Parkinson's disease prediction method based on adaptive federated learning and related equipment

Through adaptive federated learning dynamically adjusting user participation and model aggregation, the communication cost and model convergence speed problems of Parkinson's disease prediction in cross-device federated learning are solved, and efficient and low-cost early prediction of Parkinson's disease and personalized treatment plans are achieved.

CN120511002APending Publication Date: 2025-08-19SOUTH CHINA UNIV OF TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202510417995.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-04-03
Publication Date
2025-08-19

AI Technical Summary

Technical Problem

In the cross-device federated learning scenario, in the Parkinson's disease prediction task, how to effectively select the most suitable user equipment to participate in each round of training, while reducing communication costs, solve the problem of reduced global model convergence speed and inconsistent prediction accuracy caused by data acquisition differences.

Method used

Adaptive federated learning method is adopted to dynamically evaluate the local training progress of the device, combine computing power and data distribution, intelligently adjust the set of users participating in each round of training, optimize communication efficiency and global model convergence speed, and adaptively adjust the participation rate and aggregation ratio by using local convergence speed to reduce unnecessary communication frequency.

Benefits of technology

While maintaining high prediction accuracy, it effectively reduces communication costs, improves the intelligence level of Parkinson's disease prediction, and provides efficient and low-cost neurological disease monitoring solutions for smart wearable devices.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120511002A_ABST
    Figure CN120511002A_ABST
Patent Text Reader

Abstract

The invention discloses a Parkinson's disease prediction method based on adaptive federal learning and related equipment. The method comprises the following steps: step 1, initializing and broadcasting a global model; 2, user local training; step 3, the user uploads the local convergence speed of the local training to the server; step 4, the server adaptively calculates and adjusts the participation rate of the current round of communication, the selected users and parameters of the next round of local training of the users according to the local convergence speed, and sends calculation results to all the users; 5, updating the global model and broadcasting again; step 6, performing iterative training and model optimization; and step 7, predicting new patient data by using the global model obtained by training to realize early prediction of Parkinson's disease. According to the method, on the premise of ensuring the global model precision, the equipment with higher convergence speed and higher contribution degree is adaptively selected to participate in training, so that unnecessary communication overhead is reduced, and the model convergence efficiency is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the fields of medical health and federated learning, and in particular to a Parkinson's disease prediction method based on adaptive federated learning and related equipment. Background Art

[0002] Parkinson's disease is a common chronic neurodegenerative disease, the incidence of which increases significantly with age. It is estimated that by 2030, the number of Parkinson's patients worldwide will increase significantly, placing a huge burden on the medical system. Early diagnosis and prognosis are crucial to delaying disease progression and improving patients' quality of life. However, existing assessment methods, such as the Unified Parkinson's Disease Rating Scale (UPDRS), rely on the subjective judgment of clinicians and have certain assessment inconsistencies. In addition, traditional machine learning prediction methods usually rely on centralized data storage, but due to the privacy sensitivity of medical data and limited data sharing, centralized methods are difficult to promote in practical applications.

[0003] In recent years, Federated Learning (FL), as a distributed machine learning method, has provided a new solution for Parkinson's disease prediction in an environment with restricted data privacy. FL allows multiple devices or institutions to collaboratively train a global model without directly sharing data, thereby improving the model's generalization ability and protecting data privacy. However, in the cross-device federated learning scenario, data collection usually relies on smart wearable devices, and these devices have large differences in computing power, data quality, and online status. In addition, due to the different activity habits, physiological states, and device usage of different users, the data collected by each device is not independent and identically distributed, which reduces the convergence speed of the global model and affects the prediction accuracy. Therefore, in the Parkinson's disease prediction task, how to effectively select the most appropriate user device to participate in each round of training while reducing communication costs is a challenge faced by federated learning applications. Summary of the Invention

[0004] In order to at least solve one of the technical problems existing in the prior art to a certain extent, the purpose of the present invention is to provide a Parkinson's disease prediction method and related equipment based on adaptive federated learning.

