Model parameter updating method and system based on graph federal learning

Abstract

The invention discloses a model parameter updating method and system based on graph federated learning, and relates to the field of federated learning and graph neural networks, and the method comprises the steps: completing the training of a graph convolutional neural network through employing a federated learning frame, and enabling the training to be capable of training a model on a graph distributed on a plurality of clients; a neighbor feature aggregation function is used, so that the clients do not need to communicate in each training iteration, and the communication burden is remarkably reduced; an aggregation mechanism based on an attention mechanism is used, a server is allowed to dynamically allocate weights for each local model participating in training, and the robustness and accuracy of the model are improved.

Description

Technical Field

[0001] The present application belongs to the field of mobile communication technology, and in particular relates to a model parameter updating method and system based on graph federated learning. Background Art

[0002] At present, in order to ensure the security of data privacy of participants in the joint training model, federated learning technology has emerged. Federated learning is a promising distributed collaborative machine learning paradigm that aims to solve the data isolation problem. It allows local models to promote each other while ensuring the privacy of local data. In addition, the learning of personalized federated learning models under non-independent and identically distributed conditions has also received widespread attention. In federated learning, participants only share model parameters or embedded features to avoid leaking sensitive data. This architecture combines multiple encryption technologies to effectively protect the privacy of local data. At the same time, there may be relationships between federated learning participants, and the similarities and trust information of participants can be revealed through the constructed graph topology. It is still challenging to use this topological information to improve the performance of federated learning. Therefore, combining federated learning with graph machine learning is a promising solution.

[0003] The challenges currently faced by graph federated learning are as follows: In a federated learning environment, data is segmented and stored on multiple clients participating in training. The interactions and connections between these data are usually reflected through cross-client edges, which play a vital role in graph neural networks. They not only connect the graph nodes of data stored on different clients, but also capture the complex relationships and dependencies between graph nodes, which is crucial for the accuracy and robustness of the model. However, traditional federated learning methods often do not fully consider the existence and importance of these cross-client edges, resulting in the model's inability to fully capture global structural information when utilizing graph data, thus affecting the performance of the model.

[0004] At the same time, federated learning systems face special challenges when processing non-Euclidean spatial data, which often contain complex graph structures, such as social networks, bioinformatics networks, etc. In order to improve the effectiveness of federated learning on such data, researchers have been exploring more advanced aggregation strategies. However, traditional federated learning methods often adopt a simple average aggregation strategy, which does not fully consider the uniqueness and quality differences of each local model, and may lead to a decrease in the performance of the aggregated global model. Summary of the invention

[0005] In order to solve the above-mentioned problem that cross-client edges are easily ignored and the traditional federated learning method has poor aggregation effect, the present invention proposes a model parameter updating method and system based on graph federated learning.

[0006] In a first aspect of the present invention, the present invention proposes a model parameter updating method based on graph federated learning, which is applied to a graph federated learning server, and the method comprises the following steps:

[0007] Initialize the global model parameters and broadcast them to each client;

[0008] Receive local model parameters after training from the client;

[0009] Evaluate the local model parameters of each client to obtain a first confidence level of the local model parameters of each client; the first confidence level of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client;

[0010] Perform quantum state encoding on the local model parameters of each client to obtain the quantum state model parameters of each client;

[0011] The quantum state model parameters of each client are operated through a quantum self-attention circuit to obtain a second confidence level of the local model parameters of each client; the second confidence level of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients;

[0012] updating the global model parameters of the server according to the first confidence and the second confidence of the local model parameters of each client;

[0013] Broadcast the updated global model parameters to all clients;

[0014] If the global model parameters converge, the update ends, otherwise the above process is repeated.

