A0755
Title: Modeling and active learning for experiments with quantitative-sequence factors
Authors: Abhyuday Mandal - University of Georgia (United States) [presenting]
Abstract: A new type of experiment which targets finding optimal quantities of a sequence of factors is drawing much attention in medical science, bio-engineering and many other disciplines. Such studies require simultaneous optimization for both quantities and sequence orders of several components, which is defined as a new type of factors: quantitative-sequence (QS) factors. Due to the large and semi-discrete solution spaces in such experiments, it is non-trivial to efficiently identify the optimal (or near-optimal) solutions using only a few experimental trials. To address this challenge, we propose a novel active learning approach, named QS-learning, to enable effective modeling and efficient optimization for experiments with QS factors. The QS-learning consists of three parts: a novel mapping-based additive Gaussian process (MaGP) model, an efficient global optimization scheme (QS-EGO), and a new class of optimal designs (QS-design) for collecting initial data. Theoretical properties of the proposed method are investigated and techniques for optimization using analytical gradients are developed. The performance of the proposed method is demonstrated via a real drug experiment on lymphoma treatment and several simulation studies.