[0005] The first technical solution adopted by the present invention is:

[0006] A Parkinson's disease prediction method based on adaptive federated learning includes the following steps:

[0007] Step 1: The server initializes the trained global model and broadcasts the global model to all users participating in federated learning;

[0008] Step 2: After receiving the global model, the user performs local training based on local data and the global model;

[0009] Step 3: The user uploads the local convergence speed of local training to the server;

[0010] Step 4: The server adaptively calculates and adjusts the participation rate of this round of communication, the selected users, and the parameters of the user's next round of local training based on the local convergence speed, and sends the calculated results to all users;

[0011] Step 5: The selected user uploads their local training model to the server for aggregation. After aggregation, the server broadcasts the global model to all users.

[0012] Step 6: During the global model training process, multiple rounds of user-server communication will be carried out, and steps 1-5 will be repeated. After multiple rounds of optimization, a Parkinson's disease prediction model that meets specific conditions will be formed;

[0013] Step 7: Use the trained global model to predict new patient data to achieve early prediction of Parkinson's disease, thereby assisting doctors in developing personalized treatment plans.

[0014] This method is based on the local convergence rate. While minimizing the loss of accuracy, it effectively controls the communication frequency between users and servers, balances the progress of each user's local training, and allows the global model to cover the entire data set as much as possible, thereby improving the model's accuracy and accelerating the convergence speed.

[0015] Furthermore, the step 2 includes:

[0016] Each user has a task to learn, which is trained on the user's device based on the local dataset. The task learning model of the mth user is θ m ;

[0017] The prediction function of the task learned by the mth user on the jth sample is Refers to the input sample After model θ m Output predicted value.

[0018] Furthermore, the goal of the mth user learning task is to minimize the loss function, which is expressed as follows:

[0019]

[0020] Where, l(θ m ) represents the loss function of the model, n m represents the number of data samples, represents the label of the task of the jth sample in the mth user, λ represents the penalty coefficient, ||θ m ||2 represents the l2 regularization term.

[0021] Furthermore, the step 3 includes:

[0022] During the K rounds of local training before communicating with the server, the change in user loss value is defined as the local convergence rate of local training. The local convergence rate v of the local training of the mth user is m for:

[0023]

[0024] Where l(K / 2) is the loss value of the user in the (K / 2)th iteration, and l(K) is the loss value of user m in the Kth iteration;

[0025] The user uploads the obtained local convergence speed to the server.

[0026] Furthermore, the step 4 includes:

[0027] The local convergence rate v of the server after receiving all users' uploaded local training m Finally, calculate the average speed of all users' s-th round of communication:

[0028]

[0029] Where M is the total number of users;

[0030] By comparing the average speed of this round with the average speed of the previous round, the participation rate β of this round is determined. If the convergence speed is faster than that of the previous round, the participation rate of users in this round is reduced, thereby reducing the number of user communications; otherwise, the participation rate is maintained and the overall model convergence is accelerated.

[0031] After determining the participation rate, the server arranges each user in descending order according to the local convergence speed of their local training, and selects the fastest users to participate in this round of aggregation, ultimately allowing the model to achieve a specific accuracy or loss value in fewer communication rounds.

[0032] Furthermore, the step 5 includes:

[0033] Before the server receives the model uploaded by the user for aggregation, it calculates the local convergence speed v m Calculate the aggregation ratio α for each user i :

[0034]

[0035] Where, v iis the local convergence rate of the i-th selected user, v j is the local convergence rate of all selected users, I is the total number of users participating in communication and aggregation in each round;

[0036] Let the faster user models account for a larger proportion in the aggregation process, and the slower user models account for a smaller proportion, and then aggregate the models according to this proportion:

[0037]

[0038] Where θ is the aggregated model, θ i The model uploaded by the i-th user.

