Discovering General-Purpose Active Learning Strategies
This work addresses the challenge of reducing annotation costs in machine learning for practitioners, though it appears incremental as it builds on existing active learning frameworks with a reinforcement learning approach.
The authors tackled the problem of discovering general-purpose active learning strategies that are transferable across domains and compatible with various machine learning models, achieving results that consistently outperform state-of-the-art baselines on multiple unrelated domains.
We propose a general-purpose approach to discovering active learning (AL) strategies from data. These strategies are transferable from one domain to another and can be used in conjunction with many machine learning models. To this end, we formalize the annotation process as a Markov decision process, design universal state and action spaces and introduce a new reward function that precisely model the AL objective of minimizing the annotation cost. We seek to find an optimal (non-myopic) AL strategy using reinforcement learning. We evaluate the learned strategies on multiple unrelated domains and show that they consistently outperform state-of-the-art baselines.