Directions in ML: AI for Adaptive Experiment Design


December 15, 2020


Yisong Yue


California Institute of Technology


Experiment design is hallmark of virtually all research disciplines. In many settings, one important challenge is how to automatically design experiments over large action/design spaces. Furthermore, it is also important for such a procedure to be adaptive, i.e., to adapt to the outcomes of previous experiments. In this talk, I will describe recent progress in using data-driven algorithmic techniques for adaptive experiment design, also known as active learning and Bayesian optimization in the machine learning community. Building upon the Gaussian process (GP) framework, I will describe case studies in personalized clinical therapy and nanophotonic structure design. Motivated by these applications, I will show how to incorporate real-world considerations such as safety, preference elicitation, and multi-fidelity experiment design into the GP framework, with new algorithms, theoretical guarantees, and empirical validation. Time permitting, I will also briefly overview a few other case studies as well.