A multi-label data clustering and segmentation method, system, device and storage medium based on federated learning

Abstract

The present invention discloses a multi-label data clustering and segmentation method, system, device and storage medium based on federated learning. The central server extracts image features from the original multi-label data set and splices them with the multi-label information of the images into feature vectors. The central server performs clustering processing on the multi-label data set according to the feature vectors, and divides the data set into data that follows the distribution type according to the distribution type set by the global parameters and the clustering results and the number of federated clients, and assigns them to each federated client. Image preprocessing is performed on the data of each client, and it is segmented into a training set and a test set, and the configuration information and the division of the training set and the test set are saved, completing the clustering and segmentation of the multi-label data. The present invention simulates multi-label classification tasks for federated learning through a clustering algorithm, provides various data distribution situations required for experiments, and provides an effective solution to the data segmentation problem of processing multi-label data sets in federated learning.

Description

Technical Field

[0001] The present invention relates to the technical field of distributed machine learning, and in particular, to a multi-label data clustering and segmentation method, system, device and storage medium based on federated learning. Background Art

[0002] Federated learning is a distributed machine learning method that allows multiple devices or servers to jointly cooperate in training a model while protecting their respective local data from being shared, aiming to solve problems such as data privacy and security and data silos, and is particularly suitable for scenarios dealing with sensitive information, such as privacy recognition, medical health and other fields, and is a new paradigm in the field of artificial intelligence in recent years.

[0003] However, although federated learning has many advantages in terms of privacy protection and data security, it still has many challenges in practical applications, especially when dealing with non-independent and identically distributed (Non-IID) data and multi-label data. Non-IID data may lead to inconsistent performance of the federated model and low communication efficiency, and it is difficult to optimize. Multi-label data may cause the model to be unable to fully utilize the dependencies between labels, bringing additional challenges to the training and aggregation of the model. Currently, the academic community is actively exploring the method design, optimization strategies and communication protocols of federated learning to achieve more efficient, accurate and reliable distributed learning.

[0004] However, the research of federated learning needs to carry out comparative experiments on data distributions with different characteristics, and there is a lack of data segmentation methods for multi-label data. Therefore, this solution proposes a multi-label clustering data segmentation method based on federated learning, which can successfully simulate various data distribution situations in the multi-label classification task of federated learning to test relevant federated learning methods and meet the needs of federated learning academic research. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the prior art and propose a multi-label data clustering and segmentation method, system, device and storage medium based on federated learning. The present invention distributes data to each client according to several data distribution situations in the multi-label classification task of federated learning for a multi-label image data set, providing a good environment for experimental verification.

[0006] The purpose of the present invention is achieved by the following technical solutions: In the first aspect, the present invention provides a multi-label data clustering and segmentation method based on federated learning, and the method includes the following steps:

[0007] S1: The central server loads the original multi-label data set, extracts image features using a convolutional neural network, and splices them with the original multi-label information of the image into a feature vector;

[0008] S2: The central server performs clustering processing on the feature vectors obtained by splicing the multi-label dataset according to step S1;

[0009] S3: The central server divides the dataset into data following the distribution type based on the clustering results according to the distribution type set by the global parameters and the number of federated clients, and assigns it to each federated client;

[0010] S4: Perform image preprocessing on the data of each federated client, divide it into a training set and a test set, and save the configuration information and the division of the training set and the test set to complete the clustering and segmentation of multi-label data.

[0011] Furthermore, in step S1, the central server loads the original multi-label dataset, inputs it into the pre-trained convolutional neural network ResNet to extract image features, and generates feature vectors for each image.

[0012] Furthermore, in step S1, the multi-label data of the pictures is converted into multi-label vectors in one-hot encoding form and spliced with the feature vectors of the images to comprehensively represent the visual information and semantic information of the images.

[0013] Furthermore, in step S3, according to the global preset parameters, the clustered dataset is divided into subsets with the same number as the participating clients according to five different multi-label distribution methods, and are assigned to each participating client one by one. The five multi-label distribution methods are: balanced and independent and identically distributed, unbalanced and independent and identically distributed, pathological partitioning, simulating real-world situation distribution, and mixed distribution. The mixed distribution is composed of pathological partitioning and simulating real-world situation distribution.

