Adaptive Data Augmentation for Contrastive Learning
This work addresses a domain-specific issue in unsupervised computer vision by improving data efficiency in contrastive learning, though it is incremental as it builds on existing frameworks like MoCo v2.
The paper tackles the problem of fixed data augmentation in contrastive learning, which degrades representation quality due to evolving network needs, and proposes AdDA to adaptively adjust augmentations, achieving a +1.11% improvement on ImageNet-100 classification with MoCo v2.
In computer vision, contrastive learning is the most advanced unsupervised learning framework. Yet most previous methods simply apply fixed composition of data augmentations to improve data efficiency, which ignores the changes in their optimal settings over training. Thus, the pre-determined parameters of augmentation operations cannot always fit well with an evolving network during the whole training period, which degrades the quality of the learned representations. In this work, we propose AdDA, which implements a closed-loop feedback structure to a generic contrastive learning network. AdDA works by allowing the network to adaptively adjust the augmentation compositions according to the real-time feedback. This online adjustment helps maintain the dynamic optimal composition and enables the network to acquire more generalizable representations with minimal computational overhead. AdDA achieves competitive results under the common linear protocol on ImageNet-100 classification (+1.11% on MoCo v2).