Targeted Active Learning for Bayesian Decision-Making
Abstract: Active learning is usually applied to acquire labels of informative data points in supervised learning, to maximize accuracy in a sample-efficient way. However, maximizing the supervised learning accuracy is not the end goal when the results are used for decision-making, for example in personalized medicine or economics. We argue that when acquiring samples sequentially, the common practice of separating learning and decision-making is sub-optimal, and we introduce an active learning strategy that takes the down-the-line decision problem into account. Specifically, we adopt a Bayesian experimental design approach, in which the proposed acquisition criterion maximizes the expected information gain on the posterior distribution of the optimal decision. We compare our targeted active learning strategy to existing alternatives on both simulated and real data and show improved performance in decision-making accuracy.
Submission Length: Regular submission (no more than 12 pages of main content)
Previous TMLR Submission Url: https://openreview.net/forum?id=VldyVuH0eX
Changes Since Last Submission: The manuscript has been revised to (i) explicitly detail the link between Bayesian optimization and the proposed active learning method, and (ii) more clearly define the scenarios in which the problem formulation is particularly relevant for real-world applications. In addition to these changes, we have rewritten the manuscript to address the reviewers' comments.
Regarding the connection to Bayesian optimization (i), we have added a section titled “Entropy-Search Multi-Fidelity Bayesian Optimization” to the Related Work section. This new section reveals that, when the problem is reformulated as a multi-fidelity Bayesian optimization, there is a link between MF-PES (a multi-fidelity version of the Predictive Entropy Search method by Hernández-Lobato et al. (2014)) and the proposed active learning method.
Regarding the motivation for the work (ii), we have added a paragraph to the Introduction section. This paragraph presents a key motivational example for the work: “...the active learning task addressed in our paper involves identifying the most informative patient-treatment pairs from a predefined pool, which may include sources such as electronic health records (EHR). Accessing data from these records involves stringent legal justification due to privacy concerns, mandating that a physician must have a legitimate reason for retrieval and use in patient treatment.”
Assigned Action Editor: ~Benjamin_Guedj1
Submission Number: 2135
Loading