Title: A data representation from neuroscience

Abstract: An intriguing representation for data, “expand-and-sparsify”, appears in the olfactory system of the fly and the sensory systems of several other organisms. It maps an input vector to a much higher dimensional sparse representation, using a random linear transformation followed by winner-take-all thresholding.

I’ll show that this representation has a variety of formal algorithmic and statistical properties, including
(1) locality preservation, (2) universal approximation, and (3) adaptivity to manifold structure, that make it attractive for machine learning and statistics. In particular, mimicking the fly’s circuitry yields algorithms for similarity search and for novelty detection that have provable guarantees as well as being effective in practice.

This talk is based on work with Saket Navlakha (Salk Institute), Chuck Stevens (Salk Institute), and Chris Tosh (Columbia).

Biography: Sanjoy Dasgupta is currently on sabbatical at the Institute for Advanced Study, away from his usual position as Professor of Computer Science and Engineering at UC San Diego. He works on algorithms for machine learning, with a particular focus on unsupervised and minimally supervised learning. He is author of a textbook on “Algorithms” (with Christos Papadimitriou and Umesh Vazirani).