HIGH DIMENSIONAL DATA CLASSIFICATION STRATEGY BASED ON REDUCED DIMENSION DICTIONARY LEARNING

In order to solve the problem of high-dimensional data and non-linearity in dictionary learning, a high dimensional data classification strategy based on reduced dimension dictionary learning is proposed. In the dimension reduction stage, the automatic encoder was used to learn a nonlinear mapping, which could reduce the dimension and retain the nonlinear structure of high-dimensional data. In the dictionary learning stage, label embedding was used for local constraints. In the learning process, the decomposable nonlinear local structure was retained, the ability to distinguish classes was enhanced, and the mapping function and dictionary were optimized. Experimental results on several benchmark data sets show that the proposed method can effectively solve the problems of high-dimensional data and nonlinearity in dictionary learning.