Symposium SB08: Bio-Based Polymers and Composites for Sustainable Manufacturing
Thank you!

Symposium DS04: Accelerating Data-Driven Materials Research for Energy Applications

Roman Garnett, Washington University in St. Luis

Active Search for Efficient Discovery of Visible Light-Activated Azoarene Photoswitches with Long Half-Lives

Written by Matthew Nakamura and Md Afzalur Rab

A photoswitch is a kind of molecule that can transform its structure, geometry, and chemical properties when the molecule is excited with electromagnetic radiation. In this talk, Roman Garnett from Washington University of St. Luis discussed an uncommon statistical method called “active search”— a variant of Bayesian optimization—to discover potential azoarene photoswitches.

In statistics, sequential analysis is a kind of hypothetical testing where the sample size is not known in advance. So the data are tested as they are collected and sampling is stopped according to some predefined rules. Bayesian optimization is a type of sequential analysis, where the size of samples are taken as undefined. Bayesian optimization is useful to test black box type functions where only inputs and outputs are known but intermediate processes are unknown.

Garnett highlighted the challenge of identifying rare, valuable subsets of photoswitches within a vast pool of possibilities and introduced nonmyopic-yet-efficient policies to address this complexity. Garnett emphasized the significance of active search in optimizing the discovery process, particularly where optimizing specific properties is crucial to overall performance of new materials. The discussion also showcased a successful application of active search to the discovery of photoswitches with desirable properties. Overall, Garnett’s talk provided valuable insights into the potential of intelligent experimental design utilizing active search to enhance the efficiency of discovery processes in various scientific domains.


The comments to this entry are closed.