiment, the generating a corresponding graph group according to the obtained broadcast signals sent by the user devices includes: acquiring broadcast signals sent by each user equipment; determining association information between the user equipment through broadcast signals sent by the user equipment; and generating a corresponding graph group through the association information among the user equipment.
Exemplary, the broadcast signals sent by the user devices are obtained, and the association information between the user devices is determined through the broadcast signals sent by the user devices. For example, the broadcast signals sent by the user equipment are acquired, and the identification user equipment corresponding to the broadcast signals sent by the user equipment is determined, so that the association relationship between the user equipment and the corresponding identification user equipment is determined. Acquiring a first broadcast signal sent by first user equipment, determining that second user equipment receives the first broadcast signal through the first broadcast signal, and determining that association information exists between the first user equipment and the second user equipment. Or, acquiring a first broadcast signal sent by the first user equipment and a second broadcast signal sent by the second user equipment, and detecting that the first broadcast signal and the second broadcast signal are received by the third user equipment, wherein the first user equipment and the second user equipment have an association relationship with the third user equipment respectively; or, the first broadcast signal sent by the first user equipment is obtained, and if the first broadcast signal is detected to be received by the second user equipment and the third user equipment respectively, the first user equipment has an association relationship with the second user equipment and the third user equipment respectively. When the association information among the user equipment is acquired, each user equipment is used as a point, and the points are connected through the association information among the user equipment, so that a corresponding graph group is generated.
And step S1012, determining whether the user equipment are clustered together or not by carrying out aggregation calculation on the vertexes of the user equipment in the graph group.
Illustratively, in generating a graph community, vertices of individual user devices in the graph community are obtained. And determining whether the user equipment are clustered together or not by carrying out aggregation calculation on the vertexes of the user equipment. For example, each user equipment is a vertex in the generated graph group, the vertex of the first user equipment and the vertex of the second user equipment are subjected to aggregation calculation, and if the vertex of the first user equipment is connected with the vertex of the second user equipment, the first user equipment and the second user equipment are determined to be clustered in the same way; if the vertex of the first user equipment is not connected with the vertex of the second user equipment, determining that the first user equipment and the second user equipment are not clustered together.
In an embodiment, the determining whether the user equipment are clustered together by performing aggregate calculation on vertices of the user equipment in the graph group includes; calculating the vertex of each user equipment in the graph group through a preset aggregation formula to obtain the aggregation parameters of each user equipment; if the aggregation parameters are the first preset threshold values, determining that the user equipment are the same cluster; and if each aggregation parameter is a second preset threshold value, determining that the aggregation parameters are not the same cluster.
Exemplary, a preset aggregation formula is obtained to calculate the vertex of the user equipment in the graph group, so as to obtain the aggregation parameters corresponding to each user. For example, a preset aggregation formula is obtainedWherein M is a preset module value, L represents the number of edges contained in the graph group, N represents the number of vertexes, K i Representing the degree, K, of vertex i j Representing the degree of vertex j, A ij The value of (C) is a preset value in the adjacency matrix i Representing clusters of vertices i, C j Representing the cluster of vertices j, delta is then the Kronecker-delta function. By presetting an aggregation formula, delta (C) i C j ) Parameters.
And determining whether the user equipment are in the same cluster or not according to the aggregation parameters. If each aggregation parameter is a first preset threshold value, each aggregation parameter is determinedThe user equipment is the same cluster; and if the aggregation parameters are the second preset threshold value, determining that the aggregation parameters are not the same cluster. For example, when the polymerization parameter is delta (C i C j ) When parameters are given, if delta (C i C j ) If the parameter is 1, determining that the user equipment j and the user equipment i are clustered in the same way; if delta (C) i C j ) And if the parameter is 0, determining that the user equipment j and the user equipment i are not clustered together.
And step S1013, determining the user equipment in the same cluster, and taking the user equipment in the same cluster as a cluster.
When the user equipment in the same cluster is acquired, the user equipment in the same cluster is taken as a cluster. For example, when a plurality of user devices in the same cluster are acquired, the plurality of user devices in the same cluster are regarded as one cluster.
Step S102, determining target user equipment in the cluster according to the cluster, and taking the target user equipment as a central node.
