Collaborative modeling device based on federated learning
The collaborative modeling device enhances federated learning by ensuring data security and network stability through local data processing and secure transmission, addressing network instability and security issues in existing systems.
Patent Information
- Application Number
- CN202310746070.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-21
- Publication Date
- 2024-12-24
AI Technical Summary
Existing federated learning systems face issues with network instability and low information security during Internet transmission, which affect the stability and security of data exchange among participants.
A collaborative modeling device utilizing a network transmission module, encryption calculation module, neural network computing module, and storage module to ensure data security and network stability by processing data locally and transmitting models through wired and wireless networks, with encryption and preprocessing to meet preset conditions.
Improves network stability and ensures information security in federated learning by maintaining local data within the domain and using encryption to standardize data, facilitating stable model training and secure transmission.
Smart Images
Figure CN119180188A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to machine learning technology, and in particular to a collaborative modeling device based on federated learning. Background Art
[0002] With the development of artificial intelligence technology, machine learning has penetrated into all areas of life. With the support of big data, cloud computing and algorithm models, machine learning capabilities and application scenarios have been expanded. In specific implementation, centralized machine learning training has high requirements on server performance, and centralized data processing has large-scale data leakage problems. In response to the above problems, the introduction of the federated learning framework has become an important direction for the development of machine learning.
[0003] In the existing technology, federated learning enables multiple participants to jointly participate in model training through collaborative modeling; the model training of multiple participants is coordinated based on a central server, and the training data of each participant is stored locally. The shared model is trained based on local data and uploaded to the federated learning platform. Distributed model training is used to achieve large-scale data training, thereby improving the generalization ability of the model and maintaining the privacy and security of the data.
[0004] In the existing technology, during the implementation of federated learning, the various participants usually do not have the network connection stability at the LAN level, and need to conduct federated modeling through the Internet. During the Internet transmission process, there are information security issues in information transmission; therefore, in the existing federated learning, there are technical problems of unstable network and low information security assurance capabilities. Summary of the invention
[0005] The present application provides a collaborative modeling device based on federated learning, which is used to solve the technical problems of network instability and low information security assurance capability in existing federated learning.
[0006] The present application provides a collaborative modeling device based on federated learning, the device comprising:
[0007] A network transmission module, used to connect to a user data system to obtain a first target data set for federated learning;
[0008] An encryption calculation module, used for performing encryption calculation on the first target data set to obtain a second target data set;
[0009] A neural network computing module, used for training a local model based on a second target data set to obtain a federated model;
[0010] A storage module, used to store the first target data set, the second target data set and the federated model;
[0011] The control module is used to control the operation of the collaborative modeling device, generate control instructions, and determine the operation information record.
[0012] Optionally, the control module is specifically used for:
[0013] Determine whether the first target data set meets the preset conditions, and if so, control the encryption calculation module to preprocess the first target data set, and perform encryption calculation on the preprocessed first target data set to obtain the second target data set;
[0014] The preprocessing includes de-duplication processing, missing fill processing and / or regularization processing.
[0015] Optionally, the network transmission module is also used to connect to the federated learning platform and transmit the federated model to the federated learning platform.
[0016] Optionally, the network transmission module includes:
[0017] A wired network transmission unit, used to connect to a user data system, obtain a first target data set, and transmit the first target data set to a storage module; and / or used to connect to a federated learning platform and transmit a federated model to the federated learning platform;
[0018] The wireless network transmission unit connects to the federated learning platform and transmits the federated model to the federated learning platform.
[0019] Optionally, the control module is further configured to:
[0020] Obtaining the Internet connection status of the wired network transmission unit and the wireless network transmission unit;
[0021] When the Internet connection status of the wired network transmission unit is stable, the wired network unit is used to connect to the federated learning platform, and the federated model is transmitted to the federated learning platform.
[0022] Optionally, the control module is further configured to:
[0023] When the Internet connection status of the wired network transmission unit is unstable, the wireless network connection unit is used to connect to the federated learning platform and transmit the federated model to the federated learning platform.
