Broad Learning System Based on Nonlinear Transformation and Feedback Adjustment

Abstract

Broad Learning System has been used in many applications. For example, face recognition, image classification and segmentation, time series prediction. A broad learning system algorithm based on nonlinear transformation and feedback adjustment proposed to improve the accuracy of the traditional broad learning system model. This paper analyzes the impact of data on the model from the perspective of probability statistics and feature mapping, and finds the best nonlinear mapping function from the angle of data tilt to accurate data sets. In terms of the accuracy of model training, fine-tuning the broad learning system in the form of a feedback model, set the appropriate number of fine-tuning and fine-tuning rates to improve the accuracy of the model training; In addition, combined with nonlinear transformation and feedback adjustment model, new algorithms and corresponding diagrams are given. In this paper, weather data sets are used to prove the rationality and effectiveness of the algorithm framework.