Summary
International Symposium on Nonlinear Theory and its Applications
2017
Session Number:A2L-C
Session:
Number:A2L-C-1
Winner-Take-All Neural Network with Distributed Winner Search Circuit
Kazuki Hanada, Shoya Ueda, Hidetaka Ito, Hiroomi Hikawa,
pp.140-143
Publication Date:2017/12/4
Online ISSN:2188-5079
Summary:
This paper proposes a new winner-take-all (WTA) circuit, in which winner search is distributed among neurons in the network. The winner neuron that has the shortest vector distance is searched by a simple logic circuit. In the circuit, the shortest vector distance is searched by a bit-by-bit comparison scheme. With this approach, the winner search method is embedded in all neurons, and global winner search module that hinders expandability is not needed. Winner-take all network (WTANN) with the proposed WTA was designed with VHDL, and its performance was examined by simulations and experiments. It was revealed that the speed of the winner search provided by the proposed WTA is extremely faster than the other WTA circuits.