[0039] The second technical solution adopted by the present invention is:

[0040] A Parkinson's disease prediction system based on adaptive federated learning, including:

[0041] The parameter distribution module is used to initialize the global model trained by the server and broadcast the global model to all users participating in federated learning;

[0042] The local training module is used by users to perform local training based on local data and the global model after receiving the global model;

[0043] The speed upload module is used by users to upload the local convergence speed of local training to the server;

[0044] The speed calculation module is used by the server to adaptively calculate and adjust the participation rate of the current round of communication, the selected users, and the parameters of the user's next round of local training based on the local convergence speed, and send the calculated results to all users;

[0045] The model aggregation module is used by selected users to upload their locally trained models to the server for aggregation. After aggregation, the server broadcasts the global model to all users.

[0046] The model iteration module is used for global model training, which involves multiple rounds of communication between the user and the server. After each round of communication, the parameters are returned to the parameter distribution module. After multiple rounds of optimization, a Parkinson's disease prediction model that meets specific conditions is formed.

[0047] Model prediction is used to use the trained global model to predict new patient data and achieve early prediction of Parkinson's disease.

[0048] The third technical solution adopted by the present invention is:

[0049] An electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, at least one program, the code set, or the instruction set is loaded and executed by the processor to implement a Parkinson's disease prediction method based on adaptive federated learning as described above.

[0050] The fourth technical solution adopted by the present invention is:

[0051] A computer-readable storage medium stores at least one instruction, at least one program, a code set, or an instruction set, wherein the at least one instruction, at least one program, the code set, or the instruction set is loaded and executed by a processor to implement a Parkinson's disease prediction method based on adaptive federated learning as described above.

[0052] The fifth technical solution adopted by the present invention is:

[0053] A computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned Parkinson's disease prediction method based on adaptive federated learning.

[0054] The beneficial effect of the present invention is that, under the premise of ensuring the accuracy of the global model, the present invention adaptively selects devices with faster convergence speed and higher contribution to participate in training, so as to reduce unnecessary communication overhead and improve the model convergence efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following introduction is made to the drawings of the embodiments of the present invention or the related technical solutions in the prior art. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative work.

[0056] Figure 1 This is a flowchart of the steps of a Parkinson's disease prediction method based on adaptive federated learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application and are not to be construed as limiting the present application. For the step numbers in the following embodiments, they are provided only for the convenience of explanation and are not intended to limit the order of the steps. The order of execution of the steps in the embodiments can be adaptively adjusted according to the understanding of those skilled in the art.

[0058] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments and are not intended to limit the embodiments of the present application. The singular forms of "a", "said", and "the" used in the embodiments of the present application and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise. In addition, unless otherwise clearly defined, words such as setting, installing, and connecting should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present invention in combination with the specific content of the technical solution.

[0059] In the description of this application, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on this application.

[0060] In the description of this application, "several" means one or more, "many" means more than two, "greater than," "less than," and "exceed" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. The terms "first" and "second" are used solely to distinguish technical features and are not to be construed as indicating or implying relative importance, or as implicitly specifying the number or order of the technical features indicated.

[0061] In the description of this application, "and / or" describes the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship.

[0062] To address existing technical issues, this paper provides a Parkinson's disease prediction method based on adaptive federated learning. This method dynamically evaluates the local training progress of a device and, based on its computing power, communication status, and data distribution, intelligently adjusts the set of users participating in each training round, thereby optimizing communication efficiency and global model convergence speed. This method effectively reduces communication costs while maintaining high prediction accuracy, enhancing the intelligent level of Parkinson's disease prediction and providing an efficient and low-cost solution for neurological disease monitoring based on smart wearable devices.

[0063] Example 1

[0064] like Figure 1 As shown, this embodiment provides a Parkinson's disease prediction method based on adaptive federated learning. This method dynamically evaluates the local training progress of the device and, in combination with the device's computing power, communication status, and data distribution, intelligently adjusts the set of users participating in each round of training, thereby optimizing communication efficiency and global model convergence speed. This method can effectively reduce communication costs while maintaining high prediction accuracy, improving the intelligence level of Parkinson's disease prediction, and providing an efficient and low-cost solution for neurological disease monitoring based on smart wearable devices. The method specifically includes the following steps:

[0065] S1. Global model initialization and broadcast: The server initializes the trained global model and broadcasts it to all users participating in federated learning.

