Distributed cooperative learning over networks via wavelet approximation

Abstract

This paper investigates the problem of the distributed cooperative learning over networks via the wavelet approximation. On the basis of the wavelet approximation (WA) theory, the novel distributed cooperative learning (DCL) method, called DCL-WA, is proposed in this paper. The wavelet series is used to approximate the function of network nodes. For the networked systems, DCL method is used to train the optimal weight coefficient matrices of wavelet series, so as to get the best approximation function of network nodes. An illustrative example is presented to show the efficiency of the proposed strategy.