BCAM Scientific Seminar: An information theoretical approach to nonequilibrium neural computation

POSTPONED TO TUESDAY 7 JUNE 2022

Location: BCAM Seminar Room and online

Abstract:
Effective neural information processing entails flexible architectures integrating multiple sensory streams that vary in time with internal and external events. Physically, neural computation is, in a thermodynamic sense, an out-of-equilibrium, non-stationary process that changes dynamically. Cognitively, nonequilibrium neural activity results in dynamic changes in sensory streams and internal states. Classical neuroscience theory focuses on stationary, equilibrium information paradigms (e.g., efficient coding theory), which often fail to describe the role of nonequilibrium fluctuations in neural processes. In consequence, there is a pressing demand for mathematical tools to study the dynamics of large-scale, non-equilibrium networks systems and to analyse high-dimensional datasets recorded from them. In this talk, we will introduce a unifying framework based on information theory covering different aspects of nonequilibrium neural computation: 

(i) Although nonequilibrium network properties are receiving increasing attention from neuroscience and biological sciences communities, the thermodynamics of large-scale nonequilibrium systems and their phase transitions are started to b studied only very recently. Inspired by the success of the equilibrium Ising model in investigating disordered systems in the thermodynamic limit, we will inspect the nonequilibrium thermodynamics of the asymmetric Sherrington-Kirkpatrick model as a prototypical model of large-scale nonequilibrium processes
(ii) Neural activity is often found to self-organize near critical regimes at which their fluctuations are maximized, making it difficult to characterize its behaviour due to the explosion of activity patterns. We will study how to approximate the behaviour and thermodynamic properties of non-equilibrium neural populations using approximation techniques based in information geometry and information theoretical arguments allowed by the flat geometry associated with the Kullback-Leibler divergence. This results in an unified framework including approximations from mean-fields to belief propagation, with applications to model inference from data recordings.
(iii) There is an emerging interest in the design of intelligent systems based in advances in nonequilibrium systems theories and Bayesian inference theory. We will study how the information theoretical methods above can be used to design adaptive systems exploiting properties of nonequilibrium phase transitions.

Link to the session: 
https://us06web.zoom.us/j/82235486532?pwd=MkxOMkFmVjlUTlNHYjU4ZGhsSy9GZz09