[0015] In a second aspect of the present invention, the present invention proposes a model parameter updating method based on graph federated learning, which is applied to a client of graph federated learning. The method comprises the following steps:

[0016] Receive global model parameters from a server; the global model parameters are updated based on the first confidence and the second confidence of the local model parameters of each client; the first confidence of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client; the second confidence of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients, and the similarity between the local model parameters of each client and the local model parameters of other clients is obtained by calculating the quantum state model parameters of each client through a quantum self-attention circuit; the quantum state model parameters of each client are obtained by quantum state encoding the local model parameters of each client;

[0017] Constructing a graph convolutional neural network according to the global model parameters and multi-hop neighbor aggregation features;

[0018] Training the global model parameters to obtain updated local model parameters;

[0019] Sending updated local model parameters to the server;

[0020] If no global model parameters are received from the server within a preset time, the updating is terminated, otherwise the above process is repeated; the preset time is used to indicate the convergence time of the global model parameters.

[0021] In a third aspect of the present invention, the present invention proposes a model parameter updating system based on graph federated learning, which is deployed on a server participating in graph federated learning. The model parameter updating system based on graph federated learning includes:

[0022] Processor; and

[0023] a memory arranged to store computer executable instructions which, when executed, cause the processor to:

[0024] Initialize the global model parameters and broadcast them to each client;

[0025] Receive local model parameters after training from the client;

[0026] Evaluate the local model parameters of each client to obtain a first confidence level of the local model parameters of each client; the first confidence level of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client;

[0027] Perform quantum state encoding on the local model parameters of each client to obtain the quantum state model parameters of each client;

[0028] The quantum state model parameters of each client are operated through a quantum self-attention circuit to obtain a second confidence level of the local model parameters of each client; the second confidence level of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients;

[0029] updating the global model parameters of the server according to the first confidence and the second confidence of the local model parameters of each client;

[0030] Broadcast the updated global model parameters to all clients;

[0031] If the global model parameters converge, the update ends, otherwise the above process is repeated.

[0032] In a fourth aspect of the present invention, the present invention proposes a model parameter updating system based on graph federated learning, which is deployed on a client participating in graph federated learning. The model parameter updating system based on graph federated learning includes:

[0033] Processor; and

[0034] a memory arranged to store computer executable instructions which, when executed, cause the processor to:

[0035] Receive global model parameters from a server; the global model parameters are updated based on the first confidence and the second confidence of the local model parameters of each client; the first confidence of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client; the second confidence of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients, and the similarity between the local model parameters of each client and the local model parameters of other clients is obtained by calculating the quantum state model parameters of each client through a quantum self-attention circuit; the quantum state model parameters of each client are obtained by quantum state encoding the local model parameters of each client;

[0036] Constructing a graph convolutional neural network according to the global model parameters and multi-hop neighbor aggregation features;

[0037] Training the global model parameters to obtain updated local model parameters;

[0038] Sending updated local model parameters to the server;

[0039] If no global model parameters are received from the server within a preset time, the updating is terminated, otherwise the above process is repeated; the preset time is used to indicate the convergence time of the global model parameters.

[0040] Beneficial effects of the present invention:

[0041] In a federated graph convolutional neural network, in order to maintain the accuracy of the model, it is necessary to continuously exchange information about node neighbors during the training process, which results in significant communication overhead. The present invention significantly reduces the communication cost by only performing a limited number of communications in the pre-training step. The present invention also uses an attention-based aggregation mechanism that can dynamically assign weights to each local model involved in the training. This attention-based approach not only takes into account the individual differences of local models, but also adjusts their influence in the aggregation process according to their contribution to the global model. In this way, the present invention can identify and reduce the negative impact of local models with poor performance or uneven data distribution on the global model. This not only improves the fault tolerance of the model when facing clients with uneven data quality, but also enhances the accuracy of the model under various data distributions. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of a model parameter updating method based on graph federated learning according to an embodiment of the present invention;

[0043] Figure 2 is a flow chart of a method for updating model parameters in a pre-training phase according to an embodiment of the present invention;

[0044] Figure 3 is a flow chart of a model parameter updating method based on graph federated learning according to another embodiment of the present invention;

[0045] Figure 4 is a flow chart of a method for updating model parameters in a pre-training phase according to another embodiment of the present invention;

[0046] Figure 5 It is a signaling diagram of a model parameter updating method based on graph federated learning according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] The embodiments of this specification provide a method for updating model parameters based on graph federated learning. The executor of the method may be a participating device of graph federated learning. The participating device of graph federated learning may be a client or a server, wherein the server may be an independent server or a server cluster composed of multiple servers. The client may be a personal computer or a terminal device with display function such as a smart phone, smart TV, tablet computer, portable computer, etc.