[0024] Optionally, the neural network computing module is specifically used for:
[0025] Perform calculation and back-propagation optimization on the local model based on the second target data set;
[0026] When a control instruction is received, model parameters of the local model are determined, and based on the control instruction, the local model is determined as a federal model, and the model parameters and the federal model are transmitted to the storage module.
[0027] Optionally, the control module is further configured to:
[0028] When the number of back-propagation optimizations reaches a preset threshold, the network transmission module is controlled to communicate with the federated learning platform and determine the modeling instructions;
[0029] If the modeling instruction is an end modeling instruction, a control instruction is sent to the network computing module based on the end modeling instruction.
[0030] Optionally, the device also includes: a display module for displaying operation information records; wherein the operation information records include: the start, completion, error and idle status information of the federated model; and / or the usage information of the system memory, operator and network of the modeling device.
[0031] Optionally, the control module is further configured to:
[0032] The operation status of the modeling device is determined based on the operation information record, and when an error occurs in the modeling device, the display module is controlled to display prompt information indicating an operation error of the modeling device.
[0033] The collaborative modeling device based on federated learning provided by the present application includes: connecting to the user data system through a network transmission module to obtain a first target data set for federated learning; performing encryption calculation on the first target data set through an encryption calculation module to obtain a second target data set; training a local model based on the second target data set through a neural network calculation module to obtain a federated model; storing the first target data set, the second target data set and the federated model through a storage module; controlling the operation of the collaborative modeling device through a control module, generating control instructions, and determining operation information records; thereby utilizing the network transmission module to connect with the user's data system to obtain data for model training from the local area, and ensuring that local data does not go out of the domain, thereby ensuring data security; utilizing encryption calculation to process data to achieve data standardization, thereby facilitating further model training; utilizing the neural network calculation module to train the processed data and obtain the trained federated model; transmitting the federated model to the federated learning platform through the Internet transmission unit of the network transmission module, thereby achieving shared modeling and ensuring the stability of network transmission; thereby achieving the technical effect of improving the network stability of federated learning and ensuring information security. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0035] Figure 1 A schematic diagram of a collaborative modeling device based on federated learning provided in an embodiment of the present application Figure 1 ;
[0036] Figure 2 A flowchart of applying the collaborative modeling device based on federated learning provided in an embodiment of the present application to federated modeling;
[0037] Figure 3 A schematic diagram of the structure of a wired network transmission unit in a collaborative modeling device based on federated learning provided in an embodiment of the present application;
[0038] Figure 4 A schematic diagram of the structure of a wireless network transmission unit in a collaborative modeling device based on federated learning provided in an embodiment of the present application;
[0039] Figure 5 A schematic diagram of the structure of a display module in a collaborative modeling device based on federated learning provided in an embodiment of the present application;
[0040] Figure 6 A schematic diagram of a collaborative modeling device based on federated learning provided in an embodiment of the present application Figure 2 .
[0041] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION
[0042] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.
[0043] First, the terms involved in this application are explained:
[0044] Federated learning (Federated machine learning / Federated Learning): Federated learning is a machine learning framework that can effectively help multiple institutions use data and conduct machine learning modeling while meeting the requirements of user privacy protection, data security, and government regulations. As a distributed machine learning paradigm, federated learning can effectively solve the problem of data silos, allowing participants to jointly model without sharing data, which can technically break data silos and achieve AI collaboration.
[0045] In the prior art, federated learning enables multiple participants to participate in model training through collaborative modeling; the model training of multiple participants is coordinated based on a central server, and the training data of each participant is stored locally. The shared model is trained based on local data and uploaded to the federated learning platform. Distributed model training is used to achieve large-scale data training, thereby improving the generalization ability of the model and maintaining the privacy and security of the data. In the prior art, during the implementation of federated learning, each participant usually does not have the network connection stability of the LAN level, and needs to conduct federated modeling through the Internet. During the Internet transmission process, there are information security issues in information transmission; therefore, in the existing federated learning, there are technical problems of unstable network and low information security capabilities.