[0014] Furthermore, for the simulating real-world situation distribution, perform Dirichlet segmentation on the multi-label dataset based on the clustering results to simulate the data distribution in the real world.

[0015] Furthermore, for the mixed distribution, divide a certain proportion of the data using the simulating real-world situation distribution and assign it to the corresponding proportion of clients, and divide the remaining data using the mixed pathological partitioning and assign it to the remaining clients. This proportion is set in the global parameters.

[0016] Furthermore, in step S3, perform preprocessing on the images assigned to each client, including operations such as resizing, random cropping, horizontal flipping, and normalization.

[0017] In the second aspect, the present invention also provides a multi-label data clustering and segmentation system based on federated learning. The system includes a feature extraction module, a data clustering module, a data segmentation module, and a configuration storage module;

[0018] The feature extraction module is used for the central server to extract image features from the multi-label dataset and splice them with the original multi-label information of the image into feature vectors;

[0019] The data clustering module is used for the central server to perform clustering processing on the multi-label dataset according to the feature vectors;

[0020] The data segmentation module is used for the central server to divide the dataset into data conforming to the distribution type based on the clustering result and allocate it to each federated client according to the distribution type set by the global parameter and the number of federated clients;

[0021] The configuration storage module is used to perform image preprocessing on the data of each client, divide it into a training set and a test set, and save the configuration information and the division of the training set and the test set to complete the clustering and segmentation of multi-label data.

[0022] In a third aspect, the present invention also provides a multi-label data clustering and segmentation device based on federated learning, including a memory and one or more processors. Executable code is stored in the memory. When the processor executes the executable code, the multi-label data clustering and segmentation method based on federated learning as described above is implemented.

[0023] In a fourth aspect, the present invention also provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, the multi-label data clustering and segmentation method based on federated learning as described above is implemented.

[0024] In a fifth aspect, the present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the multi-label data clustering and segmentation method based on federated learning as described above is implemented.

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] 1. By proposing an original multi-label clustering data segmentation method, the present invention effectively solves the problems of uneven distribution and data heterogeneity faced by the prior art when processing multi-label image datasets. This method not only considers the multi-label characteristics of the dataset, but also optimizes according to the actual needs and number of clients in the federated learning environment, thus providing a more accurate and efficient data preprocessing method for the federated learning multi-label classification task.

[0027] 2. By splitting the dataset and using probability methods such as Dirichlet distribution, the present invention can simulate different data distribution scenarios in the multi-label classification task of federated learning, including but not limited to independent and identically distributed / non-independent and identically distributed and balanced / unbalanced distributions in the multi-label scenario. This flexible processing ability enables the present invention to perform customized data preparation for specific federated learning scenarios, greatly improving the experimental efficiency.

[0028] 3. Before data splitting, the central server first performs clustering on the multi-label dataset. This step helps to reveal the internal structure and relationships of the data and improve the utilization efficiency of the data. Through clustering preprocessing, it can be ensured that the dataset assigned to each client is not only balanced in quantity but also representative in the feature space, thereby enhancing the generalization ability of the federated learning model. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0030] Figure 1 It is the overall step flow chart of the multi-label data clustering and splitting method based on federated learning of the present invention.

[0031] Figure 2 It is the data splitting schematic diagram of the multi-label data clustering and splitting method based on federated learning of the present invention.

[0032] Figure 3 It is the structural diagram of a multi-label data clustering and splitting device based on federated learning of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0033] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0034] Federated learning is a privacy-preserving distributed machine learning technique. Each client independently trains a model using local data and then uploads only the model parameters or model updates to a central server. The central server aggregates this data to update the global model and then sends the improved model back to each client. In this way, federated learning not only ensures performance but also guarantees the data security of each client. The present invention proposes a multi-label clustering data segmentation method based on federated learning, which divides and assigns the images and labels in the dataset to each client according to different distribution types, simulating various data distribution situations that may occur in federated learning for academic experiments.

