FACE RECOGNITION AND CONSTRUCTION METHOD AND SYSTEM BASED ON NON-LINEAR NON-NEGATIVE MATRIX DECOMPOSITION, AND STORAGE MEDIUM

摘要

A face recognition construction method and system based on non-linear non-negative matrix decomposition, and a storage medium. The method comprises: depicting the degree of loss after matrix decomposition by using l _2,p-norm; enhancing the sparsity representation of a feature by using l ₁-norm of a matrix, and adding a regularization term regarding a matrix H into a loss function; constructing a target function F(W,H) by means of a degree of loss depicting step and a sparsity enhancement step; and obtaining an update iteration formula for kernel non-negative matrix decomposition of fractional power inner product. The method can solve the problem that a kernel non-negative matrix decomposition algorithm is sensitive to a singular value, can enhance the sparsity of the algorithm to feature representation, and also solves the problem that super parameters of a polynomial kernel function can only be integers.