Nonmyopic Multifidelity Active Search
This work addresses the challenge of efficiently identifying rare items with limited labeling budgets, offering a practical solution for domains where multiple data sources are available, though it appears incremental by extending classical policies.
The paper tackles the problem of active search for rare, valuable items by introducing a multifidelity approach that uses cheaper surrogates like simulations alongside expensive oracles, resulting in significantly better performance than benchmarks on real-world datasets.
Active search is a learning paradigm where we seek to identify as many members of a rare, valuable class as possible given a labeling budget. Previous work on active search has assumed access to a faithful (and expensive) oracle reporting experimental results. However, some settings offer access to cheaper surrogates such as computational simulation that may aid in the search. We propose a model of multifidelity active search, as well as a novel, computationally efficient policy for this setting that is motivated by state-of-the-art classical policies. Our policy is nonmyopic and budget aware, allowing for a dynamic tradeoff between exploration and exploitation. We evaluate the performance of our solution on real-world datasets and demonstrate significantly better performance than natural benchmarks.