Stepwise Model Selection for Sequence Prediction via Deep Kernel Learning

• DOI: 10.48550/arxiv.2001.03898
• Publication Date: 2020-01-01
• Authors: Yao Zhang; Daniel Jarrett; Mihaela van der S...
• Keywords: Sequence (biology); Kernel (algebra)

An essential problem in automated machine learning (AutoML) is that of model selection. A unique challenge the sequential setting fact optimal itself may vary over time, depending on distribution features and labels available up to each point time. In this paper, we propose a novel Bayesian optimization (BO) algorithm tackle selection setting. This accomplished by treating performance at time step as its own black-box function. order solve resulting multiple function jointly efficiently, exploit potential correlations among functions using deep kernel (DKL). To best our knowledge, are first formulate stepwise (SMS) for sequence prediction, design demonstrate an efficient joint-learning purpose. Using real-world datasets, verify proposed method outperforms both standard BO multi-objective algorithms variety prediction tasks.