Dynamic control of neural manifolds

Abstract

In the central nervous system, sequences of neural activity form trajectories on low dimensional neural manifolds. The neural computation underlying flexible cognition and behavior relies on dynamic control of these structures. For example different tasks or behaviors are represented on different subspaces, requiring fast timescale subspace rotation to move from one behavior to the next. For flexibility in a particular behavior, the neural trajectory must be dynamically controllable within that behaviorally determined subspace. To understand how dynamic control of neural trajectories and their underlying subspaces may be implemented in neural circuits, we first characterized the relationship between features of neural activity sequences and aspects of the low dimensional projection. Based on this, we propose neural mechanisms that can act within local circuits to modulate activity sequences thereby controlling neural trajectories in low dimensional subspaces. In particular, we show that gain modulation and transient synaptic currents control the speed and path of neural trajectories and clustered inhibition determines manifold orientation. Together, these neural mechanisms may enable a substrate for fast timescale computation on neural manifolds.