Closing the Discovery Loop with Low-Dimensional Models and Causal Perturbations

Join us for a talk by Tatiana Engel, Associate Professor at Princeton University. This talk is part of the Kempner Seminar Series, a research-level seminar series that covers topics related to the basis of intelligence in natural and artificial systems.

Patterned photostimulation optogenetics enables unprecedented precision in perturbing neural circuits, yet often yields weak or counterintuitive effects due to the complexity of neural code. I will present modeling approaches that leverage low-dimensional structure in neural population activity to design effective and informative stimulation patterns.

First, I will show that recurrent neural networks (RNNs) trained to replicate neural recordings are not sufficiently constrained by low-dimensional task-driven activity, undermining their ability to reliably predict perturbation effects. Instead, low-dimensional neural response data uniquely constrain a latent low-rank connectivity in RNNs, enabling accurate prediction of perturbation effects within the task-relevant subspace. Second, I will show how we combined optogenetic activation of glomerular patterns in the mouse olfactory bulb with low-dimensional models of glomerular odor responses to relate glomerular activity to olfactory percepts. We discovered that the olfactory code is low-dimensional and inherently structured for efficient odor representations, enabling the design of novel activation patterns with desired perceptual properties.