Bayesian Active Learning for Structured Output Design
This work addresses the challenge of efficiently designing inputs for complex, multi-output systems in fields like materials science, though it appears incremental as it builds on existing active learning and Gaussian process frameworks.
The paper tackles the problem of finding input parameters to achieve a desired structured output in inverse problems, such as crystal growth modeling, by proposing a Bayesian active learning method with new acquisition functions that incorporate correlations between multiple outputs, and demonstrates effectiveness on synthetic and real data, including a materials informatics application.
In this paper, we propose an active learning method for an inverse problem that aims to find an input that achieves a desired structured-output. The proposed method provides new acquisition functions for minimizing the error between the desired structured-output and the prediction of a Gaussian process model, by effectively incorporating the correlation between multiple outputs of the underlying multi-valued black box output functions. The effectiveness of the proposed method is verified by applying it to two synthetic shape search problem and real data. In the real data experiment, we tackle the input parameter search which achieves the desired crystal growth rate in silicon carbide (SiC) crystal growth modeling, that is a problem of materials informatics.