Minimax Active Learning
This work provides a more robust active learning strategy for practitioners in machine learning by mitigating the weaknesses of existing uncertainty-based and diversity-based approaches.
The paper addresses label-efficient learning by developing a semi-supervised minimax entropy-based active learning algorithm. This method combines uncertainty and diversity in an adversarial manner, achieving superior performance over state-of-the-art methods on various image classification and semantic segmentation benchmarks.
Active learning aims to develop label-efficient algorithms by querying the most representative samples to be labeled by a human annotator. Current active learning techniques either rely on model uncertainty to select the most uncertain samples or use clustering or reconstruction to choose the most diverse set of unlabeled examples. While uncertainty-based strategies are susceptible to outliers, solely relying on sample diversity does not capture the information available on the main task. In this work, we develop a semi-supervised minimax entropy-based active learning algorithm that leverages both uncertainty and diversity in an adversarial manner. Our model consists of an entropy minimizing feature encoding network followed by an entropy maximizing classification layer. This minimax formulation reduces the distribution gap between the labeled/unlabeled data, while a discriminator is simultaneously trained to distinguish the labeled/unlabeled data. The highest entropy samples from the classifier that the discriminator predicts as unlabeled are selected for labeling. We evaluate our method on various image classification and semantic segmentation benchmark datasets and show superior performance over the state-of-the-art methods.