[0035] The multi-label classification task is a machine learning task. Different from traditional single-label classification tasks, each image can be assigned multiple labels simultaneously, which may result in dependencies between labels and an unbalanced label distribution. This makes it a difficult problem to effectively partition the multi-label dataset when using federated learning to handle the multi-label classification task. A multi-label data clustering segmentation method based on federated learning proposed by the present invention first performs clustering on the multi-label dataset and then assigns the data to each client according to different distribution types based on the clustering results, effectively solving the problem of data allocation when processing the multi-label dataset during the federated learning training process. Moreover, during the clustering process, this method introduces a deep learning method, uses a convolutional neural network to extract features from the multi-label dataset, and incorporates the feature information into the label vector for clustering processing, increasing the dimension of the data.

[0036] The Dirichlet Distribution is a probability distribution defined on a k - 1 dimensional simplex and is used to describe the probability distribution of k categories / events. Simply put, it can be regarded as a generalization of the Beta distribution in a multi-dimensional space and is used to describe the parameters of the multinomial distribution. The Dirichlet distribution is composed of parameters α1, α2,..., α k which, where each α k corresponds to a dimension or category in the distribution, K is the total number of categories, and these parameters are usually called concentration parameters, which affect the shape of the distribution. When all α k > 1, the distribution tends to be more uniform; when α k < 1, the distribution tends to be more concentrated in one or several dimensions. The Dirichlet distribution has strong flexibility and can simulate various differently shaped distributions by adjusting the concentration parameters, and the values of the concentration parameters have an intuitive interpretation. Therefore, the present invention uses the Dirichlet distribution to simulate the non-independent and identically distributed situation in reality.

[0037] Please combine Figure 1 、 Figure 2As shown, the present invention provides a multi-label data clustering and segmentation method based on federated learning, comprising the following steps:

[0038] S1: The central server loads the original multi-label dataset, extracts image features using a convolutional neural network, and concatenates them with the original multi-label information of the image to form a feature vector, comprehensively representing the visual and semantic information of the image.

[0039] In step S1, taking the dataset VPA (a privacy image dataset, each picture corresponding to multiple privacy labels, 2017) as an example, the central server first receives a multi-label dataset, which contains a large number of images and their corresponding multiple labels. It reads the index file (stored in CSV format) in the dataset. The index file includes the picture file name and its corresponding label number, and reads the corresponding picture in the file system according to the file name. Then, the picture is cropped to a unified size and input into the pre-trained ResNet-18 model for image feature extraction. The last fully connected layer is removed, only the feature extraction part is retained, and the features are flattened into a one-dimensional vector, which is used as the image feature vector. In addition, for the label number of each picture, binary conversion is performed using one-hot encoding to obtain the image label vector. Finally, the image feature vector and the image label vector are concatenated and fused into a larger feature vector to expand the image data dimension for convenient clustering processing.

[0040] S2: The central server performs clustering processing on the multi-label dataset using the K-Means algorithm based on the feature vector obtained by concatenation in step S1.

[0041] In step S2, the central server needs to perform clustering processing on the multi-label dataset to divide data subsets for subsequent federated learning tasks. For this purpose, the central server first receives the label feature vectors extracted for each sample in step S1, and these vectors will be used as the basis for clustering analysis. Considering the characteristics of federated learning, the present invention determines the number of clusters K according to the number of clients participating in federated learning to ensure that each client can receive sufficient and diverse data, thereby improving the model training effect. Subsequently, the central server executes the K-Means clustering algorithm to divide the label feature vectors of all samples into K different clusters. Each cluster represents a specific region in the label space and contains samples with similar label features. According to the clustering result, the central server divides the original multi-label dataset into K non-repeating data subsets, and each subset corresponds to a specific cluster. Finally, the central server stores the clustering result as a CSV file and adds a column of "clustering category" data to the original index file to record the cluster to which each sample belongs, facilitating subsequent data allocation and the federated learning training process.

[0042] S3: The central server divides the dataset according to the clustering results and distributes it to each client according to different distribution types set by the parameters.

