Experiments with Skinner, Pavlov and Hume Synapses: Learning to Discriminate Letters

Abstract

It is assumed that learning in the brain happens by adjusting synaptic efficacies between neurons. How the synapses are altered are no yet certain. Hokland argues that the present dominant hypothesis for synaptic altering are unrealistic models for learning in the human brain. In Connectology - Research Programme for Brain-Psychology Hokland describes a neural network theory where he presents three mechanisms for synaptic learning, based on the work of A. Harry Klopf. The mechanisms in this thesis are called Skinner, Pavlov and Hume. These three learning mechanisms will be implemented in an artificial neural network. As input to the artificial neural network I will use an input image of letters. The task at hand will be to find out if the learning mechanisms in this specific network will be able to discriminate between the letters.