[0049] Figure 1 is a flow chart of a model parameter updating method based on graph federated learning according to an embodiment of the present invention. Figure 1 As shown, the method is applied to a server for graph federated learning, and comprises the following steps:

[0050] 101. Initialize the global model parameters and broadcast them to each client;

[0051] In some embodiments, at the beginning of training, the server initializes the global model parameters And broadcast it to all clients under graph federated learning, where the superscript t represents the iteration round, initially t = 0, and the subscript g is used to indicate the global model of the server.

[0052] 102. Receive local model parameters after training from the client;

[0053] In some embodiments, after entering the training, the server receives the local model parameter w returned by the client. k , where the subscript k is used to indicate the local model of the kth client.

[0054] 103. Evaluate the local model parameters of each client to obtain a first confidence level of the local model parameters of each client; the first confidence level of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client;

[0055] In some embodiments, evaluating the local model parameters of each client to obtain the first confidence level of the local model parameters of each client includes:

[0056] The local model parameters of each client are measured similarly with the global model parameters of the server to obtain the similarity distance between the local model parameters of each client and the global model parameters of the server;

[0057] The similarity distance between the local model parameters of each client and the global model parameters of the server is processed by attention using a standardized function to obtain the first confidence of the local model parameters of each client.

[0058] For example, assuming is the local model w of the k-th client in layer l k The trainable parameters of is the l-layer global model w g The trainable parameters of The similarity distance between the local model parameters of the kth client and the global model parameters of the server at layer l can be obtained. It can further characterize the attention coefficient of the local model parameters of the kth client to the global model parameters of the server at layer l; in order to make the coefficients in the parameters of each layer easy to use and not affect the convergence speed due to the possible order of magnitude change of the parameters, the SoftMax function is used to standardize the attention coefficient, so as to obtain the first confidence of the local model parameters of the kth client The superscript L represents the total number of layers, and the subscript k represents the kth client.

[0059] 104. Perform quantum state encoding on the local model parameters of each client to obtain the quantum state model parameters of each client;

[0060] In some embodiments, the server sends each client's local model parameter w k Encoded as quantum states Specifically expressed as:

[0061]

[0062] Where H represents the Hadamard gate, Represents a tensor operator, U enc represents quantum amplitude coding, and n represents the number of clients.

[0063] 105. Calculate the quantum state model parameters of each client through a quantum self-attention circuit to obtain a second confidence level of the local model parameters of each client; the second confidence level of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients;

[0064] In some embodiments, the process of calculating the similarity of the local model parameters between each client includes:

[0065] The quantum state model parameters of all clients are subjected to corresponding parameter evolution respectively, and the Z direction is measured according to the Pauli-Z gate to obtain the first measurement expectation and the second measurement expectation corresponding to the quantum state model parameters of each client;

[0066] According to the difference between the first measurement expectation and the second measurement expectation corresponding to the quantum state model parameters of each client, the quantum attention matrix of the corresponding client is obtained;

[0067] According to the quantum attention matrix of each client, the similarity of the local model parameters between each client and other clients is calculated;

[0068] According to the similarity between the local model parameters of each client and other clients, a second confidence level of the local model parameters of each client is calculated.

[0069] In some embodiments, the server transmits the quantum states of all clients to The server sends the quantum state of all clients to the quantum self-attention circuit for calculation. The quantum self-attention circuit is sent to calculate the parameter similarity between clients, and then the second confidence of the local model parameters of each client is obtained based on the similarity between each client and other clients.