When a cluster of the same cluster is acquired, any one user equipment in the cluster is determined to be used as target user equipment, and the target user equipment is used as a central node. For example, the cluster includes a first user device, a second user device, and the like, where the first user device is determined to be a target user device, and the number of the target user devices is one.
In one embodiment, referring specifically to fig. 3, step S102 includes: substep S1021 to substep S1022.
And step S1021, obtaining social centrality information of each user equipment in the cluster.
When a cluster of the same cluster is acquired, the social centrality information of each user equipment in the cluster is acquired, and the social centrality information and other nodes have higher social relations as selection criteria of the central node.
In an embodiment, the acquiring social centrality information of each of the user devices in the cluster includes: acquiring social relations among all the user equipment in the cluster; obtaining social centrality vector information of each user equipment through the social relationship; and calculating the social centrality vector information of each user equipment to obtain the social centrality information of each user equipment.
Exemplary, social relationships between individual user devices in a cluster are obtained. For example, if the first user equipment is connected with the second user equipment and the third user equipment respectively, the connection relationship between the first user equipment and the second user equipment and the connection relationship between the first user equipment and the third user equipment respectively are used as the social relationship of the first user equipment. And obtaining the social center vector information of each user equipment through the social relationship of each user equipment. For example, if the first user device is connected to the second user device and the third user device respectively, the social center vector information of the first user device is S 1 (S 2 ,S 3 ) The method comprises the steps of carrying out a first treatment on the surface of the Or the first user equipment is respectively connected with the second user equipment, the third user equipment and the fourth user equipment, and the social center vector information of the first user equipment is S 1 (S 2 ,S 3 ,S 4 ). And calculating the social center vector information of each user equipment to obtain the social center information of each user equipment. For example, when the social center vector information of the first user equipment is obtained as S 1 (S 2 ,S 3 ) And when the social centrality information of the first user equipment is determined to be 2. Or, after obtaining the social center vector information of the first user equipment as S 1 (S 2 ,S 3 ,S 4 ) And determining the social centrality information of the first user equipment to be 3.
Substep S1022, determining a corresponding target user equipment according to the social centrality information of each user equipment, and taking the target user equipment as a central node.
Exemplary, when the social-centering information of each user device is obtained, the target user device is determined by comparing the social-centering information of each user device. For example, when the sociality information of the first user equipment is acquired to be 3, and when the sociality information of the second user equipment is acquired to be 4, the second user equipment is determined to be the target user equipment, and the determined target user equipment is taken as a central node.
Step S103, receiving the model parameters sent by each user equipment, and sending each model parameter to the central node.
Illustratively, each user device includes a pre-set model therein, including a pre-set neural network model, a deep learning model, a pre-trained language model, and the like. And when model parameters in the current preset model sent by each user equipment are received, the model parameters in the current preset model sent by the user equipment are sent to the central node.
And step S104, acquiring the aggregate model parameters after the central node performs aggregate federation learning on the model parameters.
The central node comprises preset aggregate federation models, transmits an uploading request to the central node, receives an encryption public key transmitted by the central node, encrypts model parameters of each preset model through the encryption public key, and transmits the encrypted model parameters to the central node. And when the central node receives the encrypted model parameters, decrypting each encrypted model parameter respectively to obtain the decrypted model parameters of each preset model. And learning each model parameter by presetting an aggregation federation model in the central node to obtain a corresponding aggregation model parameter. The aggregate federation model comprises a aggregate transverse federation model, a aggregate longitudinal federation model, an aggregate federation migration model and the like.
It should be noted that federal learning refers to a method of machine learning modeling by combining different clients or participants. In federal learning, the client does not need to expose own data to other clients and coordinators (also called servers), so that the federal learning can well protect user privacy and ensure data security, and can solve the problem of data islanding. Federal learning has the following advantages: data isolation is carried out, data cannot be revealed to the outside, and the requirements of user privacy protection and data security are met; the quality of the federal learning model is ensured to be lossless, negative migration is avoided, and the federal learning model is ensured to have better effect than a split independent model; the encryption exchange of information and model parameters can be ensured under the condition that the independence of each client is maintained, and growth is obtained at the same time.
Step 105, sending the aggregate model parameters to each user device, and updating the model parameters of the preset model in each user device.
