Past talk · AI/ML Seminar Series
Variational Methods for Bayesian Optimal Experimental Design
PhD Student, Department of Computer Science, University of California, Irvine
- Date & time
- Monday, February 13, 2023 · 1:00 PM
- Location
- Donald Bren Hall 4011
Abstract
Bayesian optimal experimental design is a sub-field of statistics focused on developing methods to make efficient use of experimental resources. Any potential design is evaluated in terms of a utility function, such as the expected information gain (EIG); unfortunately, under most circumstances the EIG is intractable to evaluate. In this talk we build off successful variational approaches, which optimize a parameterized variational model with respect to bounds on the EIG. We present a novel neural architecture that allows experimenters to optimize a single variational model that can estimate the EIG for potentially infinitely many designs. We demonstrate the effectiveness of our technique on generalized linear models, showing that our method greatly improves accuracy over existing approximation strategies with far better sample efficiency.
About the speaker
Noble Kennamer recently completed his PhD at UC Irvine under Alexander Ihler, where he worked on variational methods for optimal experimental design and applications of machine learning to the physical sciences. He is starting as a Research Scientist at Netflix.