[0070] Exemplarily, the encoded quantum states are respectively evolved through corresponding parameter circuits, and finally the measurement expectation is obtained by performing the measurement in the Z direction according to the Pauli-Z gate, which is specifically expressed as:

[0071]

[0072]

[0073] Among them, U K and U Q All are quantum circuits, U K and U Q In addition to the parameters θ K and θ Q Otherwise, the circuit is the same. <Z K > and <Z Q >respectively represent the measurement expectation obtained by performing measurement in the Z direction according to the Pauli-Z gate. Represents the conjugate transpose of a matrix. Z represents the Pauli-Z gate.

[0074] Based on the first measurement expectation and the second measurement expectation obtained by the above calculation formula, the corresponding attention matrix is ​​obtained, which is specifically expressed as:

[0075]

[0076] According to the obtained attention matrix, further calculation is performed to obtain the attention similarity of any l of each client k, and then the original similarity value is mapped to a probability distribution with a sum of 1 through the softmax function, which is specifically expressed as:

[0077]

[0078] It is understandable that the computational cost of the traditional self-attention mechanism rises sharply with the increase in the number of parameters, while the self-attention mechanism implemented using quantum circuits can more effectively manage the computational requirements of a large number of parameters with the help of the parallel processing capabilities of quantum computing, and more efficiently cope with the computational challenges of large-scale parameters, thereby optimizing the use of computing resources. The embodiment of the present invention can dynamically assign weights to each local model involved in the training. This self-attention-based method not only takes into account the individual differences of local models, but also adjusts their influence in the aggregation process according to their contribution to the global model. In this way, the negative impact of local models with poor performance or uneven data distribution on the global model can be identified and reduced. This not only improves the fault tolerance of the model when facing clients with uneven data quality, but also enhances the accuracy of the model under various data distributions.

[0079] 106. Update the global model parameters of the server according to the first confidence and the second confidence of the local model parameters of each client;

[0080] In some embodiments, the server updates the global model based on a first confidence level determined by the similarity of model parameters between the client and the server, and a second confidence level determined by the similarity of model parameters between the clients.

[0081] Exemplarily, the update rule of the global model is specifically expressed as:

[0082]

[0083] Among them, p and q are adjustable hyperparameters. represents the first confidence of the local model parameters of the k-th client at layer l during the t-th iteration, Represents the second confidence of the local model parameters of the kth client at layer l during the tth iteration; these two confidences fuse the information of the server and other clients. This method can dynamically assign weights to the local models of each client participating in the training, not only considering the individual differences of each local model, but also adjusting its influence in the aggregation process according to its contribution to the global model. In this way, the negative impact of local models with poor performance or uneven data distribution on the global model can be identified and reduced. This not only improves the fault tolerance of the model when facing clients with uneven data quality, but also enhances the accuracy of the model under various data distributions. .

[0084] 107. Broadcast the updated global model parameters to all clients;

[0085] In some embodiments, the updated global model parameters Broadcast to all clients, the global model parameters can be Transformed into Complete a round of iterative training.

[0086] 108. If the global model parameters converge, the update is terminated, otherwise the above process is repeated.

[0087] In some embodiments, it is determined whether the global model has converged. If not, steps 101-107 are repeated multiple times. If the global model has converged, the training is terminated and the updating of the model parameters is stopped.

[0088] In some preferred embodiments, Figure 2 As shown, the global model parameters are initialized and broadcast to each client, and the following steps are also included:

[0089] 111. Receive the neighbor feature ciphertext of the local map node from the client;

[0090] 121. Aggregate the neighbor feature ciphertexts of the local map nodes of all clients to obtain the multi-hop neighbor feature ciphertexts of the local map nodes;

[0091] In some embodiments, after receiving the ciphertext of neighbor features of all local graph nodes from the client, the server performs a neighbor aggregation operation, which can be expressed as:

[0092]

[0093] Where c(j) represents the index of the client containing graph node j, is the indicator function. When z is equal to c(j), that is, when client z contains the information of graph node j, the indicator function The value of is 1; otherwise, the value is 0. is the encryption operator, Represents the neighbor nodes of graph node i. The server operates on all received neighbor feature arrays and accumulates the multi-hop neighbor features of each graph node.

[0094] 131. Send the multi-hop neighbor feature ciphertext of the local map node to the client.

