Edinburgh Napier University

  The paper presents a new approach to the interpretation of chaotic neural activity. It is suggested that such activity forms neural objects represented as spatio-temporal firing patterns and the neural computations are performed through the interaction of such neural objects. To introduce the concepts of the proposed interpretation the so-called Sierpinski brain is described. This model brain is composed of simple neural networks, which produce Sierpinski triangles as their chaotic spatio-temporal firing pattern. It is shown how such Sierpinski triangles can be used to perform general approximation, prediction and classification tasks. This paper discusses how learning occurs in the context of the Sierpinski brain and how the presented ideas can be interpreted in the context of biological brains.