Application of Matrix Decomposition in Machine Learning

Matrix decomposition is a mathematical method widely adopted in computer science for its reliability and outstanding performance. It is often chosen for use in the computer vision field, recommender system, image compression, and denoising. This paper introduces the classic matrix decomposition algorithms, including eigenvalue decomposition, UV decomposition, singular value decomposition (SVD), and their application in computer science. Additionally, the paper discusses the flaws of the adoption of singular value decomposition in image compression and denoising, a common problem in matrix decomposition‐‐‐sparse matrix. Finally, this paper introduces the PCANet, the combination of matrix decomposition and neural network, which inspires improving the interpretability of the deep learning model.