Active Learning with Expected Error Reduction
This work addresses the computational bottleneck in active learning for deep learning practitioners, offering a more efficient method with competitive performance, though it is incremental as it builds on existing EER and Bayesian frameworks.
The paper tackles the computational inefficiency of Expected Error Reduction (EER) in active learning for deep neural networks by reformulating it with Bayesian methods, resulting in a method that outperforms most state-of-the-art approaches on standard and WILDS datasets, except for one more costly method in data shift scenarios.
Active learning has been studied extensively as a method for efficient data collection. Among the many approaches in literature, Expected Error Reduction (EER) (Roy and McCallum) has been shown to be an effective method for active learning: select the candidate sample that, in expectation, maximally decreases the error on an unlabeled set. However, EER requires the model to be retrained for every candidate sample and thus has not been widely used for modern deep neural networks due to this large computational cost. In this paper we reformulate EER under the lens of Bayesian active learning and derive a computationally efficient version that can use any Bayesian parameter sampling method (such as arXiv:1506.02142). We then compare the empirical performance of our method using Monte Carlo dropout for parameter sampling against state of the art methods in the deep active learning literature. Experiments are performed on four standard benchmark datasets and three WILDS datasets (arXiv:2012.07421). The results indicate that our method outperforms all other methods except one in the data shift scenario: a model dependent, non-information theoretic method that requires an order of magnitude higher computational cost (arXiv:1906.03671).