Roman Garnett, Assistant Professor of Computer Science & Engineering at Washington Univesity, will present.

Active Search for Automating Scientific Discovery

Abstract: With the explosive growth of the internet and availability of increasingly cheap large-scale storage, amassing huge datasets is becoming commonplace. In many real-world scenarios, however, conducting detailed analysis of an identified data point is very expensive, requiring human intervention or a costly experiment. In such situations, it is critical that we allocate limited resources effectively. In "active" machine learning, we consider how to collect the most-useful data to achieve our goals effectively with a limited budget for labeling. We will introduce active machine learning and discuss a particular important setting: "active search," where we seek to discover rare, valuable points from a large database of alternatives. We will discuss the surprising difficulty of this problem and introduce efficient, nonmyopic polices to solve it, demonstrating our method on large-scale drug and materials discovery simulations.

Bio: Roman Garnett has been an assistant professor in the Computer Science & Engineering department since 2015. His research seeks to accelerate scientific discovery through intelligent Bayesian algorithms for sequential experimental design. He recently received the NSF CAREER award for this effort.
 

Roman Garnett, Assistant Professor of Computer Science & Engineering at Washington Univesity, will present.

Active Search for Automating Scientific Discovery

Abstract: With the explosive growth of the internet and availability of increasingly cheap large-scale storage, amassing huge datasets is becoming commonplace. In many real-world scenarios, however, conducting detailed analysis of an identified data point is very expensive, requiring human intervention or a costly experiment. In such situations, it is critical that we allocate limited resources effectively. In "active" machine learning, we consider how to collect the most-useful data to achieve our goals effectively with a limited budget for labeling. We will introduce active machine learning and discuss a particular important setting: "active search," where we seek to discover rare, valuable points from a large database of alternatives. We will discuss the surprising difficulty of this problem and introduce efficient, nonmyopic polices to solve it, demonstrating our method on large-scale drug and materials discovery simulations.

Bio: Roman Garnett has been an assistant professor in the Computer Science & Engineering department since 2015. His research seeks to accelerate scientific discovery through intelligent Bayesian algorithms for sequential experimental design. He recently received the NSF CAREER award for this effort.