| 研究実績の概要 |
1)We proposed a novel dictionary learning algorithm with the log-regularizer and simultaneously with the coherence penalty based on proximal operators. Our proposed algorithm simply employs a decomposition scheme and alternating optimization, which transforms the overall problem into a set of single-vector variable sub-problems, with either one dictionary atom or one coefficient vector. Although the sub-problems are still nonsmooth and even nonconvex, remarkably they can be solved by proximal operators, and the closed-form solutions of the dictionary atoms and the coefficient vectors are obtained directly and explicitly. Was published in Digital Signal Processing (Elsevier), Vol. 63, No. 4, pp. 86-99.
2)We proposed two analysis dictionary learning algorithms for sparse representation with analysis model. The problem is formulated with the L1-norm regularizer and with two penalty terms on the analysis dictionary: the term of -log det (W’W) and the coherence penalty term. As the processing scheme, we employ a block coordinate descent framework, so that the overall problem is transformed into a set of minimizations of univariate sub-problems with respect to a single-vector variable. Each sub-problem is still nonsmooth, but it can be solved by a proximal operator and then the closed-form solutions can be obtained directly and explicitly. Was published in Neurocomputing (Elsevier), Vol. 239, C, pp. 165-180.
|
| 今後の研究の推進方策 |
1) We shall finish the part of Scheme for real-time sparse representation. The progress is going well.
2) We shall continue the research for new algorithms of sparse coding and dictionary leaning.
3) We shall apply our algorithms to Defect Detection on Thin-Wall Structure with ultrasonic transducers, an important engineering problem.
4) We shall apply our algorithms to Device-Free Localization with wireless transmitters/receivers, another important engineering problem.
5) We shall try to research on the relationship of dictionary learning problem and Deep learning.
|
