Width learning method based on main component analysis

Abstract

The invention relates to the field of one-dimensional or two-dimensional data processing, and particularly to a data identification method based on main component analysis and width learning. According to the method, main component analysis is used for performing dimension reduction on original data for obtaining a characteristic node as input of a width learning network; and furthermore for displaying a relatively non-obvious characteristic node in the dimension-reduced data, a linear combination principle of the characteristic node is utilized for obtaining an enhancing node for the input ofthe width learning network; the enhancing node is continuously adjusted according to the precision and speed of the training result of a width learning data processing model, thereby perfecting the width learning training model. Compared with prior art, the width learning method has advantages of high identification precision and short training time at a data processing prospect. Furthermore thewidth learning method facilitates real-time network parameter updating.

Description

technical field [0001] The invention relates to the technical field of one-dimensional or two-dimensional data processing, in particular to a data recognition method based on principal component analysis and width learning. Background technique [0002] With the development of technology, more and more jobs can be done by computers to improve efficiency. These technologies can be collectively referred to as artificial intelligence. As we all know, deep learning is already one of the important fields of artificial intelligence. Deep learning has been successfully applied in many fields, especially in big data processing, and has achieved unprecedented recognition accuracy. Commonly used deep learning networks are Deep Belief Networks (DBN) [1],[2] , Deep Boltzmann Machines (DBM) [3] , Convolutional Neural Networks (CNN) [4],[5] . However, the disadvantages of deep learning are becoming increasingly prominent. Deep networks have very complex network structures and hyperp...

AI Technical Summary

Problems solved by technology

[0017] The technical problem to be solved by the present invention is that the structure of the existing deep learning method is complex, the parameter adjustment is uncertain, and the training time is relatively long. Further speaking, the BroadLearning System (broad learning system) proposed by C.L.Philip Chen and Zhulin Liu is Directly inputting high-dimensional original data into the width learning network will still cause relatively complex network structure and a large amount of calculation. A width learning method based on principal component analysis is proposed. This method will use principal component analysis to analyze the original Data dimensionality reduction is used as the feature node input by the width learning network. In order to further display the relatively inconspicuous feature nodes in the data after dimension reduction, the enhanced node input by the width learning network is obtained by the principle of linear combination of feature nodes. The accuracy and speed of the data processing model training results are used as the basis to continuously adjust the enhancement nodes and improve the width learning training model

Images