A fast and efficient competitive learning design algorithm based on weight vector training in transform domain

This paper presents a new competitive learning (CL) algorithm which performs the training in the wavelet domain. In the algorithm, the winning neural units during the training process are identified using the partial distance search (PDS) technique so that little multiplication is required. The PDS can be performed over the lower resolution representation of codewords in the wavelet transform domain to further reduce the computation time required for training. Simulation results show that, at the expense of a possible slight decrease in performance, the algorithm requires less than 5% of the computational time required by the traditional CL algorithm in many cases.