Distributed Analysis Dictionary Learning Based on Consensus Constraints

Abstract

Distributed dictionary learning is to learn a global dictionary so that all data distributed in a network has sparse representation in the domain of the dictionary. Existing works are based on sparse synthesis model. We consider this problem based on the sparse analysis model, and propose a distributed analysis dictionary learning (ADL) algorithm using consensus constraints. In particular, local dictionaries corresponding to local data at each node are introduced, and the distributed ADL problem is formulated as a minimization problem with consensus constraints on local dictionaries and the global dictionary. An optimization method consisting of a sparse coding stage and a dictionary update stage is then developed. Experimental results have shown that the performance of the proposed algorithm is comparable to centralized ADL algorithms.