Fast computation of a gated dipole field.

Abstract

We address the need to develop efficient algorithms for numerical simulation of models, based in part or entirely on adaptive resonance theory. We introduce modifications that speed up the computation of the gated dipole field (GDF) in the Exact ART neural network. The speed increase of our solution amounts to at least an order of magnitude for fields with more than 100 gated dipoles. We adopt a 'divide and rule' approach towards the original GDF differential equations by grouping them into three categories, and modify each category in a separate way. We decouple the slow-dynamics part - the neurotransmitters from the rest of system, solve their equations analytically, and adapt the solution to the remaining fast-dynamics processes. Part of the node activations are integrated by an unsophisticated numerical procedure switched on and off according to rules. The remaining activations are calculated at equilibrium. We implement this logic in a Generalized Net (GN) - a tool for parallel processes simulation which enables a fresh look at developing efficient models. Our software implementation of generalized nets appears to add little computational overhead.