[0095] This embodiment takes into account that in the federated graph convolutional neural network, in order to maintain the accuracy of the model, it is necessary to continuously exchange node neighbor information during the training process, which leads to significant communication overhead. This embodiment adds the pre-training steps of steps 111-131 before updating the model parameters, and only requires a limited number of communications, which significantly reduces the communication cost and saves the communication overhead.

[0096] Figure 3 is a flow chart of a model parameter updating method based on graph federated learning according to an embodiment of the present invention. Figure 3As shown, the method is applied to a client of graph federated learning, and includes the following steps:

[0097] 201. Receive global model parameters from a server; the global model parameters are updated based on the first confidence and the second confidence of the local model parameters of each client; the first confidence of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client; the second confidence of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients, and the similarity between the local model parameters of each client and the local model parameters of other clients is obtained by calculating the quantum state model parameters of each client through a quantum self-attention circuit; the quantum state model parameters of each client are obtained by quantum state encoding the local model parameters of each client;

[0098] In some embodiments, during the training round, the client continuously receives the global model parameters sent by the server in the iteration round. The superscript t represents the iteration round, initially t=0, and the subscript g is used to indicate the global model of the server.

[0099] 202. Construct a graph convolutional neural network according to the global model parameters and multi-hop neighbor aggregation features;

[0100] In some embodiments, constructing a graph convolutional neural network according to the global model parameters and multi-hop neighbor aggregation features includes:

[0101] Construct the graph node embedding vector of each convolutional layer based on the multi-hop neighbor aggregation features;

[0102] A graph convolutional neural network is constructed based on the graph node embedding vectors of each convolutional layer and the global model parameters.

[0103] Exemplarily, a multi-layer graph convolutional neural network consists of multiple convolutional layers, each of which builds node embedding by aggregating the features of adjacent nodes and follows the propagation rule:

[0104]

[0105] Where W (l) is the weight of each layer of the convolutional neural network to be learned, A ij is the weighted adjacency matrix, φ is the activation function, is the embedding of each neighbor j in the previous layer. The softmax function is used as the activation function of the last layer, which is specifically expressed as:

[0106]

[0107] Among them, zi is the i-th component of the input, and e is the base of the natural logarithm. This allows the output to be interpreted as the probability of a node in each class. ReLU is used as the activation function in the first few layers. Specifically, it can be expressed as:

[0108] f(a)=max(0,a)

[0109] Among them, a is the input value, and f(a) is the output of the activation function. The loss function uses the cross entropy loss function, which is specifically expressed as:

[0110]

[0111] Where n k is the amount of data owned by the kth client. is the i-th feature contained in the data of the k-th client. is the i-th label contained in the data of the k-th client, and these labels are encoded using one-hot representation. The GCN model is denoted as G, and the node feature X is input into the model to obtain the output G(X).

[0112] 203. Training the global model parameters to obtain updated local model parameters;

[0113] In some embodiments, after receiving the global model sent by the server, the client uses its own data to train the global model. The client uses the gradient descent method to update the model in local training. The specific update rules for each round are specifically expressed as follows:

[0114]

[0115] in represents the weights of L GCN layers in the e-th local training step of client k in the t-th round of global training The series connection, f k represents the local loss function and η is the learning rate.

[0116] 204. Send the updated local model parameters to the server;

[0117] In some embodiments, in each iteration round t, the client trains the local model w of the client locally for τ times. k is sent to the server.

[0118] 205. If no global model parameters are received from the server within a preset time, the updating is terminated, otherwise the above process is repeated; the preset time is used to indicate the convergence time of the global model parameters.

[0119] In some embodiments, if the client does not receive the global model parameters sent by the server within a preset time, it indicates that the server believes that the global model parameters have reached convergence, and the client can end the model update.

