Active learning using region-based sampling
This work addresses active learning for general data in metric spaces, providing a novel approach but is incremental in its method development.
The authors tackled the problem of active learning in metric spaces by introducing a region-based sampling algorithm that identifies conflicting label neighborhoods to target queries, achieving label complexity bounds without data assumptions.
We present a general-purpose active learning scheme for data in metric spaces. The algorithm maintains a collection of neighborhoods of different sizes and uses label queries to identify those that have a strong bias towards one particular label; when two such neighborhoods intersect and have different labels, the region of overlap is treated as a ``known unknown'' and is a target of future active queries. We give label complexity bounds for this method that do not rely on assumptions about the data and we instantiate them in several cases of interest.