Journal of VLSI Signal Processing Systems
SIAM Journal on Matrix Analysis and Applications
Nonnegative Matrix and Tensor Factorizations: Applications to Exploratory Multi-way Data Analysis and Blind Source Separation
ICONIP'12 Proceedings of the 19th international conference on Neural Information Processing - Volume Part I
Hi-index | 0.00 |
Nonnegative Matrix Factorization (NMF) is an important tool in spectral data analysis. Various types of numerical optimization algorithms have been proposed for NMF, including multiplicative, projected gradient descent, alternating least squares and active-set ones. In this paper, we discuss the Tikhonov regularized version of the FC-NNLS algorithm (proposed by Benthem and Keenan in 2004) that belongs to a class of active-set methods in the context of its application to spectroscopy data. We noticed that starting iterative updates from a large value of a regularization parameter, and then decreasing it gradually to a very small value considerably reduces the risk of getting stuck into unfavorable local minima of a cost function. Moreover, our experiments demonstrate that this algorithm outperforms the well-known NMF algorithms in terms of Peak Signal-to-Noise Ratio (PSNR).