I am a theoretical physicist exploring learning in physical systems, particularly in the context of biologically inspired and physically realizable learning rules. My research at the University of Pennsylvania (Liu and Balasubramanian groups) and formerly at the University of Chicago (Murugan group) established such learning rules for flow networks, elastic networks, and self-folding origami, highlighting their potential as designer multi-functional, dynamically controlled metamaterials.
I am specifically interested in the analogies between learning in physical networks and its counterparts in neuroscience and computational machine learning. These fundamental connections suggest the use of physical learning to train new classes of smart metamaterials and machines. Moreover, physical learning models help shed light on how abstract learning theory is modified by real-world physical constraints.
In addition to learning in physical systems, I like to think about connections between reinforcement learning and statistical physics, and especially the notion of curious exploration in complex environments. In earlier work I studied dynamics of dense granular suspensions (Zhang & Jaeger groups, UChicago), isobaric ensembles of supercooled liquid models (Eisenberg group, TAU), self-avoiding random walks, and atmospheric modeling (Soreq NRC, Israel).