A NEW K-WINNERS-TAKE-ALL NEURAL-NETWORK AND ITS ARRAY ARCHITECTURE

In this paper, a new neural-network model called WINSTRON and its nove l array architecture are proposed. Based on a competitive learning alg orithm that is originated from the coarse-fine competition, WINSTRON c an identify the fi larger elements or the k smaller ones in a data set . We will then prove that WINSTRON converges to the correct state in a ny situation, In addition, the convergence rates of WINSTRON for three special data distributions will be derived. In order to realize WINST RON, its array architecture with low hardware complexity and high comp uting speed is also detailed. Finally, simulation results are included to demonstrate its effectiveness and its advantages over three existi ng networks.