Abstract

The analysis of neural population activity during behaviour consistently uncovers low-dimensional mathematical structures that capture a large fraction of neural variability. These structures or “neural manifolds” are defined by the dominant patterns of covariation across neurons. Remarkably, recent studies focusing on neural manifolds and the activity within them –the “latent dynamics”– have shed light into questions about cognition, motor control, and learning that had remained elusive when focusing on the activity of individual neurons.

In this talk, I will discuss some of our ongoing efforts to understand the emergence of these population-wide activity patterns using a combination of recordings from monkeys, mice, and humans, as well as computational models. First, I will discuss our recent work on how animals consistently perform the same behaviour on different days, and how motor cortical populations drive a variety of behaviours, which animals learn to adapt in the face of a perturbation even after a few attempts. Next, I will show that, even if each animal has a brain that is unique, individuals from the same species that are engaged in the same behaviour share preserved latent dynamics. This suggests that evolution by natural selection may act upon the brain by specifying circuits that enable the generation of the appropriate latent dynamics to support species-specific behaviours.

Thus, the study of neural manifolds and their associated latent dynamics provides insights into how individual animals consistently and flexibly perform a variety of behaviours, and may enable comparative studies across groups of individuals from the same of even different species.