Exemplary, after learning each model parameter through the preset aggregate federal model in the central node is obtained, the corresponding aggregate model parameter is obtained, and then the aggregate model parameter is sent to each user device, and the model parameter of the preset model in each user device is updated.
In the embodiment of the application, the corresponding cluster is generated by acquiring the broadcast signals of each user equipment, so that the target user equipment in the cluster is determined, the user equipment is used as a central node to receive the model parameters of each user equipment to perform aggregation federal learning, and the model parameters of the preset model in each user equipment are updated by the aggregation model parameters, so that joint FL model training can be provided without using a predetermined centralized cloud server, and the problem of single-point failure of the predetermined centralized server is effectively avoided.
Referring to fig. 4, fig. 4 is a flow chart of another federal learning method based on self-organizing clusters according to an embodiment of the present application.
As shown in fig. 4, the federation learning method based on the ad hoc cluster includes steps S201 to S203.
Step S201, determining whether the preset model is in a convergence state.
Illustratively, it is determined whether the preset model is in a converged state. For example, the aggregation model parameter is compared with the previously recorded aggregation model parameter, and if the aggregation model parameter is the same as the previously recorded aggregation model parameter, or the difference between the aggregation model parameter and the previously recorded aggregation model parameter is smaller than a preset difference, the preset model is determined to be in a convergence state.
Step S202, if the preset model is in a convergence state, the preset model is used as a corresponding aggregation model.
Exemplary, if the aggregation model parameter information is the same as the previously recorded aggregation model parameter, or the difference between the aggregation model parameter information and the previously recorded aggregation model parameter is smaller than the preset difference, the preset model is used as the corresponding aggregation model.
Step S203, if the preset model is not in a convergence state, receiving second model parameters sent by each user device, and training the preset model through the second model parameters.
If the preset model is determined not to be in the convergence state, second model parameters of the preset model in each user device are continuously acquired, aggregation federation learning is conducted on each second model parameter through the center node, and second aggregation model parameters after aggregation federation learning are acquired. And sending the second aggregation model parameters to each user equipment, and updating the aggregation model parameters of the preset model in the user equipment.
In the embodiment of the application, whether the preset model is in the convergence state or not is detected, and training is continuously carried out on the preset model when the preset model is not in the convergence state, so that the preset model is ensured to be in the convergence state, and inaccurate preset results of the preset model when the preset model is not in the convergence state are effectively avoided.
Referring to fig. 5, fig. 5 is a schematic block diagram of a federation device based on an ad hoc cluster according to an embodiment of the present application.
As shown in fig. 5, the federation device 400 based on self-organizing clusters includes: a generating module 401, a determining module 402, a receiving and transmitting module 403, an obtaining module 404 and an updating module 405.
A generating module 401, configured to obtain broadcast signals sent by each user equipment, and generate a corresponding cluster;
a determining module 402, configured to determine, according to the cluster, a target user equipment in the cluster, and take the target user equipment as a central node;
a receiving and sending module 403, configured to receive the model parameters sent by each piece of user equipment, and send each piece of model parameters to the central node;
an obtaining module 404, configured to obtain aggregate model parameters returned after the central node performs aggregate federation learning on each model parameter;
and the updating module 405 is configured to send the aggregate model parameters to each of the user devices, and update the model parameters of the preset model in each of the user devices.
The generating module 401 is specifically further configured to:
generating a corresponding graph group according to the obtained broadcast signals sent by the user equipment;
determining whether the user equipment are clustered together or not by carrying out aggregation calculation on the vertexes of the user equipment in the graph group;
and determining the user equipment of the same cluster, and taking the user equipment of the same cluster as a cluster.
The generating module 401 is specifically further configured to:
acquiring broadcast signals sent by each user equipment;
determining association information between the user equipment through broadcast signals sent by the user equipment;
and generating a corresponding graph group through the association information among the user equipment.
The generating module 401 is specifically further configured to:
calculating the vertex of each user equipment in the graph group through a preset aggregation formula to obtain the aggregation parameters of each user equipment;
if the aggregation parameters are the first preset threshold values, determining that the user equipment are the same cluster;
and if each aggregation parameter is a second preset threshold value, determining that the aggregation parameters are not the same cluster.