[0066] S2. User local training: After receiving the global model, the user performs local training based on local data and the global model.

[0067] In some embodiments, each user has a task to learn, which is trained on the user's device based on a local dataset. The task learning model for the mth user is θ m .

[0068] The prediction function of the task learned by the mth user on the jth sample is Refers to the input sample After model θ m The predicted value of the output. The goal of the mth user learning task is to minimize the loss function, which is expressed as follows:

[0069]

[0070] S3. Local convergence rate calculation: The user uploads the local convergence rate of local training to the server.

[0071] The user first uploads the local convergence speed of local training to the server. Based on the local convergence speed of each user received, the server adaptively calculates and adjusts the participation rate of this round of communication, the selected users, and the parameters of the user's next round of local training. While minimizing the loss of accuracy, it effectively controls the communication frequency between the user and the server, balances the progress of each user's local training, and allows the global model to cover the entire data set as much as possible, thereby improving the model's accuracy and accelerating the convergence speed.

[0072] For example, the local convergence rate of local training is defined by the change in user loss value during the K rounds of local training before communicating with the server. The local convergence rate v of the local training of the mth user is m for:

[0073]

[0074] Where l(K / 2) is the loss value of the user in the (K / 2)th iteration, and l(K) is the loss value of user m in the Kth iteration.

[0075] The user uploads the obtained local convergence speed to the server.

[0076] S4. Calculate the average speed of all users: The server adaptively calculates and adjusts the participation rate of this round of communication, the selected users, and the parameters of the user's next round of local training based on the local convergence speed, and sends the calculated results to all users.

[0077] In some embodiments, the server receives the local convergence rate v of the local training uploaded by all users. m Finally, calculate the average speed of all users' s-th round of communication:

[0078]

[0079] Where M is the total number of users;

[0080] By comparing the average speed of this round with the average speed of the previous round, the participation rate β of this round is determined. When the convergence speed is faster than that of the previous round, the user participation rate of this round can be halved to reduce the number of user communications. Otherwise, the participation rate can be maintained and the overall model convergence can be accelerated.

[0081] After determining the participation rate, the server arranges each user in descending order according to the local convergence speed of their local training, and selects the fastest users to participate in this round of aggregation, ultimately allowing the model to achieve a specific accuracy or loss value in fewer communication rounds.

[0082] S5. Global model update and re-broadcast: The selected users will upload their local training models to the server for aggregation. After aggregation, the server will send the global model to all users in the form of broadcast.

[0083] As an implementation manner, step S5 specifically includes the following steps:

[0084] S51, before the server receives the model uploaded by the user for aggregation, it calculates the local convergence speed v m Calculate the aggregation ratio α for each user i :

[0085]

[0086] S52. Let the faster user models account for a larger proportion in the aggregation process, and the slower user models account for a smaller proportion, and then aggregate the models according to this proportion:

[0087]

[0088] S6. Iterative training and model optimization: During the global model training process, multiple rounds of user-server communication will be carried out, and steps S1-S5 will be repeated. After multiple rounds of optimization, a Parkinson's disease prediction model that meets specific conditions is formed.

[0089] S7. Disease prediction: Use the trained global model to predict new patient data to achieve early prediction of Parkinson's disease, thereby assisting doctors in developing personalized treatment plans.

[0090] In summary, the method of the present invention has at least the following advantages and beneficial effects compared to the prior art:

[0091] (1) The present invention introduces a dynamic adjustment strategy for model aggregation, which adaptively selects user models for aggregation based on the weights of different users and various parameters. This optimization strategy can better adapt to the heterogeneity of data, thereby improving the accuracy and reliability of the generated model for Parkinson's prediction.

[0092] (2) The present invention adopts an adaptive user selection strategy to select users for the next round of training based on the training speed of the previous round, which greatly improves the generation efficiency of the model.

