Greedy Policy Search: A Simple Baseline for Learnable Test-Time Augmentation
This addresses the need for optimized test-time augmentation in machine learning, offering a simple baseline method for practitioners.
The paper tackled the problem of learning test-time augmentation policies, demonstrating that their method achieves superior predictive performance on image classification, better in-domain uncertainty estimation, and improved robustness to domain shift.
Test-time data augmentation$-$averaging the predictions of a machine learning model across multiple augmented samples of data$-$is a widely used technique that improves the predictive performance. While many advanced learnable data augmentation techniques have emerged in recent years, they are focused on the training phase. Such techniques are not necessarily optimal for test-time augmentation and can be outperformed by a policy consisting of simple crops and flips. The primary goal of this paper is to demonstrate that test-time augmentation policies can be successfully learned too. We introduce greedy policy search (GPS), a simple but high-performing method for learning a policy of test-time augmentation. We demonstrate that augmentation policies learned with GPS achieve superior predictive performance on image classification problems, provide better in-domain uncertainty estimation, and improve the robustness to domain shift.