The determining module 402 is specifically further configured to:
acquiring social centrality information of each user equipment in the cluster;
and determining corresponding target user equipment through the social centrality information of each user equipment, and taking the target user equipment as a central node.
The determining module 402 is specifically further configured to:
acquiring social relations among all the user equipment in the cluster;
obtaining social centrality vector information of each user equipment through the social relationship;
and calculating the social centrality vector information of each user equipment to obtain the social centrality information of each user equipment.
Wherein, federal device based on self-organizing clusters is specifically further used for:
determining whether the preset model is in a convergence state;
if the preset model is in a convergence state, the preset model is used as a corresponding aggregation model;
and if the preset model is not in a convergence state, receiving second model parameters sent by each user equipment, and training the preset model through the second model parameters.
It should be noted that, for convenience and brevity of description, specific working processes of the above-described apparatus and each module and unit may refer to corresponding processes in the foregoing embodiment of the federal learning method based on the ad hoc cluster, which are not described herein again.
The apparatus provided by the above embodiments may be implemented in the form of a computer program which may be run on a computer device as shown in fig. 6.
Referring to fig. 6, fig. 6 is a schematic block diagram of a computer device according to an embodiment of the present application. The computer device may be a terminal.
As shown in fig. 6, the computer device includes a processor, a memory, and a network interface connected by a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.
The non-volatile storage medium may store an operating system and a computer program. The computer program comprises program instructions that, when executed, cause the processor to perform any one of a federal learning method based on ad hoc clusters.
The processor is used to provide computing and control capabilities to support the operation of the entire computer device.
The internal memory provides an environment for the execution of a computer program in a non-volatile storage medium that, when executed by a processor, causes the processor to perform any of a number of federal learning methods based on ad hoc clusters.
The network interface is used for network communication such as transmitting assigned tasks and the like. It will be appreciated by those skilled in the art that the structure shown in fig. 6 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
It should be appreciated that the processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processors, digital signal processors (Digita l Signa l Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), field-programmable gate arrays (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. Wherein the general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
Wherein in one embodiment the processor is configured to run a computer program stored in the memory to implement the steps of:
acquiring broadcast signals sent by each user equipment and generating a corresponding cluster;
according to the cluster, determining target user equipment in the cluster, and taking the target user equipment as a central node;
receiving model parameters sent by each user equipment, and sending each model parameter to the central node;
acquiring aggregate model parameters after the central node performs aggregate federation learning on each model parameter;
and sending the aggregate model parameters to each user equipment, and updating the model parameters of a preset model in each user equipment.
In one embodiment, the processor is configured to, when acquiring the broadcast signals sent by each user equipment and generating the corresponding cluster implementation, implement:
generating a corresponding graph group according to the obtained broadcast signals sent by the user equipment;
determining whether the user equipment are clustered together or not by carrying out aggregation calculation on the vertexes of the user equipment in the graph group;
and determining the user equipment of the same cluster, and taking the user equipment of the same cluster as a cluster.
In one embodiment, the processor is configured to, when generating the corresponding group of graphs according to the obtained broadcast signals sent by the respective user devices, implement:
acquiring broadcast signals sent by each user equipment;
determining association information between the user equipment through broadcast signals sent by the user equipment;
and generating a corresponding graph group through the association information among the user equipment.
In one embodiment, the processor is configured to, when determining whether the user devices are implemented in the same cluster by performing aggregate computation on vertices of the user devices in the graph group, implement:
calculating the vertex of each user equipment in the graph group through a preset aggregation formula to obtain the aggregation parameters of each user equipment;
if the aggregation parameters are the first preset threshold values, determining that the user equipment are the same cluster;
and if each aggregation parameter is a second preset threshold value, determining that the aggregation parameters are not the same cluster.
In one embodiment, the processor is configured to, when determining, according to the cluster, a target user equipment in the cluster and implementing the target user equipment as a central node, implement:
acquiring social centrality information of each user equipment in the cluster;
and determining corresponding target user equipment through the social centrality information of each user equipment, and taking the target user equipment as a central node.
In one embodiment, the processor is configured, when executing the acquiring social centrality information of each of the user devices in the cluster, to execute:
acquiring social relations among all the user equipment in the cluster;
obtaining social centrality vector information of each user equipment through the social relationship;
and calculating the social centrality vector information of each user equipment to obtain the social centrality information of each user equipment.
