Abstract

Bayesian optimization is a popular algorithm for optimizing low-dimensional functions in a data-efficient manner. In this talk, I will discuss my practical experience with Bayesian optimization when applied to robotic applications. Along this journey, I will introduce several interesting real-world problem settings that I have encountered, and the corresponding algorithms designed as a result, as well as the new insights gained. To conclude, I will briefly discuss some future challenges and potential directions.

Notes

• Link to the paper: http://proceedings.mlr.press/v108/aglietti20a

• Javier González Hernández is a Principal Researcher at Microsoft Research Cambridge. Personal website can be found here.