[Connectionist models of memory]

Even if computer science, at its birth, had strong links with the neurosciences, it is today mainly oriented toward efficiency and robustness. For example, memory in a computer has few relationships with memory in a living being. Nevertheless, some domains in computer science are interested in this kind of modelling. In particular, connectionism, whose goal is to elaborate artificial neural networks, uses a formalism for its calculus inspired from calculus in the brain. Different kinds of memory that can be emulated by artificial neural networks, inspired by statistics or biology, are presented here. Their relationships with human memory are discussed together with their tentative interest for the biologist or the therapist.