Plenary talk: "Correlations and complexity: characterizing neural code dynamics"

FERNANDO MONTANI

Ilhéus (BA)

Congreso; ENFE 17 (Encontro Nacional de Física Estatística); 2017

To understand how sensory information is processed in the brain, we need to investigate how information is represented by the activity of a population of neurons. At the level of how individual neurons process sensory information, it is necessary to develop suitable stochastic models to describe the enormous variability of trains of action potentials. For this, we use the path integral method to determine the analytical solution of the Hodgkin and Huxley equation by considering a non-Gaussian color noise to represent possible feedbacks due to the surrounding network. However at the collective level information can be carried not only on spike firing rates, but also for the temporal structure of the series of shots, and correlations of spikes through neurons. In particular the correlations across neurons are widely found in the cortex, and the evidence shows that the correlations of pairs by themselves are not sufficient to explain the patterns of multi-neuronal firing. We then analyze how the input statistics of neurons can lead to higher order correlations in a neuronal population defining a new scenario for possible synergistic or redundant states in the neural code. Finally we analyzed human recordings during different motor-type activities and quantified the complexity according to the different dynamics.