[0043] In step S3, the central server performs a data splitting operation on the multi-label dataset after label clustering according to the parameters preset by the user. It is split into several subsets with the same number as the number of clients participating in the federated learning according to different multi-label distribution methods, and distributed to each participating client to simulate several experimental scenarios of multi-label classification in federated learning, such as Figure 2 As shown, "niid", "balance", and "partition" are parameters set by the user when running the code, used to split the multi-label dataset into 5 different data distributions. Among them, "niid" indicates whether to use non-independent and identically distributed, "balance" indicates whether the data volume is balanced among clients, and "partition" indicates the specific distribution specified. The 5 distribution types and the corresponding parameters are as follows:

[0044] 1. Balanced and independent and identically distributed (parameters: niid = False, balance = True)

[0045] 2. Unbalanced and independent and identically distributed (parameters: niid = False, balance = False)

[0046] 3. Ill-conditioned partition (parameters: niid = True, balance = False, partition = 'cpat')

[0047] 4. Simulate real-world distribution (parameters: niid = True, balance = False, partition = 'cdir')

[0048] 5. Mixed distribution (parameters: niid = True, balance = False, partition = 'chaos')

[0049] For balanced and independent and identically distributed, this is the most ideal data distribution scenario. All clients will obtain all the clustered data, ensuring the homogeneity (independent and identically distributed) of the data distribution, and each client obtains roughly the same samples from each cluster to achieve the balance of the data volume. This setting is used to simulate an ideal federated learning scenario where the data is evenly distributed among clients. It helps the model learn effective feature representations, thereby improving the classification accuracy and the generalization ability of the model.

[0050] For unbalanced and independent and identically distributed data, all clients will still receive data from all clusters. The difference is that the number of samples each client obtains from each cluster is random, but still within a reasonable range to ensure that each client receives at least a certain number of samples. This setting simulates a more realistic independent and identically distributed scenario where the data is evenly distributed among clients, but the amount of data may vary. The data is independent and identically distributed among clients, but there is an imbalance in the labels, that is, the number of samples of some labels is much more than that of other labels, resulting in a weaker recognition ability of the model for labels with fewer samples and reducing the classification performance.

[0051] For pathological partitioning, each client only receives data from some clusters, and a user-defined parameter is used to control the number of clusters each client receives, resulting in heterogeneity (non-independent and identically distributed) in the data distribution. Moreover, each client obtains a random number of samples from the clusters it is assigned, and each client may have data for only a very small number of labels, and the label distribution is unbalanced, further increasing the heterogeneity of the data volume. This setting is used to simulate a non-independent and identically distributed scenario where there are significant differences in data among clients, such as different types of data collected by different devices. This is the data distribution that has the greatest impact on the model performance. The model may overfit to the minority classes on most clients and it is difficult to learn a general feature representation, which will greatly reduce the model's performance and generalization ability.

[0052] For the distribution simulating real-world situations, different from the previous three cases, the Dirichlet Distribution is used to assign the data points in each cluster to each client. This results in each client receiving data from all clusters, but the number of samples for each cluster is randomly determined according to the Dirichlet Distribution, thus forming a more complex non-independent and identically distributed situation, which is the data distribution closest to the real-world situation. This setting is suitable for simulating a non-independent and identically distributed scenario with more diverse data distribution, where the data differences among clients may be more subtle and complex, and the label distribution is unbalanced, reflecting the complexity of the real world. Thus, it better simulates the federated data distribution in real-world situations and helps to research and develop more robust federated learning methods.

[0053] For the mixed distribution, in order to more comprehensively simulate the data heterogeneity in real-world federated learning, the mixed distribution combines the characteristics of pathological partitioning and the distribution simulating real-world scenarios. Specifically, the mixed distribution first divides the multi-label dataset into two parts according to the ratio α set by the user. The part with the ratio of α adopts the pathological partitioning strategy, where each client only receives data from specific clusters, and the number of clusters and the number of samples are controlled by the parameters set by the user. This extreme partitioning method will significantly increase the heterogeneity of the data distribution and may result in the situation that some clients are extremely lacking in data of specific labels, simulating the scenario where the data types collected by different devices are significantly different. The remaining part with the ratio of 1 - α adopts the distribution strategy simulating real-world scenarios based on the Dirichlet distribution. Each client obtains data from all clusters, but the number of samples in each cluster is randomly determined by the Dirichlet distribution, making the data distribution more complex and closer to the real situation, simulating the scenario where the data distribution is diverse and the differences between clients are more subtle. By adjusting the ratio α, the mixed distribution can flexibly control the degree of pathological partitioning and the distribution simulating real-world scenarios, thereby simulating a wider range of non-independent and identically distributed scenarios, better reflecting the complexity of the federated learning data distribution in the real world, and helping researchers develop more robust and generalizable federated learning algorithms.

