Title :
A Proximal Method for Dictionary Updating in Sparse Representations
Author :
Guan-Ju Peng ; Wen-Liang Hwang
Author_Institution :
Inst. of Inf. Sci., Taipei, Taiwan
Abstract :
In this paper, we propose a new dictionary updating method for sparse dictionary learning. Our method imposes the ℓ0 norm constraint on coefficients as well as a proximity regularization on the distance of dictionary modifications in the dictionary updating process. We show that the derived dictionary updating rule is a generalization of the K-SVD method. We study the convergence and the complexity of the proposed method. We also compare its performance with that of other methods.
Keywords :
approximation theory; computational complexity; learning (artificial intelligence); signal processing; singular value decomposition; K-SVD method; dictionary modifications; dictionary updating process; proximal method; proximity regularization; sparse dictionary learning; sparse representations; Approximation methods; Convergence; Dictionaries; Encoding; Estimation; Optimization; Signal processing algorithms; Dictionary learning; supervised learning;
Journal_Title :
Signal Processing, IEEE Transactions on
DOI :
10.1109/TSP.2015.2434323