[0046] The collaborative modeling device based on federated learning provided by the present application includes: connecting to the user data system through a network transmission module to obtain a first target data set for federated learning; performing encryption calculation on the first target data set through an encryption calculation module to obtain a second target data set; training a local model based on the second target data set through a neural network calculation module to obtain a federated model; storing the first target data set, the second target data set and the federated model through a storage module; controlling the operation of the collaborative modeling device through a control module, generating control instructions, and determining operation information records; thereby utilizing the network transmission module to connect with the user's data system to obtain data for model training from the local area, and ensuring that local data does not go out of the domain, thereby ensuring data security; utilizing encryption calculation to process data to achieve data standardization, thereby facilitating further model training; utilizing the neural network calculation module to train the processed data and obtain the trained federated model; transmitting the federated model to the federated learning platform through the Internet transmission unit of the network transmission module, thereby achieving shared modeling and ensuring the stability of network transmission; thereby achieving the technical effect of improving the network stability of federated learning and ensuring information security.
[0047] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.
[0048] Figure 1 A schematic diagram of a collaborative modeling device based on federated learning provided in an embodiment of the present application Figure 1 .like Figure 1As shown, a collaborative modeling device 100 based on federated learning provided in an embodiment of the present application includes: a network transmission module 101, an encryption calculation module 102, a neural network calculation module 103, a storage module 104 and a control module 105.
[0049] A network transmission module 101 is used to connect to a user data system to obtain a first target data set for performing federated learning;
[0050] In this embodiment, the network transmission module 101 includes a wired network transmission unit and a wireless network transmission unit; the wired network transmission unit includes two ports, a local transmission port and an Internet transmission port; the local transmission port is used to connect to a local user database to obtain a local first target data set; the Internet transmission port is used to connect to a federated learning platform to transmit the federated model after training; the purpose is to separate local data transmission from Internet data transmission to maintain the information security of local data; the wireless network transmission unit is used to use a stable wireless network including a 5G network to perform data transmission when the Internet transmission port of the wired network transmission unit is disconnected or the signal is unstable, in order to maintain the network stability of the collaborative modeling device in federated learning.
[0051] An encryption calculation module 102, configured to perform encryption calculation on the first target data set to obtain a second target data set;
[0052] In this embodiment, the encryption calculation module 102 encrypts the first target data set, and obtains the second target data set by performing data deduplication, missing filling and regularization processing on the first target data set; wherein the first target data set can be: image data, audio data, video data, sensor sequence and table data.
[0053] A neural network calculation module 103, used to train the local model based on the second target data set to obtain a federated model;
[0054] In this embodiment, the neural network computing module 103 performs local model calculation and back propagation optimization on the second target data set. After the number of calculations and back propagation optimizations reaches the rounds set by the user, model aggregation communication is performed through the Internet interface of the wired network transmission unit or the wireless network transmission unit and the federated learning platform; the model aggregation communication is to obtain the model parameters trained by the participants participating in the federated modeling, which are used to train the global model; the model parameters after aggregation are saved, and the next round of local model optimization is entered.
[0055] A storage module 104, used to store the first target data set, the second target data set and the federated model;
[0056] In this embodiment, after the network transmission module 101 obtains the first target data set, the first target data set is stored in the storage module, so that the encryption calculation module 102 can take out the first target data set from the storage module 104 and determine the second target data set; the second target data set is stored in the storage module 104, which is used by the neural network calculation module 103 to determine the federated model based on the second target data set; the federated model is stored in the storage module 104, so that the network transmission module 101 and the federated learning platform can communicate and transmit the federated model.
[0057] In the first example, the storage module is composed of TiPlus5000 NVMe memory, which is connected to the PCIe bus of the central processing unit through the M.2 interface. It is responsible for the storage of federated learning data, temporary models during training, and federated models after training is completed.