[0054] S4: Perform image preprocessing on the data of each client, divide it into a training set and a test set according to a certain ratio, and save the configuration information and the division of the training set and the test set.

[0055] In step S4, each federated client performs image preprocessing and data augmentation on the specific images according to the dataset assigned to it, including operations such as adjusting pixels, cropping, horizontal flipping, format conversion, normalization, etc., and divides it into a training set and a test set for each client according to the ratio of 3:1. Finally, save the configuration information as a CSV file and save the dataset of each client as an NPZ file for subsequent experiments.

[0056] Corresponding to the foregoing embodiment of a multi-label data clustering and segmentation method based on federated learning, the present invention also provides an embodiment of a multi-label data clustering and segmentation system based on federated learning. The system includes a feature extraction module, a data clustering module, a data segmentation module, and a configuration storage module; for the implementation process of each module, please refer to the specific steps of the foregoing embodiment of the multi-label data clustering and segmentation method.

[0057] The feature extraction module is used for the central server to extract the image features in the multi-label dataset and splice them into feature vectors with the image labels.

[0058] The data clustering module is used for the central server to load the original multi-label dataset and perform clustering processing on the multi-label dataset according to the feature vectors;

[0059] The data segmentation module is used for the central server to segment the dataset to each federated client according to the clustering result according to different distribution types set by each federated client parameter;

[0060] The configuration storage module is used to perform image preprocessing on the data of each client, segment it into a training set and a test set, and save the configuration information and the division of the training set and the test set to complete the multi-label data clustering segmentation.

[0061] Corresponding to the foregoing embodiment of a multi-label data clustering segmentation method based on federated learning, the present invention also provides an embodiment of a multi-label data clustering segmentation device based on federated learning.

[0062] See Figure 3 , an embodiment of a multi-label data clustering segmentation device based on federated learning provided by the embodiment of the present invention includes a memory and one or more processors. Executable code is stored in the memory. When the processor executes the executable code, it is used to implement a multi-label data clustering segmentation method in the foregoing embodiment.

[0063] An embodiment of a multi-label data clustering segmentation device based on federated learning provided by the present invention can be applied to any device with data processing capabilities. The any device with data processing capabilities can be a device or apparatus such as a computer. The device embodiment can be implemented by software, or by hardware or a combination of software and hardware. Taking software implementation as an example, as a logically meaningful device, it is formed by the processor of any device with data processing capabilities where it is located reading the corresponding computer program instructions in the non-volatile memory into the memory for operation. From the hardware level, as Figure 3 shown, it is a hardware structure diagram of any device with data processing capabilities where a multi-label data clustering segmentation device based on federated learning provided by the present invention is located. Except for Figure 3 the processor, memory, network interface, and non-volatile memory shown, any device with data processing capabilities where the device in the embodiment is located usually also includes other hardware according to the actual functions of the any device with data processing capabilities, which will not be elaborated here.

[0064] The implementation processes of the functions and roles of each unit in the above device are specifically detailed in the implementation processes of the corresponding steps in the above method, which will not be elaborated here.

[0065] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the descriptions of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of the present invention. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0066] An embodiment of the present invention also provides a computer-readable storage medium, on which a program is stored. When the program is executed by a processor, it implements a multi-label data clustering and segmentation method based on federated learning in the above embodiments.

[0067] The computer-readable storage medium may be an internal storage unit of any device with data processing capabilities described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium may also be an external storage device of any device with data processing capabilities, such as a plug-in hard disk, a Smart Media Card (SMC), an SD card, a Flash Card, etc. equipped on the device. Further, the computer-readable storage medium may also include both an internal storage unit and an external storage device of any device with data processing capabilities. The computer-readable storage medium is used to store the computer program and other programs and data required by any device with data processing capabilities, and can also be used to temporarily store the data that has been output or will be output.