[0120] In some preferred embodiments, Figure 4 As shown, before receiving the global model parameters from the server, the method further includes:

[0121] 211. Encrypt the neighbor feature of the local graph node using a homomorphic encryption algorithm to generate a ciphertext of the neighbor feature of the local graph node;

[0122] In some embodiments, each client sends the neighbor characteristics of the local graph node to the server, which are specifically expressed as:

[0123]

[0124] Where c(j) represents the index of the client containing graph node j, is the indicator function. When k is equal to c(j), that is, when client z contains the information of graph node j, the indicator function The value of is 1; otherwise, the value is 0. is the encryption operator, A ij represents the weighted adjacency matrix, x j A feature vector representing a graph node.

[0125] 212. Send the ciphertext of the neighbor feature of the local graph node to the graph federated learning server;

[0126] 213. Receive the multi-hop neighbor feature ciphertext of the local map node from the server;

[0127] In some embodiments, after receiving the ciphertext of neighbor features of all local graph nodes from the client, the server performs a neighbor aggregation operation, which can be expressed as:

[0128]

[0129] Where c(j) represents the index of the client containing graph node j, is the indicator function. When z is equal to c(j), that is, when client z contains the information of graph node j, the indicator function The value of is 1; otherwise, the value is 0. is the encryption operator, Represents the neighbor nodes of graph node i, A ij represents the weighted adjacency matrix, x j Represents the feature vector of graph node j. The server operates on all received neighbor feature arrays and accumulates the multi-hop neighbor features of each node.

[0130] 214. Decrypt the multi-hop neighbor feature ciphertext of the local map node to generate a multi-hop neighbor aggregation feature of the local map node.

[0131] This embodiment takes into account that in the federated graph convolutional neural network, in order to maintain the accuracy of the model, it is necessary to continuously exchange information about node neighbors during the training process, which leads to significant communication overhead. This embodiment adds the pre-training steps of steps 211-214 before updating the model parameters, and only requires a limited number of communications, which significantly reduces the communication cost and saves the communication overhead.

[0132] Figure 5 is a schematic diagram of a model parameter update scenario based on graph federated learning according to an embodiment of the present invention. Figure 5 As shown, the server and client applied to graph federated learning, the method includes:

[0133] The client uses a homomorphic encryption algorithm to encrypt the neighbor features of the local graph node and generate the ciphertext of the neighbor features of the local graph node;

[0134] The client sends the encrypted feature text of the neighbor of the local graph node to the graph federated learning server;

[0135] The server aggregates the neighbor feature ciphertexts of the local graph nodes of all clients to obtain the multi-hop neighbor feature ciphertexts of the local graph nodes;

[0136] The server sends the multi-hop neighbor feature ciphertext of the local map node to the client;

[0137] The client decrypts the ciphertext of the multi-hop neighbor feature of the local map node to generate a multi-hop neighbor aggregation feature of the local map node;

[0138] The server initializes the global model parameters and broadcasts them to each client;

[0139] The client constructs a graph convolutional neural network based on global model parameters and multi-hop neighbor aggregation features;

[0140] The client trains the global model parameters to obtain updated local model parameters;

[0141] The server evaluates the local model parameters of each client to obtain a first confidence level of the local model parameters of each client;

[0142] The server performs quantum state encoding on the local model parameters of each client to obtain the quantum state model parameters of each client;

[0143] The server operates the quantum state model parameters of each client through the quantum self-attention circuit to obtain the second confidence level of the local model parameters of each client;

[0144] The server updates the global model parameters of the server according to the first confidence and the second confidence of the local model parameters of each client;

[0145] The server broadcasts the updated global model parameters to all clients;

[0146] If the global model parameters converge, the update ends, otherwise the above process is repeated.

[0147] A model parameter updating system based on graph federated learning according to an embodiment of the present invention is deployed on a server participating in graph federated learning. The model parameter updating system based on graph federated learning includes:

[0148] Processor; and

[0149] a memory arranged to store computer executable instructions which, when executed, cause the processor to:

[0150] Initialize the global model parameters and broadcast them to each client;

[0151] Receive local model parameters after training from the client;

[0152] Evaluate the local model parameters of each client to obtain a first confidence level of the local model parameters of each client; the first confidence level of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client;

[0153] Perform quantum state encoding on the local model parameters of each client to obtain the quantum state model parameters of each client;

[0154] The quantum state model parameters of each client are operated through a quantum self-attention circuit to obtain a second confidence level of the local model parameters of each client; the second confidence level of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients;

[0155] updating the global model parameters of the server according to the first confidence and the second confidence of the local model parameters of each client;

[0156] Broadcast the updated global model parameters to all clients;

[0157] If the global model parameters converge, the update ends, otherwise the above process is repeated.

