What are neural computations, and how do they give rise to animal behavior? Can we study behavior to infer the computations performed by the brain? Is it possible to locate the underlying neural circuit and to infer the manner it which it implements these computations?
My research group is interested in brain functions that can be characterized by a computational goal encompassing sensory inputs and motor outputs. Our favorite examples are vocal production and vocal learning, which we study in songbirds using reductionist experimental and theoretical aspproaches.
We perform electrophysiological recordings to read out the neural code in singing birds, we design behavioral experiments to study vocal communication and social learning in bird groups. Currently we are trying to identify the simplest mechanistic equations to describe vocal learning dynamics.
Our work also focuses on translating songbird research to the domain of natural language processing (NLP) and back. We expect much cross-fertilization between these research areas: On the one hand, NLP approaches are readily applicable to birdsong research, because there are many analytical similarities between songbird-syllable and human-word sequences. On the other hand, we recently discovered that songbirds’ strategy of assigning vocal errors during learning is equivalent to word mover's distance, a highly successful computational strategy for document and sentence retrieval. Thus, songbirds have used retrieval strategies millions of years before computational linguists have invented them. Most certainly, many more analogies are about to be discovered
Teaching
INI-431, 227-1045-00 Readings in Neuroinformatics
INI-434, 227-1049-00 Block: Insights Into Neuroinformatics
INI-435, 227-0395-00 Neural Systems