| ABSTRACT |
The accurate and efficient measurement of observations is critical for empirical scientific inquiry. To ensure measurement episodes are also optimal, and thereby maximize inference, there has been a growing interest by researchers in the design of adaptive experiments that lead to rapid accumulation of information about the phenomenon under study with the fewest possible experimental measurements. Further, in an autonomous adaptive experiment, the design and execution of an experiment are performed with minimal intervention by scientists after the experiment is initiated. Building on the foundational work in statistics and machine learning, our lab has developed and applied our own adaptive methodologies, dubbed adaptive design optimization (ADO) and Gaussian Process Active Learning (GPAL). In this talk, I will provide an overview of these closed-loop experimentation algorithms, along with example applications in cognitive science and materials science. |
