Title :
Analogue Globally Stable WTA Neural Circuit
Author :
Tymoshchuk, Pavlo ; Lobur, Mykhaylo
Author_Institution :
Dept. of CAD/CAM, Lviv Polytech. Nat. Univ.
Abstract :
A new inhibitory analogue WTA (winner-take-all) neural circuit which identifies maximal among N unknown input signals is proposed. As a building block the second order analogue globally stable neural network of Hopfield type is used. The connection matrix belongs to the class of diagonally stable block diagonal matrices and the activation functions are piecewise linear or sigmoid. The mathematical justification of the circuit functioning, comparative evaluation with analogs and computer simulation results are given
Keywords :
Hopfield neural nets; continuous time systems; transfer functions; Hopfield type; analogue winner-take-all neural circuit; block diagonal matrix; circuit functioning; computer simulation; linear activation function; sigmoid activation function; Circuits; Computer simulation; Convergence; Hopfield neural networks; Neural networks; Neurons; Pattern classification; Pattern recognition; Piecewise linear techniques; Signal processing; Inhibitory analogue WTA neural circuit; computer simulation results; diagonally stable block diagonal matrix; sigmoid activation function;
Conference_Titel :
Perspective Technologies and Methods in MEMS Design, 2006. MEMSTECH 2006. Proceedings of the 2nd International Conference on
Conference_Location :
Lviv
Print_ISBN :
966-553-517-X
DOI :
10.1109/MEMSTECH.2006.288654
