Computational Neuroscience, Institute of Physiology, University of Bern, Switzerland
Wednesday 4 May 2005
B10 Seminar Room
Alexandra House, 17 Queen Square, LONDON, WC1N 3AR
Learning on Multiple Time Scales: Theory and In Vivo Experiments
Storing memories of ongoing, everyday experiences requires a high degree of plasticity, but retaining these memories demands protection against changes induced by further activity and experience. Models in which memories are stored through switch-like transitions in synaptic efficacy are good at storing but bad at retaining memories if these transitions are likely, and poor at storage but good at retention if they are unlikely. We construct and study a model in which each synapse has a cascade of states with different levels of plasticity, connected by metaplastic transitions. This cascade model combines high levels of memory storage with long retention times and significantly outperforms alternative models. As a result, we suggest that memory storage requires synapses with multiple states exhibiting dynamics over a wide range of timescales. We then analyze and model the behavioural and the neural data of the experiment performed by W.F. Asaad, G. Rainer and E.K. Miller (Neuron, 1998) in which two monkeys are trained to learn and forget visuo-motor associations. Synaptic plasticity on multiple timescales is shown to be crucial to allow the model to quickly forget the old cue-motor response associations, as well as to capture the statistics of correct associations across many blocks of trials. The cascade model has been developed with L.F. Abbott and P.J. Drew. The model of the experimental data with X.-J. Wang, W.F. Asaad and E.K. Miller.