In one embodiment, the processor is configured to, when sending the aggregate model parameters to each of the user devices and updating the model parameters of the preset model in each of the user device updates, implement:
determining whether the preset model is in a convergence state;
if the preset model is in a convergence state, the preset model is used as a corresponding aggregation model;
and if the preset model is not in a convergence state, receiving second model parameters sent by each user equipment, and training the preset model through the second model parameters.
Embodiments of the present application also provide a computer readable storage medium having a computer program stored thereon, where the computer program includes program instructions, where the method implemented when the program instructions are executed may refer to various embodiments of the federal learning method based on ad hoc clusters of the present application.
The computer readable storage medium may be an internal storage unit of the computer device according to the foregoing embodiment, for example, a hard disk or a memory of the computer device. The computer readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like, which are provided on the computer device.
Further, the computer-readable storage medium may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function, and the like; the storage data area may store data created from the use of blockchain nodes, and the like.
The blockchain referred to in the application is a novel application mode of computer technologies such as storage of a preset model, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. The Blockchain (Blockchain), which is essentially a decentralised database, is a string of data blocks that are generated by cryptographic means in association, each data block containing a batch of information of network transactions for verifying the validity of the information (anti-counterfeiting) and generating the next block. The blockchain may include a blockchain underlying platform, a platform product services layer, an application services layer, and the like.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or system that comprises the element.
The foregoing embodiment numbers of the present application are merely for describing, and do not represent advantages or disadvantages of the embodiments. While the invention has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions of equivalents may be made and equivalents will be apparent to those skilled in the art without departing from the scope of the invention. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (7)
1. A federal learning method based on ad hoc clusters, comprising:
acquiring broadcast signals sent by each user equipment and generating a corresponding graph group;
determining whether the user equipment are clustered together or not by carrying out aggregation calculation on the vertexes of the user equipment in the graph group;
determining the user equipment of the same cluster, and taking the user equipment of the same cluster as a cluster;
according to the cluster, determining target user equipment in the cluster, and taking the target user equipment as a central node;
receiving model parameters sent by each user equipment, and sending each model parameter to the central node;
acquiring aggregate model parameters after the central node performs aggregate federation learning on each model parameter;
sending the aggregation model parameters to each user equipment, and updating the model parameters of a preset model in each user equipment;
the obtaining the broadcast signals sent by each user equipment and generating a corresponding graph group comprises the following steps:
acquiring broadcast signals sent by each user equipment;
determining association information between the user equipment through broadcast signals sent by the user equipment;
generating a corresponding graph group through the association information among the user equipment;
the method comprises the steps of determining whether the user equipment are clustered together or not by carrying out aggregation calculation on the vertexes of the user equipment in the graph group, wherein the aggregation calculation comprises the steps of;
calculating the vertex of each user equipment in the graph group through a preset aggregation formula to obtain the aggregation parameters of each user equipment;
if the aggregation parameters are the first preset threshold values, determining that the user equipment are the same cluster;
and if the aggregation parameters are the second preset threshold values, determining that the user equipment are not in the same cluster.
2. The federal learning method based on ad hoc clusters according to claim 1, wherein the determining a target user equipment in the cluster according to the clusters and taking the target user equipment as a central node comprises:
acquiring social centrality information of each user equipment in the cluster;
and determining corresponding target user equipment through the social centrality information of each user equipment, and taking the target user equipment as a central node.
3. The federal learning method based on ad hoc clusters according to claim 2, wherein the acquiring social centrality information of each of the user equipments in the clusters comprises:
acquiring social relations among all the user equipment in the cluster;
obtaining social centrality vector information of each user equipment through the social relationship;
and calculating the social centrality vector information of each user equipment to obtain the social centrality information of each user equipment.
4. The federal learning method based on self-organizing clusters according to claim 1, wherein said sending the aggregate model parameters to each of the user devices, after updating the model parameters of the preset model in each of the user device updates, further comprises:
determining whether the preset model is in a convergence state;
if the preset model is in a convergence state, the preset model is used as a corresponding aggregation model;
and if the preset model is not in a convergence state, receiving second model parameters sent by each user equipment, and training the preset model through the second model parameters.
