Anand Gopalakrishnan
Postdoctoral Researcher
E-mail: anand_gopalakrishnan@seas.harvard.edu
Personal website

I am interested in building AI systems that can discover and manipulate compositional structure in the way humans do by efficiently learning reusable concepts and combining them flexibly to solve novel problems. I draw on principles from human perception and cognition, i.e. how humans perceive objects, form abstractions, reason about relations or causes and effects to build structured world models. I completed my PhD in Computer Science at IDSIA working with Jürgen Schmidhuber, where my research was centered around the binding problem. I developed new methods to discover modular and compositional (object-centric) representations of perceptual data with neural networks.