How do we utilize previously acquired knowledge to guide behavior in novel environments?
A fundamental aspect of this question is how we recognize when a new context resembles one we’ve encountered before. Very little is known about how humans and other animals compose elements of past learning to solve similar problems in new situations. To explore these and related questions, I recently joined the Allen Institute for Neural Dynamics. My group will utilize artificial network simulations and reverse engineering with close ties to experimental groups collecting behavioral and neural data. We will examine how previous learning shapes behavior in novel environments.
This work is informed by my postdoctoral training with Krishna V. Shenoy and David Sussillo in the Neural Prosthetic Systems Laboratory (NPSL) at Stanford University, where I reverse-engineered recurrently connected neural networks to uncover shared dynamical motifs across multiple related computations.
My graduate training with Chris Harvey at Harvard University shapes my thinking about structures of knowledge in the brain. We discovered that neural activity patterns, correlated with sensation and action, often aren’t stable. Instead, they undergo large-scale changes over days and weeks — a phenomenon now called representational drift.