AUTONOMOUS LEARNING WITH COMPLEX DYNAMICS

Traditionally, associative memory models are based on point attractor dynamics, where a memory state corresponds to a stationary point in st ate space. However, biological neural systems seem to display a rich a nd complex dynamics whose function is still largely unknown. We use a neural network model of the olfactory cortex to investigate the functi onal significance of such dynamics, in particular with regard to learn ing and associative memory. The model uses simple network units, corre sponding to populations of neurons connected according to the structur e of the olfactory cortex. All essential dynamical properties of this system are reproduced by the model, especially oscillations at two sep arate frequency bands and aperiodic behavior similar to chaos. By intr oducing neuromodulatory control of gain and connection weight strength s, the dynamics can change dramatically, in accordance with the effect s of acetylcholine, a neuromodulator known to be involved in attention and learning in animals. With computer simulations we show that these effects can be used for improving associative memory performance by r educing recall time and increasing fidelity. The system is able to lea rn and recall continuously as the input changes, mimicking a real worl d situation of an artificial or biological system in a changing enviro nment. (C) 1995 John Wiley & Sons, Inc.