Past Event

Dr. Sylvain Chartier, of the University of Ottawa will be giving a colloquium talk on Thursday, April 3rd, at noon, in room 2203 of Dunton Tower. Refreshments will be available before the talk. All are welcome.

General Bidirectional Associative Memories

Being able to recognize and recall patterns of various natures and in different contexts is something that human beings accomplish routinely and with little effort. The learning of associations is one of the most basic and important processes in cognition. It is a phenomenon which has been philosophized about for 100s of years and studied empirically for just over the last 100 years. The essence of associationism is characterized within the neural networks or connectionist modeling framework in terms of recurrent and bidirectional associative memories. These brain-inspired, artificial neural network approaches offer the ability to develop attractors for each learned pattern through feedback connections. It also exhibits great stability and adaptability with regards to noise and pattern degradation, and can perform generalization tasks. In this presentation, focus will be drawn on associative memories properties for different kinds of associations. At first, simulations of the classic case of one-to-one association will be presented in context of fixed points and chaotic attractors. Then many-to-one association, temporal and complex associations will be shown. In all cases, the same learning and transmission function are used. The only modification concerned the network topology. This property gives associative memory a high internal consistency and therefore constitutes a step toward unifying various classes of neural networks within a general architecture.