[0093] Example 2

[0094] This embodiment provides a Parkinson's disease prediction system based on adaptive federated learning, including:

[0095] The parameter distribution module is used to initialize the global model trained by the server and broadcast the global model to all users participating in federated learning;

[0096] The local training module is used by users to perform local training based on local data and the global model after receiving the global model;

[0097] The speed upload module is used by users to upload the local convergence speed of local training to the server;

[0098] The speed calculation module is used by the server to adaptively calculate and adjust the participation rate of the current round of communication, the selected users, and the parameters of the user's next round of local training based on the local convergence speed, and send the calculated results to all users;

[0099] The model aggregation module is used by selected users to upload their locally trained models to the server for aggregation. After aggregation, the server broadcasts the global model to all users.

[0100] The model iteration module is used for global model training, which involves multiple rounds of communication between the user and the server. After each round of communication, the parameters are returned to the parameter distribution module. After multiple rounds of optimization, a Parkinson's disease prediction model that meets specific conditions is formed.

[0101] Model prediction is used to use the trained global model to predict new patient data and achieve early prediction of Parkinson's disease.

[0102] Since this system is a Parkinson's disease prediction system based on adaptive federated learning in an embodiment of the present invention, and the principle of solving the problem by this system is similar to that of this method, the implementation of this system can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0103] Example 3

[0104] An embodiment of the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the following Figure 1 A Parkinson's disease prediction method based on adaptive federated learning is shown.

[0105] It is understood that the memory may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory may be used to store instructions, programs, codes, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the various method embodiments described above, etc.; the data storage area may store data created based on the use of the server, etc.

[0106] The processor may include one or more processing cores. The processor utilizes various interfaces and circuits to connect various components within the server. It executes various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory, as well as accessing data stored in memory. Optionally, the processor may be implemented using at least one of the following hardware forms: digital signal processing (DSP), field-programmable gate array (FPGA), and programmable logic array (PLA). The processor may integrate one or a combination of a central processing unit (CPU) and a modem. The CPU primarily processes the operating system and application programs, while the modem handles wireless communications. It is understood that the modem may not be integrated into the processor and may be implemented separately via a single chip.

[0107] Since the electronic device is an electronic device corresponding to a Parkinson's disease prediction method based on adaptive federated learning in an embodiment of the present invention, and the principle of solving the problem by the electronic device is similar to that of the method, the implementation of the electronic device can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0108] Example 4

[0109] An embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to implement the following Figure 1 A Parkinson's disease prediction method based on adaptive federated learning is shown.

[0110] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program. The program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0111] Since the storage medium is a storage medium corresponding to a Parkinson's disease prediction method based on adaptive federated learning in an embodiment of the present invention, and the principle of solving the problem by the storage medium is similar to that of the method, the implementation of the storage medium can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0112] Example 5

[0113] In some possible implementations, various aspects of the methods of the embodiments of the present invention may also be implemented in the form of a program product, which includes program code. When the program product is run on a computer device, the program code is used to cause the computer device to perform the steps of a Parkinson's disease prediction method based on adaptive federated learning according to various exemplary embodiments of the present application described above in this specification. The executable computer program code or "code" for performing each embodiment may be written in a high-level programming language such as C, C++, Python, Smalltalk, Java, JavaScript, Visual Basic, Structured Query Language (e.g., Transact-SQL), Perl, or in various other programming languages.

[0114] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0115] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0116] The above embodiments are intended only to illustrate the technical concepts and features of the present invention. Their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. They are not intended to limit the scope of protection of the present invention. Any equivalent changes or modifications made based on the essence of the present invention are intended to be covered by the scope of protection of the present invention.

Claims

1. A Parkinson's disease prediction method based on adaptive federated learning, characterized in that: The following steps are involved: Step 1: The server initializes the trained global model and broadcasts the global model to all users participating in federated learning; Step 2: After receiving the global model, the user performs local training based on local data and the global model; Step 3: The user uploads the local convergence speed of local training to the server; Step 4: The server adaptively calculates and adjusts the participation rate of this round of communication, the selected users, and the parameters of the user's next round of local training based on the local convergence speed, and sends the calculated results to all users; Step 5: The selected users upload their local training models to the server for aggregation. After aggregation, the server sends the global model to all users. Step 6: During the global model training process, multiple rounds of user-server communication will be carried out, and steps 1-5 will be repeated. After multiple rounds of optimization, a Parkinson's disease prediction model that meets specific conditions will be formed; Step 7: Use the trained global model to predict new patient data to achieve early prediction of Parkinson's disease.