[0158] A model parameter updating system based on graph federated learning according to an embodiment of the present invention is deployed on a client participating in graph federated learning. The model parameter updating system based on graph federated learning includes:

[0159] Processor; and

[0160] a memory arranged to store computer executable instructions which, when executed, cause the processor to:

[0161] Receive global model parameters from a server; the global model parameters are updated based on the first confidence and the second confidence of the local model parameters of each client; the first confidence of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client; the second confidence of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients, and the similarity between the local model parameters of each client and the local model parameters of other clients is obtained by calculating the quantum state model parameters of each client through a quantum self-attention circuit; the quantum state model parameters of each client are obtained by quantum state encoding the local model parameters of each client;

[0162] Constructing a graph convolutional neural network according to the global model parameters and multi-hop neighbor aggregation features;

[0163] Training the global model parameters to obtain updated local model parameters;

[0164] Sending updated local model parameters to the server;

[0165] If no global model parameters are received from the server within a preset time, the updating is terminated, otherwise the above process is repeated; the preset time is used to indicate the convergence time of the global model parameters.

[0166] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0167] The embodiments of this specification are described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0168] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the relevant hardware through a program, and the program can be stored in a computer-readable storage medium, which can include: ROM, RAM, disk or CD, etc.

[0169] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A model parameter updating method based on graph federated learning, characterized in that: The method is applied to a server for graph federated learning, and comprises the following steps: Initialize the global model parameters and broadcast them to each client; Receive local model parameters after training from the client; Evaluate the local model parameters of each client to obtain a first confidence level of the local model parameters of each client; the first confidence level of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client; Perform quantum state encoding on the local model parameters of each client to obtain the quantum state model parameters of each client; The quantum state model parameters of each client are operated through a quantum self-attention circuit to obtain a second confidence level of the local model parameters of each client; the second confidence level of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients; updating the global model parameters of the server according to the first confidence and the second confidence of the local model parameters of each client; Broadcast the updated global model parameters to all clients; If the global model parameters converge, the update ends, otherwise the above process is repeated.

2. According to a model parameter updating method based on graph federated learning according to claim 1, it is characterized in that: The process of initializing the global model parameters and broadcasting them to each client also includes the following steps: Receive the neighbor feature ciphertext of the local graph node from the client; Aggregate the neighbor feature ciphertexts of the local graph nodes of all clients to obtain the multi-hop neighbor feature ciphertexts of the local graph nodes; The multi-hop neighbor feature ciphertext of the local graph node is sent to the client.

3. According to a model parameter updating method based on graph federated learning according to claim 1, it is characterized in that: The step of evaluating the local model parameters of each client to obtain the first confidence level of the local model parameters of each client includes: The local model parameters of each client are measured similarly with the global model parameters of the server to obtain the similarity distance between the local model parameters of each client and the global model parameters of the server; The similarity distance between the local model parameters of each client and the global model parameters of the server is processed by attention using a standardized function to obtain the first confidence of the local model parameters of each client.

4. According to a model parameter updating method based on graph federated learning according to claim 1, it is characterized in that: The step of calculating the quantum state model parameters of each client through a quantum self-attention circuit to obtain the second confidence level of the local model parameters of each client includes: The quantum state model parameters of all clients are subjected to corresponding parameter evolution respectively, and the Z direction is measured according to the Pauli-Z gate to obtain the first measurement expectation and the second measurement expectation corresponding to the quantum state model parameters of each client; According to the difference between the first measurement expectation and the second measurement expectation corresponding to the quantum state model parameters of each client, the quantum attention matrix of the corresponding client is obtained; According to the quantum attention matrix of each client, the similarity of the local model parameters between each client and other clients is calculated; According to the similarity between the local model parameters of each client and other clients, a second confidence level of the local model parameters of each client is calculated.

