A world according to artificial neural networks

This paper presents results from a preliminary study in the field of artificial neural networks (ANN). The overall aim of our work relates to the field of cognitive science. In this wider framework we try to investigate, reason about, and model cognitive processes in order to obtain a better understanding of the major processing device involved - the human brain. In terms of content this paper presents a novel ANN learning approach. Note that through-out the paper we assume supervised learning. In contrast to the classical ANN learning approach where an ANN algorithm alters an initial random weight assignment until a reasonable solution to a problem is obtained this approach does not alter the initial random weight assignment at all, but provides a solution to the problem by transforming the actual input data. The approach is applied to perceptrons and adalines and its quality is demonstrated on simple classification problems.