Simons Center for Systems Biology - Seminar

Abstract: Ensembles of randomly wired networks, constrained by biology (cell types, Dale’s law, geometry), can rival far more detailed models in matching statistics of neural population dynamics, thus revealing the “typical” behavior of high-dimensional brains. Large-scale population recordings now let us test these models quantitatively and identify the minimal constraint sets needed to fit experiments. I’ll present a few examples (unstructured networks, balanced networks, networks with latent variables), where constrained randomness yields precise, quantitative agreement with experiments with no tuning. I will conclude with a brief overview of similar approaches in other areas of the life sciences. The talk will be loosely based on https://arxiv.org/abs/2509.03765