5. A federal appliance based on ad hoc clusters, comprising:
the generating module is used for acquiring broadcast signals sent by all the user equipment, generating a corresponding graph group, determining whether all the user equipment are clustered together or not by carrying out aggregation calculation on the vertexes of all the user equipment in the graph group, determining the user equipment clustered together, and taking the user equipment clustered together as a cluster;
the determining module is used for determining target user equipment in the cluster according to the cluster, and taking the target user equipment as a central node;
the receiving and transmitting module is used for receiving the model parameters transmitted by each piece of user equipment and transmitting each piece of model parameters to the central node;
the acquisition module is used for acquiring the aggregate model parameters returned after the central node performs aggregate federal learning on the model parameters;
the updating module is used for sending the aggregation model parameters to each user equipment and updating the model parameters of a preset model in each user equipment;
the obtaining the broadcast signals sent by each user equipment and generating a corresponding graph group comprises the following steps:
acquiring broadcast signals sent by each user equipment;
determining association information between the user equipment through broadcast signals sent by the user equipment;
generating a corresponding graph group through the association information among the user equipment;
the method comprises the steps of determining whether the user equipment are clustered together or not by carrying out aggregation calculation on the vertexes of the user equipment in the graph group, wherein the aggregation calculation comprises the steps of;
calculating the vertex of each user equipment in the graph group through a preset aggregation formula to obtain the aggregation parameters of each user equipment;
if the aggregation parameters are the first preset threshold values, determining that the user equipment are the same cluster;
and if the aggregation parameters are the second preset threshold values, determining that the user equipment are not in the same cluster.
6. A computer device comprising a processor, a memory, and a computer program stored on the memory and executable by the processor, wherein the computer program when executed by the processor implements the steps of the ad hoc cluster-based federal learning method of any one of claims 1 to 4.
7. A computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, wherein the computer program, when executed by a processor, implements the steps of the ad hoc cluster based federal learning method according to any one of claims 1 to 4.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110193253.5A CN112990276B (en) | 2021-02-20 | 2021-02-20 | Federal learning method, device, equipment and storage medium based on self-organizing cluster |
| PCT/CN2021/097409 WO2022174533A1 (en) | 2021-02-20 | 2021-05-31 | Federated learning method and apparatus based on self-organized cluster, device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110193253.5A CN112990276B (en) | 2021-02-20 | 2021-02-20 | Federal learning method, device, equipment and storage medium based on self-organizing cluster |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112990276A CN112990276A (en) | 2021-06-18 |
| CN112990276B true CN112990276B (en) | 2023-07-21 |
Family
ID=76393764
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110193253.5A Active CN112990276B (en) | 2021-02-20 | 2021-02-20 | Federal learning method, device, equipment and storage medium based on self-organizing cluster |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112990276B (en) |
| WO (1) | WO2022174533A1 (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113255937B (en) * | 2021-06-28 | 2021-11-09 | 江苏奥斯汀光电科技股份有限公司 | Federal learning method and system for different intelligent agents in intelligent workshop |
| CN113487041B (en) * | 2021-07-15 | 2024-05-07 | 深圳市与飞科技有限公司 | Transverse federal learning method, device and storage medium |
| CN113469373B (en) * | 2021-08-17 | 2023-06-30 | 北京神州新桥科技有限公司 | Model training method, system, equipment and storage medium based on federal learning |
| CN113723509B (en) * | 2021-08-30 | 2024-01-16 | 平安科技（深圳）有限公司 | Follow-up monitoring method and device based on federal reinforcement learning and related equipment |
| CN114662340B (en) * | 2022-04-29 | 2023-02-28 | 烟台创迹软件有限公司 | Weighing model scheme determination method and device, computer equipment and storage medium |
| CN117808126B (en) * | 2024-02-29 | 2024-05-28 | 浪潮电子信息产业股份有限公司 | Machine learning method, device, equipment, federal learning system and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109271416A (en) * | 2018-09-03 | 2019-01-25 | 中国平安人寿保险股份有限公司 | Time management recommended method, electronic device and readable storage medium storing program for executing |
