INTEGER APPROXIMATION OF LUKASIEWICZ ARTIFICIAL NEURAL NETWORK

MAJEROVÁ, D.; KUKAL, J.

Abstract

Lukasiewicz algebra enriched by square root function is a good model for data processing background. It enables to realize Lukasiewicz artificial neural network as a tool for the realization of constrained sensitivity systems. The integer approximation of ANN consists of three steps: input conversion to integer domain, exact integer processing and output conversion to real domain. The language for integer processing was designed to provide fast calculations. There are only seven basic instructions. The integer processor is simulated in the Matlab environment. The minimum instruction set is a motivation for realization of fast integer multiprocessor data processing system.

Coresponding author e-mail: dana[dot]majerova[at]vscht[dot]cz

Session: Information Technologies in Automation