The Physics & Astronomy Colloquium Series and Quantitative Biology Institute presents Tatiana Engel of Princeton Neuroscience Institute, Princeton University, discussing "Unifying Single-neuron Tuning, Manifold Geometry, and Population Dynamics" on January 16.

Abstract: Responses of single cortical neurons are complex and heterogeneous, yet they often form low-dimensional manifolds in the population state space. It is commonly assumed that neural computations arise from low-dimensional population dynamics, while functional properties of individual neurons are not interpretable.

I will present our recent work that bridges single-neuron tuning with manifold geometry and population dynamics. First, we developed a flexible approach for simultaneously inferring single-trial population dynamics and tuning functions of individual neurons to the latent population state. Applied to spike data recorded during decision-making, our model revealed that all neurons encode the same dynamic decision variable, and heterogeneous firing rates result from diverse tuning of single neurons to this decision variable. Second, we prove that in firing-rate recurrent networks, when population dynamics are strictly low-dimensional, single-neuron responses necessarily cluster into functional types, with the number of response types equal to the linear dimensionality of the neural manifold. We confirm these predictions in recurrent networks trained on cognitive tasks and brain-wide neural recordings from mice during decision-making. These results reveal fundamental constraints on how recurrent circuits represent information.