[0058] The control module 105 is used to control the operation of the collaborative modeling device, generate control instructions, and determine the operation information record;
[0059] In this embodiment, the control module 105 is a central processing unit that controls the operation of the collaborative modeling device, and is used to generate control instructions based on user needs and control the operation and regulation of each module of the collaborative modeling device.
[0060] The collaborative modeling device based on federated learning provided by the present application includes: connecting to the user data system through a network transmission module to obtain a first target data set for federated learning; performing encryption calculation on the first target data set through an encryption calculation module to obtain a second target data set; training a local model based on the second target data set through a neural network calculation module to obtain a federated model; storing the first target data set, the second target data set and the federated model through a storage module; controlling the operation of the collaborative modeling device through a control module, generating control instructions, and determining operation information records; thereby utilizing the network transmission module to connect with the user's data system to obtain data for model training from the local area, and ensuring that local data does not go out of the domain, thereby ensuring data security; utilizing encryption calculation to process data to achieve data standardization, thereby facilitating further model training; utilizing the neural network calculation module to train the processed data and obtain the trained federated model; transmitting the federated model to the federated learning platform through the Internet transmission unit of the network transmission module, thereby achieving shared modeling and ensuring the stability of network transmission; thereby achieving the technical effect of improving the network stability of federated learning and ensuring information security.
[0061] Figure 2 The flowchart of the collaborative modeling device based on federated learning provided in the embodiment of the present application is applied to federated modeling. Figure 2 As shown in the figure, the federated modeling process includes:
[0062] S201, connecting to a user data system through a wired network transmission unit, acquiring a first target data set, and transmitting the first target data set to a storage module;
[0063] S202, determining whether the first target data set meets a preset condition;
[0064] In this embodiment, the preset condition is to determine the integrity and standardization of the first target data set; for example, in the second example, the first target data set is image data, and the size and color expression of the image data are inconsistent, then it is determined that the image data meets the preset condition and the image data needs to be preprocessed.
[0065] S203: If yes, control the encryption calculation module to preprocess the first target data set, and perform encryption calculation on the preprocessed first target data set to obtain a second target data set;
[0066] In this embodiment, the first target data set is image data, and the preprocessing may be image pixel unification processing, image gray value unification processing, and image vectorization processing.
[0067] In this embodiment, during the model training process, the encryption computing module can perform homomorphic encryption, differential privacy protection, and sharded secret sharing on the model parameters and gradients based on the training requirements of the federated model to ensure the security of the user's training data.
[0068] S204, controlling the neural network calculation module to calculate and perform back propagation optimization on the local model based on the second target data set;
[0069] S205, when the number of back propagation optimizations reaches a preset threshold, controlling the network transmission module to communicate with the federated learning platform and determining a modeling instruction;
[0070] In this embodiment, back propagation optimization is a machine learning optimization method based on the gradient descent algorithm; in the third exemplary embodiment, the collaborative modeling is to train an artificial neural network, and the implementation steps of back propagation optimization are: first calculate the state and activation value of each layer of the neural network until the last layer; calculate the error of each layer of the neural network, and the error calculation process is carried out from the last layer forward; calculate the gradient of each neuron connection weight; update the parameters according to the gradient descent method to reduce the error; know that the number of back propagation times reaches a preset threshold and / or the difference between two adjacent errors is lower than the set preset error difference value.
[0071] S206: If the modeling instruction is an end modeling instruction, sending a control instruction to the neural network calculation module based on the end modeling instruction;
[0072] S207, controlling the neural network calculation module to determine the model parameters of the local model when receiving the control instruction, and determining the local model as a federated model based on the control instruction, and transmitting the model parameters and the federated model to the storage module;
[0073] S208, obtaining the Internet connection status of the wired network transmission unit and the wireless network transmission unit;
[0074] S209, when the Internet connection state of the wired network transmission unit is stable, use the wired network unit to connect to the federated learning platform, and transmit the federated model to the federated learning platform;
[0075] S210. When the Internet connection status of the wired network transmission unit is unstable, use the wireless network connection unit to connect to the federated learning platform and transmit the federated model to the federated learning platform.
