Federal modeling method for nonlinear regression

Abstract

The invention belongs to the field of algorithms, and particularly discloses a federal modeling method for nonlinear regression. According to the modeling method provided by the invention, a third-party node is removed, the complexity of the system is greatly reduced, and any two parties are allowed to train the joint model without the help of a trusted coordinator; the privacy of the data is protected through an encryption algorithm; the characteristics of the model are enriched through longitudinal federation modeling, and the accuracy of the model is improved; the defect that a linear regression prediction result is large in error is overcome through a nonlinear regression algorithm. The model not only can predict a single target value, but also can predict a plurality of target values by establishing one model, and model prediction can be realized without a plurality of processes; the number of dimensions of the features is constant, the effective number of the samples is within a certain multiple of the dimensions of the features according to statistical characteristics, the number of the samples is compressed under the condition that the dimensions of the features are few, and the model training time is shortened.

Description

technical field [0001] The invention relates to the field of algorithms, in particular to a federated modeling method of nonlinear regression. Background technique [0002] Machine learning (Machine Learning, referred to as ML) refers to the process of using certain algorithms to guide computers to use known data to independently build reasonable models, and use this model to give judgments on new situations. , Mechanical failure prediction, insurance pricing, financial risk management and other applications play a very important role. Traditionally, machine learning models are trained on a centralized corpus of data, which may be collected by a single or multiple data providers. Although parallel distributed algorithms have been developed to speed up the training process, the training data itself is still collected and stored centrally in a single data center. [0003] In May 2018, the European Union passed the General Data Protection Regulation (GDPR) bill, raising the r...

AI Technical Summary

Problems solved by technology

[0007] The current federated learning technology is only designed for a small number of specific algorithms, such as linear regression, logistic regression, shallow neural network, tree model, etc., mainly focusing on the application of algorithms in the financial field, and in wider scenarios such as the industrial field , there is still a lack of research on a large number of algorithm models, especially the nonlinear regression model with high application value, and the lack of implementation of security federation modeling

Images