Research On Extreme Learning Machine Algorithm Based On Manifold Learning

Extreme learning machine(ELM)is one of the most important algorithms in the field of machine learning,which is mainly used to solve the problem of data classification and regression.It has simple structure and easy to realization,the learning speed is fast,and more importantly,it has good global optimization ability.Therefore,the study of ELM algorithm has important research significance and application value.As the successful application in many fields,the research of ELM has attracted much attention.In this paper,we give a study aiming to solve the inadequate learning of ELM.By exploiting the idea of manifold learning,we mine the potential information of data by using the manifold structure of data,and two different improved algorithms are proposed.The main research results are as follows:(1)Innovation research from the perspective of ELM algorithm model,based on manifold learning method,the idea of the neighborhood preserving embedding algorithm(Neighborhood preserving embedding,NPE)is introduced into the ELM model,a neighborhood preserving extreme learning machine algorithm is proposed(Neighborhood preserving extreme learning machine,NPELM).The algorithm maintains the essential geometric structure and intrinsic manifold information of data,and improves the classification performance of ELM by minimizing the intra-class dispersion matrix.(2)Based on the algorithm of NPELM,the inter-class dispersion is introduced,and a discriminative neighborhood preserving extreme learning machine algorithm based on manifold learning is proposed(Discriminative neighborhood preserving extreme learning machine,DNPELM).By defining the minimization of intra-class dispersion matrix and the maximization of inter-class dispersion matrix of data,the algorithm not only maintains the local geometric structure information of the data,but also gives full using of identifying information between data category,which can further improve the classification accuracy of ELM algorithm.The experimental results show that the two optimized algorithms have achieved good classification effect in the field of image recognition,and have better classification accuracy and generalization ability than other algorithms.