[0076] By executing S201 to S210, it is possible to connect the user's local data system through the collaborative modeling device, and pre-process the data to be trained in the data system to make it conform to the data specifications of the training model; train the normalized data and obtain the trained federated model, and transmit the federated model to the federated learning platform through the Internet transmission function of the transmission network to realize model sharing; realize the security maintenance of data information by separating local data transmission and Internet data transmission, and ensure the stability of the data transmission network by combining wireless network and wired network, thereby achieving the technical effect of improving the network stability of federated learning and ensuring information security.
[0077] Figure 3 This is a schematic diagram of the structure of a wired network transmission unit provided in an embodiment of the present application. Figure 3 As shown, the wired network transmission unit is composed of two RTL8111H Gigabit Ethernet core boards and RJ45 ports, which are respectively used for local intranet data transmission and Internet federal model transmission. The two Gigabit Ethernet core boards are connected to the PCLe bus of the central processing unit through the M.2 interface, and the RJ45 port is directly connected to the output of the Gigabit Ethernet core board; wherein the central processing unit is the control module in this application; in this embodiment, during the federated collaborative modeling process, the user sets the address and port of the federated learning platform server on the mobile network configuration interface of the collaborative modeling device of this embodiment, and the user connects to the user data system through the RJ port of the wired network transmission unit; in this embodiment, after the federal model is built, the wired network transmission unit uploads the federal model to the federal learning platform through the Internet RJ45 port, thereby realizing data isolation between the internal and external networks and ensuring data security.
[0078] Figure 4 This is a schematic diagram of the structure of a wireless network transmission unit provided in an embodiment of the present application. Figure 4 As shown, the wireless network transmission unit is composed of a SIM8200EA 5G communication core board and a 5G antenna. The 5G communication core board is connected to the PCLe bus of the central processing unit through the M.2 interface, and the 5G antenna is connected to the 5G communication core board through the IPEX-4 interface, wherein the central processing unit is the control module in this application; the wireless network transmission unit is responsible for connecting to the Internet through the mobile communication network to ensure that the federated learning process will not be interrupted when the wired network is disconnected, or to undertake the transmission task of the Internet federated model when the wired network transmission unit is not connected.
[0079] In this embodiment, the central processing unit is composed of an RK3399 single-chip microcomputer system, which is responsible for running the collaborative modeling device, processing data access requests of the storage module, network communication requests of the wired and wireless network transmission units, calling the encryption calculation module to complete the preprocessing and privacy encryption of the federated learning data, calling the neural network calculation module to complete the forward calculation and back propagation optimization in the federated learning training process, and displaying the federated learning training status in real time through the display module.
[0080] Figure 5 This is a schematic diagram of the structure of the display module provided in the embodiment of the present application. Figure 5 The display module is composed of an AT070TN92 liquid crystal driver board and an LCD screen. The display module is connected to a central processing unit and controlled by the central processing unit, wherein the central processing unit is the control module of the collaborative modeling device in this application; based on the control of the central processing unit, the start, completion, error and idle status information in the federated learning modeling are displayed in real time; and / or the usage information of the system memory, operator and network of the modeling device; and when an error occurs in the modeling device, a prompt message indicating an operation error of the modeling device is displayed.
[0081] Figure 6 A schematic diagram of a collaborative modeling device based on federated learning provided in an embodiment of the present application Figure 2 The device of this embodiment may be in the form of software and / or hardware. Figure 6 As shown, a collaborative modeling device 600 based on federated learning provided in an embodiment of the present application includes: a network transmission module 601, an encryption calculation module 602, a neural network calculation module 603, a storage module 604, a control module 605, and a display module 606. The network transmission module 601 includes a wired network transmission unit 607 and a wireless network transmission unit 608.
