Model training method and system for longitudinal federal learning
A model training and model technology, which is applied in the field of model training methods and systems of vertical federated learning, and can solve problems such as high time consumption and slow operation speed.
Inactive Publication Date: 2022-04-05
富算科技(上海)有限公司
View PDF7 Cites 3 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0005] The purpose of the embodiment of the present application is to provide a model training method and system for vertical federated learning to solve the need to involve semi-homomorphic encryption, public key communication between the model initiator and the data participant in the learning process of the prior art Steps such as semi-homomorphic decryption and semi-homomorphic decryption, the problem of high time-consuming and slow operation speed in the case of large amounts of data
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0070] The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.
[0071] In vertical federated learning scenarios, existing machine learning algorithms based on gradient descent optimization algorithms often rely on a reliable coordinator, especially for logistic regression algorithms. As a third-party role, the coordinator is independent of the data participant, and performs relevant intermediate result processing and communication between the data participant and the model initiator. However, in real scenarios, especially banks, operators and other institutions that are extremely strict on data security, they cannot accept this kind of algorithm that relies on a trusted third party, because it is difficult to find a target organization recognized by each participant to assume the role of coordinator . Therefore, the existing technology usually adopts semi-homom...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention provides a longitudinal federated learning model training method and system, and the method comprises the steps: obtaining a gradient intermediate value in a federated learning process of a plurality of participants according to the inner product of a data participant and a model initiator and a real label of the model initiator, and enabling the model initiator to mix a first random number for the gradient intermediate value, the data participant calculates the gradient of the data participant for the gradient intermediate value confused with the first random number and confuses a second random number, and then the model initiator removes the first random number and the data participant removes the second random number in sequence, so that both the model initiator and the data participant have respective gradient values; according to the method and the system, respective feature weights can be updated, encryption communication of gradient information by a data participant and a model initiator is realized by adopting a random number confusion mode, the method and the system are not limited by calculation times, the problem of precision loss does not exist, more complex federated learning model learning requirements can be supported, encryption time consumption is reduced, and the encryption efficiency is improved. And the processing efficiency is improved.
Description
technical field [0001] This application relates to the technical field of federated learning, in particular, to a model training method and system for vertical federated learning. Background technique [0002] As a data security computing technology, federated learning is gradually being applied in enterprises. It can realize the flow of data value among various institutions without leaving the original data, and create business value. control, advertising recommendation and other fields. Federated learning is a distributed computing architecture that supports multi-party secure computing. According to different business usage scenarios, it mainly includes three types: vertical federated learning, horizontal federated learning, and federated migration algorithms. At present, federated learning can already support a variety of machine learning algorithms. [0003] Algorithms such as logistic regression (logistic regression) is a classic machine learning model, suitable for ...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06N20/20
Inventor 尤志强卞阳
Owner 富算科技(上海)有限公司