Nonmyopic Multifidelity Active Search

TL;DR

This paper introduces a nonmyopic, budget-aware multifidelity active search method that leverages cheaper surrogate data to efficiently identify rare, valuable classes, outperforming existing benchmarks.

Contribution

It proposes a novel, computationally efficient multifidelity active search policy that balances exploration and exploitation dynamically, extending traditional active search to incorporate multiple fidelity levels.

Findings

Significantly better performance than benchmarks on real-world datasets

Effective tradeoff management between exploration and exploitation

Demonstrates the utility of cheaper surrogates in active search

Abstract

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.