[0082] A network transmission module 601 is used to connect to a user data system to obtain a first target data set for federated learning;
[0083] An encryption calculation module 602 is used to perform encryption calculation on the first target data set to obtain a second target data set;
[0084] A neural network calculation module 603 is used to train the local model based on the second target data set to obtain a federated model;
[0085] A neural network computing module 604, used to store the first target data set, the second target data set and the federated model;
[0086] The control module 605 is used to control the operation of the collaborative modeling device, generate control instructions, and determine the operation information record.
[0087] In a possible implementation, the control module 605 is specifically configured to:
[0088] Determine whether the first target data set meets the preset conditions. If yes, control the encryption calculation module 602 to preprocess the first target data set, and perform encryption calculation on the preprocessed first target data set to obtain the second target data set;
[0089] The preprocessing includes de-duplication processing, missing fill processing and / or regularization processing.
[0090] In a possible implementation, the network transmission module 601 is also used to connect to the federated learning platform and transmit the federated model to the federated learning platform.
[0091] In a possible implementation, the network transmission module 601 includes:
[0092] A wired network transmission unit 607, used to connect to a user data system, obtain a first target data set, and transmit the first target data set to the neural network calculation module 604; and / or used to connect to a federated learning platform and transmit a federated model to the federated learning platform;
[0093] The wireless network transmission unit 608 is connected to the federated learning platform and transmits the federated model to the federated learning platform.
[0094] In a possible implementation, the control module 605 is further configured to:
[0095] Obtaining the Internet connection status of the wired network transmission unit 607 and the wireless network transmission unit 608;
[0096] When the Internet connection status of the wired network transmission unit 607 is stable, the wired network unit is used to connect to the federated learning platform, and the federated model is transmitted to the federated learning platform.
[0097] In a possible implementation, the control module 605 is further configured to:
[0098] When the Internet connection status of the wired network transmission unit 607 is unstable, the wireless network connection unit is used to connect to the federated learning platform, and the federated model is transmitted to the federated learning platform.
[0099] In a possible implementation, the neural network calculation module 603 is specifically used for:
[0100] Perform calculation and back-propagation optimization on the local model based on the second target data set;
[0101] When a control instruction is received, the model parameters of the local model are determined, and based on the control instruction, the local model is determined as a federal model, and the model parameters and the federal model are transmitted to the neural network calculation module 604.
[0102] In a possible implementation, the control module 605 is further configured to:
[0103] When the number of back-propagation optimizations reaches a preset threshold, the network transmission module 601 is controlled to communicate with the federated learning platform and determine the modeling instructions;
[0104] If the modeling instruction is an end modeling instruction, a control instruction is sent to the network computing module based on the end modeling instruction.
[0105] In one possible implementation, the device also includes: a display module 606, which is used to display operation information records; wherein the operation information records include: the start, completion, error and idle status information of the modeling of the federated model; and / or the usage information of the system memory, operator and network of the modeling device.
[0106] In a possible implementation, the control module 605 is further configured to:
[0107] The operation status of the modeling device is determined based on the operation information record. When an error occurs in the modeling device, the control display module 606 displays prompt information indicating an operation error of the modeling device.