[0068] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the multi-label data clustering and segmentation method based on federated learning described above.

[0069] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-label data clustering and segmentation method based on federated learning, characterized in that The method includes the following steps: S1: The central server loads the original multi-label dataset, extracts image features using a convolutional neural network, and concatenates them with the original multi-label information of the images to form feature vectors; S2: The central server performs clustering on the multi-label dataset based on the feature vectors obtained in step S1; S3: The central server divides the dataset into data conforming to the distribution type based on the clustering results and assigns it to each federated client according to the distribution type set by the global parameters and the number of federated clients; specifically: according to the global preset parameters, the clustered dataset is divided into subsets with the same number as the participating clients according to five different multi-label distribution methods, and each is assigned to each participating client. The five multi-label distribution methods are: balanced and independent and identically distributed, unbalanced and independent and identically distributed, pathological partitioning, simulating real-world distribution, and mixed distribution, where the mixed distribution is composed of pathological partitioning and simulating real-world distribution; S4: Preprocess the images of each federated client's data, divide them into a training set and a test set, and save the configuration information and the division of the training set and the test set to complete the clustering and segmentation of multi-label data.

2. The multi-label data clustering and segmentation method based on federated learning according to claim 1, wherein, In step S1, the central server loads the original multi-label dataset, inputs it into the pre-trained convolutional neural network ResNet to extract image features, and generates feature vectors for each image.

3. The multi-label data clustering and segmentation method based on federated learning according to claim 2, wherein In step S1, the multi-label data of the images is converted into a multi-label vector in one-hot encoding form and concatenated with the feature vectors of the images to comprehensively represent the visual and semantic information of the images.

4. The multi-label data clustering and segmentation method based on federated learning according to claim 1, characterized in that For the simulating real-world distribution, perform Dirichlet segmentation on the multi-label dataset based on the clustering results to simulate the data distribution in the real world.

5. The multi-label data clustering and segmentation method based on federated learning according to claim 1, wherein For the mixed distribution, divide and assign a certain proportion of the data according to the simulating real-world distribution to the corresponding proportion of clients, and divide and assign the remaining data according to the mixed pathological partitioning to the remaining clients, and this proportion is set in the global parameters.

6. The multi-label data clustering and segmentation method based on federated learning according to claim 1, wherein In step S3, preprocess the images assigned to each client, including operations such as resizing, random cropping, horizontal flipping, and normalization.

7. A multi-label data clustering and segmentation system based on federated learning, characterized in that, The system includes a feature extraction module, a data clustering module, a data segmentation module, and a configuration storage module; The feature extraction module is used for the central server to extract image features from the multi-label dataset and concatenate them with the original multi-label information of the images to form feature vectors; The data clustering module is used for the central server to perform clustering on the multi-label dataset based on the feature vectors; The data segmentation module is used for the central server to divide the dataset into data conforming to the distribution type based on the clustering results and assign it to each federated client according to the distribution type set by the global parameters and the number of federated clients; specifically: according to the global preset parameters, the clustered dataset is divided into subsets with the same number as the participating clients according to five different multi-label distribution methods, and each is assigned to each participating client. The five multi-label distribution methods are: balanced and independent and identically distributed, unbalanced and independent and identically distributed, pathological partitioning, simulating real-world distribution, and mixed distribution, where the mixed distribution is composed of pathological partitioning and simulating real-world distribution; The configuration storage module is used to perform image preprocessing on the data of each client, segment it into a training set and a test set, and save the configuration information and the division of the training set and the test set, so as to complete the multi-label data clustering segmentation.

8. A multi-label data clustering and segmentation device based on federated learning, comprising a memory and one or more processors, wherein executable code is stored in the memory, and is characterized in that, When the processor executes the executable code, it implements a multi-label data clustering segmentation method based on federated learning as described in any one of claims 1-6.

9. A computer-readable storage medium having a program stored thereon, characterized in that, When the program is executed by the processor, it implements a multi-label data clustering segmentation method based on federated learning as described in any one of claims 1-6.

Images