A neural paradigm for controlling autonomous systems with reflex behaviour and learning capability
In this paper we present a neural paradigm for controlling the reflex behaviour of autonomous systems which are able to modify their behaviour by interaction with the environment. This paradigm incorporates the ideas expressed by Russell  about how to model the living being's reflex behaviour. In this paradigm a new type of connection is introduced: the so called high order Or connection. Learning is local and unsupervised, i.e., the change in the weight of a connection takes place as a consequence of its activation. We present two functions to update the weights which incorporate the forgetting capability. Some topologies have been simulated to provide the basic capabilities such as inhibition, stimuli association an reinforcement.
Unable to display preview. Download preview PDF.
- .-S. B. Russell, “A practical device to simulate the working of nervous discharges”, Journal of Animal Behaviour, 3, (15) (1913), pp. 15–35.Google Scholar
- .-R.P. Lippmann, “An Introduction to computing with neural nets”, IEEE ASSP Magazine, April 1987, pp. 4–22.Google Scholar
- .-C.L. Giles and T. Maxwell, “Learning, invariance, and generalization in high-order neural networks”. Applied Optics, 26, (23), (1987), pp. 4972–4978.Google Scholar
- .-M.L. Minsky and S. Papert, Perceptrons, MIT Press, Cambrige, MA, 1969.Google Scholar
- .-J. Mira, A.E. Delgado, J.R. Alvarez, A.P. Madrid, and M. Santos, “Towards more realistic self contained models of neurons: high-order, recurrence and local learning”, in J. Mira, J. Cabestany, A. Prieto eds, New Trends in Neural Computation. Lecture Notes in Computer Science 686, Springer-Verlag, 1993, pp. 55–62.Google Scholar