A Nonmyopic Approach to Cost-Constrained Bayesian Optimization at UAI

The Conference on Uncertainty in Artificial Intelligence (UAI) has been established since 1985, this year it will be held from July 17-29 as a virtual event. UAI is one of the foremost international conferences related to knowledge representation, learning and reasoning in uncertainty in machine learning.

Eric Lee, SigOpt Research Engineer,  David Eriksson, Facebook Research Scientist, and Valerio Perrone and Matthias Seeger, Amazon Applied Research Scientists, collaborated on a paper on cost-constrained Bayesian optimization that has been accepted at UAI. Bayesian optimization budgets are normally given in iterations while assuming each evaluation has the same cost. Cost-constrained Bayesian optimization measures convergence with alternative cost metrics such as time, money or energy. 

Have questions or would like to learn more about the paper? 

• On July 27th, join the lightning talk at 9:20pm PT, or catch us in the poster session right after at 10:00pm PT. Learn more about the paper here. 
• Registration is free for students before July 2. Get your tickets here: https://www.auai.org/uai2021/ 

We hope to see you at the event!