[0108] The collaborative modeling device based on federated learning provided by the present application includes: connecting to a user data system through a network transmission module to obtain a first target data set for performing federated learning; wherein the network transmission module includes a wired network transmission unit and a wireless network transmission unit, and connecting to an internal data system through the wired network transmission unit to obtain the first target data set; performing encryption calculation on the first target data set to obtain a second target data set through an encryption calculation module, wherein before performing the encryption calculation, it is necessary to determine whether the first target data set meets a preset condition, and if it meets the preset condition, it is necessary to preprocess the first target data set, and then perform encryption calculation on the preprocessed data; training a local model based on the second target data set through a neural network calculation module to obtain a federated model; storing the first target data set, the second target data set and the federated model through a storage module; controlling the operation of the collaborative modeling device through a control module, generating control instructions, and determining the operation information record; and The control module generates control instructions, controls the wired network transmission unit of the network transmission module to connect to the user's local data system, obtains local training data for model training, and ensures that local data does not leave the domain; and based on the control module, determines whether the first target data set obtained from the local user data system meets the standards for entering encryption calculation. If not, the first target data set is preprocessed to ensure data standardization; the normalized data is encrypted and the second target data set is obtained, and the security of the data is ensured based on the encryption calculation; the neural network calculation module is controlled to process the obtained second target data set, and the model training is continued to obtain the federated model; the Internet connection status of the wired network transmission unit and the Internet connection status of the wireless network transmission unit are determined, and the transmission unit with a stable connection status is selected to upload the trained federated model to the federated learning platform, thereby ensuring the stability of network transmission; thereby achieving the technical effect of improving the network stability of federated learning and ensuring information security.
[0109] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.
[0110] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.
Claims
1. A collaborative modeling device based on federated learning, characterized in that: include: A network transmission module, used to connect to a user data system to obtain a first target data set for federated learning; An encryption calculation module, used for performing encryption calculation on the first target data set to obtain a second target data set; A neural network computing module, used for training a local model based on the second target data set to obtain a federated model; A storage module, used for storing the first target data set, the second target data set and the federated model; The control module is used to control the operation of the collaborative modeling device, generate control instructions, and determine the operation information record.
2. The device according to claim 1, characterized in that The control module is specifically used for: Determine whether the first target data set meets a preset condition, and if so, control the encryption calculation module to preprocess the first target data set, and perform encryption calculation on the preprocessed first target data set to obtain the second target data set; The preprocessing includes de-duplication processing, missing fill processing and / or regularization processing.
3. The device according to claim 1, characterized in that The network transmission module is also used to connect to the federated learning platform and transmit the federated model to the federated learning platform.
4. The device according to claim 3, characterized in that The network transmission module includes: a wired network transmission unit, used to connect to the user data system, obtain the first target data set, and transmit the first target data set to the storage module; and / or used to connect to the federated learning platform and transmit the federated model to the federated learning platform; A wireless network transmission unit is connected to the federated learning platform and transmits the federated model to the federated learning platform.
5. The device according to claim 4, characterized in that The control module is also used for: Acquiring the Internet connection status of the wired network transmission unit and the wireless network transmission unit; When the Internet connection state of the wired network transmission unit is stable, the wired network unit is used to connect to the federated learning platform, and the federated model is transmitted to the federated learning platform.
6. The device according to claim 5, characterized in that The control module is also used for: When the Internet connection status of the wired network transmission unit is unstable, the wireless network connection unit is used to connect to the federated learning platform, and the federated model is transmitted to the federated learning platform.
7. The device according to claim 1, characterized in that The neural network calculation module is specifically used for: Performing calculation and back-propagation optimization on the local model based on the second target data set; When the control instruction is received, the model parameters of the local model are determined, and based on the control instruction, the local model is determined as the federated model, and the model parameters and the federated model are transmitted to the storage module.
8. The device according to claim 7, characterized in that The control module is also used for: When the number of back-propagation optimizations reaches a preset threshold, controlling the network transmission module to communicate with the federated learning platform and determining a modeling instruction; If the modeling instruction is an end modeling instruction, a control instruction is sent to the neural network calculation module based on the end modeling instruction.
9. The device according to claim 1, characterized in that The device also includes: a display module for displaying the operation information record; wherein the operation information record includes: the start, completion, error and idle state information of the modeling of the federated model; and / or the usage information of the system memory, operator and network of the modeling device.
10. The device according to claim 9, characterized in that The control module is also used for: The operation status of the modeling device is determined based on the operation information record, and when an error occurs in the modeling device, the display module is controlled to display prompt information indicating the operation error of the modeling device.