Spoločný seminár umelej inteligencie

Thu 12 Feb. 2009, 14:00
D220 (FIIT STU)

Title: How neurons learn: insights from computational models
Speaker: Ľubica Beňušková, University of Otago, Dunedin, New Zealand

All of our cognitive functions, including our sense of identity, are 
underpinned by what we have learned and what we can remember. To 
understand learning and memory formation in the brain, it is necessary to 
study how neurons change their mutual connections in response to synaptic 
stimulation, i.e. phenomenon of synaptic plasticity. I will demonstrate the 
predictive power of computational modeling by presenting a novel model 
of heterosynaptic plasticity in the hippocampal dentate gyrus, which is the 
brain structure crucially involved in formation of long-term memories. This 
model combines computational properties of both metaplasticity and spike 
timing-dependent plasticity. As a result it can reproduce (1) homosynaptic 
long-term potentiation of the stimulated input, and (2) heterosynaptic 
long-term depression of the unstimulated input, as observed in real 
experiments.