A new network model for short term memory

S. BOUZAT; D. HANSEL; G. MATO

Mar del Plata

Conferencia; XV Conference on Nonequilibrium Statistical Mechanics and Nonlinear Physics. Medyfinol’06; 2006

Un.N de Mar del Plata, Un. de Los Andes (Chile), UBA.

 Persistent activity in the prefrontal cortex is thought to be thephysical correlate of short term  memory. This activity ischaracterized by the coexistence of two states (background andactive) with different firing rates and highly variable patternsof activity. These features are not satisfactorily described bythe present models. In this work we introduce a new mechanism forthe generation of persistent activity in the cortex. It is basedon disinhibition: a subpopulation of inhibitory neurons becomesactive when the stimulus is strong enough, and it depresses theactivity of another inhibitory subpopulation. The model alsoincludes non-linear inhibitory synapses which can be thought assimplified version of the $GABA_B$ synapses. The system works in abalanced regime, where the low firing rates are achieved as a aresult of cancellation of excitatory an inhibitory inputs. Thisguarantees a strongly variable activity. We also introduce aspatial working memory model, which relies on spontaneous symmetrybreaking to generate a continuum of memory states.