Flexible working memory model with two types of plasticity

As the brain system for short-term storage and manipulation of information working memory (WM) plays an important role in complex cognitive tasks. Most WM models contain pre-formed content-specific structures of neurons for loading specific items into memory, which lacks of flexibility of natural WM. On the other hand, long-term synaptic plasticity such as spike-timing-dependent plasticity (STDP) seems to be the best mechanism to form content depended functional and structural neural ensembles without preliminary tuning. In this work, we develop a flexible model of WM based on synaptic theory considering two types of plasticity: short-term plasticity and STDP without pre-formed neuronal clusters. The numerical simulations show the formation of such populations as an STDP-driven response to an external stimuli. The model is implemented in a recurrent network of leaky integrate-and-fire neurons in excitable mode. We demonstrate formation of neuronal clusters encoding items in the WM model due to STDP and hold and reactivated by short-term plasticity mechanisms. The relevance of the proposed WM model is confirmed by the similarity of the WM capacity dependence on synaptic facilitation and depression time constants with previous results obtained earlier for the model with only short-term plasticity. Increasing the STDP learning rate parameter resulted in increased WM capacity on average.