5. A model parameter updating method based on graph federated learning, characterized in that: Applied to a client of graph federated learning, the method comprises the following steps: Receive global model parameters from a server; the global model parameters are updated based on the first confidence and the second confidence of the local model parameters of each client; the first confidence of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client; the second confidence of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients, and the similarity between the local model parameters of each client and the local model parameters of other clients is obtained by calculating the quantum state model parameters of each client through a quantum self-attention circuit; the quantum state model parameters of each client are obtained by quantum state encoding the local model parameters of each client; Constructing a graph convolutional neural network according to the global model parameters and multi-hop neighbor aggregation features; Training the global model parameters to obtain updated local model parameters; Sending updated local model parameters to the server; If no global model parameters are received from the server within a preset time, the updating is terminated, otherwise the above process is repeated; the preset time is used to indicate the convergence time of the global model parameters.

6. According to the model parameter updating method based on graph federated learning according to claim 5, it is characterized in that: The receiving of the global model parameters from the server also includes the following steps: Use the homomorphic encryption algorithm to encrypt the neighbor features of the local graph node to generate the ciphertext of the neighbor features of the local graph node; Send the ciphertext of the neighbor features of the local graph node to the graph federated learning server; Receive multi-hop neighbor feature ciphertext of the local graph node from the server; The multi-hop neighbor feature ciphertext of the local graph node is decrypted to generate a multi-hop neighbor aggregation feature.

7. A model parameter updating method based on graph federated learning according to claim 5 or 6, characterized in that: The constructing of a graph convolutional neural network according to the global model parameters and multi-hop neighbor aggregation features includes: Construct the graph node embedding vector of each convolutional layer based on the multi-hop neighbor aggregation features; A graph convolutional neural network is constructed based on the graph node embedding vectors of each convolutional layer and the global model parameters.

8. A model parameter updating system based on graph federated learning, characterized in that: Deployed on a server participating in graph federated learning, the model parameter updating system based on graph federated learning includes: Processor; and a memory arranged to store computer executable instructions which, when executed, cause the processor to: Initialize the global model parameters and broadcast them to each client; Receive local model parameters after training from the client; Evaluate the local model parameters of each client to obtain a first confidence level of the local model parameters of each client; the first confidence level of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client; Perform quantum state encoding on the local model parameters of each client to obtain the quantum state model parameters of each client; The quantum state model parameters of each client are operated through a quantum self-attention circuit to obtain a second confidence level of the local model parameters of each client; the second confidence level of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients; updating the global model parameters of the server according to the first confidence and the second confidence of the local model parameters of each client; Broadcast the updated global model parameters to all clients; If the global model parameters converge, the update ends, otherwise the above process is repeated.

9. A model parameter updating system based on graph federated learning, characterized in that: Deployed on a client participating in graph federated learning, the model parameter updating system based on graph federated learning includes: Processor; and a memory arranged to store computer executable instructions which, when executed, cause the processor to: Receive global model parameters from a server; the global model parameters are updated based on the first confidence and the second confidence of the local model parameters of each client; the first confidence of the local model parameters of each client is determined by the similarity between the global model parameters of the server and the local model parameters of each client; the second confidence of the local model parameters of each client is determined by the similarity between the local model parameters of each client and the local model parameters of other clients, and the similarity between the local model parameters of each client and the local model parameters of other clients is obtained by calculating the quantum state model parameters of each client through a quantum self-attention circuit; the quantum state model parameters of each client are obtained by quantum state encoding the local model parameters of each client; Constructing a graph convolutional neural network according to the global model parameters and multi-hop neighbor aggregation features; Training the global model parameters to obtain updated local model parameters; Sending updated local model parameters to the server; If no global model parameters are received from the server within a preset time, the updating is terminated, otherwise the above process is repeated; the preset time is used to indicate the convergence time of the global model parameters.

Images