Coefficient optimization model for a class of robust principal component analysis algorithms

Abstract

There are more and more large-scale data models with high dimensions, but these large-scale data often have strong noise
and sparse, which is troubled by information loss, noise infl uence and small sample. Nowadays, there is a demand for extracting eff ective
content from chaotic information in many fields, such as pattern recognition, machine learning and data mining, among which robust
principal component analysis is a common method to separate eff ective information from these raw data. Aiming at the traditional algorithm
of robust principal component analysis, this paper establishes a new optimization model by assigning new coefficients to the low-rank
matrix, which has a better correlation to the original matrix and improves the accuracy problem without changing the solving speed.