SpecialityScuola Superiore S. Anna, Pisa, IT

Studying functional interaction between brain regions through RNNsUnderstanding the dynamical system governing neuronal activity is crucial for unraveling how the brain performs cognitive functions. Historically, various forms of recurrent neural networks (RNNs) have been suggested as simplified models of the cortex. Recently, however, RNNs' inherent ability to capture temporal dependencies has been leveraged to develop tools for approximating unknown dynamical systems directly by training on observed time-series data (dynamical system reconstruction). Thus it became now possible to build functional models of brain regions executing specific cognitive functions by training latent non-linear RNN models on the neuronal data recorded experimentally. Moreover, when recording from multiple brain regions, dynamical system reconstruction can help us capture interactions between regions and dissect their functional relationship. In this talk, I will discuss the activity of two brain regions heavily involved in the assignment of subjective value to external stimuli, the ventral striatum and the ventral tegmental area, and their interplay during the formation of reward predictive signals in the brain.