| CN111600707A (en) * | 2020-05-15 | 2020-08-28 | 华南师范大学 | Decentralized federal machine learning method under privacy protection |
| WO2020229684A1 (en) * | 2019-05-16 | 2020-11-19 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Concepts for federated learning, client classification and training data similarity measurement |
| CN112200263A (en) * | 2020-10-22 | 2021-01-08 | 国网山东省电力公司电力科学研究院 | Self-organizing federal clustering method applied to power distribution internet of things |
| CN112329940A (en) * | 2020-11-02 | 2021-02-05 | 北京邮电大学 | Personalized model training method and system combining federal learning and user portrait |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11475350B2 (en) * | 2018-01-22 | 2022-10-18 | Google Llc | Training user-level differentially private machine-learned models |
| CN111212110B (en) * | 2019-12-13 | 2022-06-03 | 清华大学深圳国际研究生院 | Block chain-based federal learning system and method |
| CN111966698B (en) * | 2020-07-03 | 2023-06-13 | 华南师范大学 | Block chain-based trusted federation learning method, system, device and medium |
| CN112232527B (en) * | 2020-09-21 | 2024-01-23 | 北京邮电大学 | Safe distributed federal deep learning method |
-
2021
- 2021-02-20 CN CN202110193253.5A patent/CN112990276B/en active Active
- 2021-05-31 WO PCT/CN2021/097409 patent/WO2022174533A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109271416A (en) * | 2018-09-03 | 2019-01-25 | 中国平安人寿保险股份有限公司 | Time management recommended method, electronic device and readable storage medium storing program for executing |
| WO2020229684A1 (en) * | 2019-05-16 | 2020-11-19 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Concepts for federated learning, client classification and training data similarity measurement |
| CN111600707A (en) * | 2020-05-15 | 2020-08-28 | 华南师范大学 | Decentralized federal machine learning method under privacy protection |
| CN112200263A (en) * | 2020-10-22 | 2021-01-08 | 国网山东省电力公司电力科学研究院 | Self-organizing federal clustering method applied to power distribution internet of things |
| CN112329940A (en) * | 2020-11-02 | 2021-02-05 | 北京邮电大学 | Personalized model training method and system combining federal learning and user portrait |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112990276A (en) | 2021-06-18 |
| WO2022174533A1 (en) | 2022-08-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112990276B (en) | Federal learning method, device, equipment and storage medium based on self-organizing cluster | |
| US11836583B2 (en) | Method, apparatus and system for secure vertical federated learning | |
| US20210158216A1 (en) | Method and system for federated learning | |
| CN110929886B (en) | Model training and predicting method and system | |
| US20230039182A1 (en) | Method, apparatus, computer device, storage medium, and program product for processing data | |
| CN110419053B (en) | System and method for information protection | |
| CN113505882B (en) | Data processing method based on federal neural network model, related equipment and medium | |
| CN113515760B (en) | Horizontal federal learning method, apparatus, computer device, and storage medium | |
| US11843586B2 (en) | Systems and methods for providing a modified loss function in federated-split learning | |
| JP6556659B2 (en) | Neural network system, share calculation device, neural network learning method, program | |
| CN113077060A (en) | Federal learning system and method aiming at edge cloud cooperation | |
| EP3566389A1 (en) | Distributed privacy-preserving verifiable computation | |
| CN113837392A (en) | System and method for computing model validation loss in decentralized machine learning | |
| CN102272728B (en) | Method, apparatus, and computer program product for polynomial-based data transformation and utilization | |
| CN112347500A (en) | Machine learning method, device, system, equipment and storage medium of distributed system | |
| US20230106985A1 (en) | Developing machine-learning models | |
| CN113065866A (en) | Internet of things edge computing system and method based on block chain | |
| US11991156B2 (en) | Systems and methods for secure averaging of models for federated learning and blind learning using secure multi-party computation | |
| CN115496204B (en) | Federal learning-oriented evaluation method and device under cross-domain heterogeneous scene | |
| CN111079153A (en) | Security modeling method and device, electronic equipment and storage medium | |
| CN116489637B (en) | Mobile edge computing method oriented to meta universe and based on privacy protection | |
| US20200183586A1 (en) | Apparatus and method for maintaining data on block-based distributed data storage system | |
| WO2023124219A1 (en) | Joint learning model iterative update method, apparatus, system, and storage medium | |
| CN115983275A (en) | Named entity identification method, system and electronic equipment | |
| CN115270176A (en) | Radix estimation method, system, computing device and computer storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |