The Biophysical Principles Underlying Computation in Neural Substrates

Iran Roman will presenting and dicussing the following:

Abstract

This presentation offers an overview of NeuroAI’s development, starting with seminal biophysical models such as the Hodgkin-Huxley equations, which elucidate the ionic mechanisms underlying neuronal action potentials. We then discuss the Wilson-Cowan model, capturing the interactions between excitatory and inhibitory neuronal populations. Advancements in the 2010s used recurrent neural networks to paralleling complex computations observed in both macaque and RNNs. We then discuss dynamical systems approaches, emphasizing their efficacy in modeling the temporal evolution of human brain activity, including unsupervised Hebbian Learning. Finally, we explore bifurcation theory’s role in identifying critical parameters that govern perception-action coupling within neural substrates, illustrating how minor variations can lead to significant shifts in neural computation.

Related Work

(Mante et al., 2013)

(Zemlianova et al., 2024)

(Roman et al., 2023)

(Driscoll et al., 2024)