2. A Parkinson's disease prediction method based on adaptive federated learning according to claim 1, characterized in that: The step 2 comprises: Each user has a task to learn, which is trained on the user's device based on the local dataset. The task learning model of the mth user is θ m ; The prediction function of the task learned by the mth user on the jth sample is Refers to the input sample After model θ m Output predicted value.

3. The Parkinson's disease prediction method based on adaptive federated learning according to claim 2, characterized in that: The goal of the mth user learning task is to minimize the loss function, which is expressed as follows: Where, l(θ m ) represents the loss function of the model, n m represents the number of data samples, represents the label of the task of the jth sample in the mth user, λ represents the penalty coefficient, ||θ m ||2 represents the l2 regularization term.

4. The Parkinson's disease prediction method based on adaptive federated learning according to claim 1, characterized in that: The step 3 comprises: During the K rounds of local training before communicating with the server, the change in user loss value is defined as the local convergence rate of local training. The local convergence rate v of the local training of the mth user is m for: Where l(K / 2) is the loss value of the user in the (K / 2)th round of iteration, and l(K) is the loss value of user m in the Kth round of iteration; The user uploads the local convergence speed to the server.

5. The Parkinson's disease prediction method based on adaptive federated learning according to claim 1, characterized in that: The step 4 comprises: The local convergence rate v of the server after receiving all users' uploaded local training m Finally, calculate the average speed of all users' s-th round of communication: Where M is the total number of users; By comparing the average speed of this round with the average speed of the previous round, the participation rate β of this round is determined. If the convergence speed is faster than that of the previous round, the participation rate of users in this round is reduced, thereby reducing the number of user communications; otherwise, the participation rate is maintained and the overall model convergence is accelerated. After determining the participation rate, the server arranges each user in descending order according to the local convergence speed of their local training, and selects the fastest users to participate in this round of aggregation, ultimately allowing the model to achieve a specific accuracy or loss value in fewer communication rounds.

6. The Parkinson's disease prediction method based on adaptive federated learning according to claim 1, characterized in that: The step 5 comprises: Before the server receives the model uploaded by the user for aggregation, it calculates the local convergence speed v m Calculate the aggregation ratio α for each user i : Where, v i is the local convergence rate of the i-th selected user, v j is the local convergence rate of all selected users, I is the total number of users participating in communication and aggregation in each round; Let the faster user models account for a larger proportion in the aggregation process, and the slower user models account for a smaller proportion, and then aggregate the models according to this proportion: Where θ is the aggregated model, θ i The model uploaded by the i-th user.

7. A Parkinson's disease prediction system based on adaptive federated learning, characterized in that: include: The parameter distribution module is used to initialize the global model trained by the server and broadcast the global model to all users participating in federated learning; The local training module is used by users to perform local training based on local data and the global model after receiving the global model; The speed upload module is used by users to upload the local convergence speed of local training to the server; The speed calculation module is used by the server to adaptively calculate and adjust the participation rate of the current round of communication, the selected users, and the parameters of the user's next round of local training based on the local convergence speed, and send the calculated results to all users; The model aggregation module is used by selected users to upload their locally trained models to the server for aggregation. After aggregation, the server sends the global model to all users; The model iteration module is used for global model training, which involves multiple rounds of communication between the user and the server. After each round of communication, the parameters are returned to the parameter distribution module. After multiple rounds of optimization, a Parkinson's disease prediction model that meets specific conditions is formed. Model prediction is used to use the trained global model to predict new patient data and achieve early prediction of Parkinson's disease.

8. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set, or an instruction set, and the at least one instruction, the at least one program, the code set, or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The computer program product comprises computer instructions, which are used to perform the method according to any one of claims 1 to 6 when executed by a processor.

Citation Information

Cited By

  • Multi-modal federated learning method and system based on projection adaptation mechanism

    CN121390360A