Implementing a "psychophysical" pattern classifier in a decrementing network

Abstract

The biological systems work starting from the sensations produced by the stimulis received from the environment. The relationship between the physical magnitude of the stimulus and the sensation produced has beeen the scope of the Psychophysics. We describe a Pattern classifier, which uses the sensations produced by the patterns and the inputs to be classified. We develope the classifying algorithm which is based on a logarithmic distance, and we show how it can be implemented in a network of simple devices. Finally, we compare the performance of this classifier versus a Gaussian classifier in a specific task.