Gradient Regularized Contrastive Learning for Continual Domain Adaptation

Abstract

Human beings can quickly adapt to environmental changes by leveraging learning experience. However, adapting deep neural networks to dynamic environments by machine learning algorithms remains a challenge. To better understand this issue, we study the problem of continual domain adaptation, where the model is presented with a labelled source domain and a sequence of unlabelled target domains. The obstacles in this problem are both domain shift and catastrophic forgetting. We propose Gradient Regularized Contrastive Learning (GRCL) to solve the obstacles. At the core of our method, gradient regularization plays two key roles: (1) enforcing the gradient not to harm the discriminative ability of source features which can, in turn, benefit the adaptation ability of the model to target domains; (2) constraining the gradient not to increase the classification loss on old target domains, which enables the model to preserve the performance on old target domains when adapting to an in-coming target domain. Experiments on Digits, DomainNet and Office-Caltech benchmarks demonstrate the strong performance of our approach when compared to the other state-of-the-art methods.

Accepted by AAAI2021 (poster). arXiv admin note: text overlap with arXiv:2007.12942

References in corpus (13)

• Overcoming catastrophic forgetting in neural networks
• A Simple Framework for Contrastive Learning of Visual Representations
• Deep Domain Confusion: Maximizing for Domain Invariance
• Improved Baselines with Momentum Contrastive Learning
• Continual Learning Through Synaptic Intelligence
• Efficient Lifelong Learning with A-GEM
• Contrastive Multiview Coding
• LSDA: Large Scale Detection Through Adaptation
• Domain Agnostic Learning with Disentangled Representations
• A Unified Feature Disentangler for Multi-Domain Image Translation and Manipulation
• Cross-domain Self-supervised Learning for Domain Adaptation with Few Source Labels
• Continuous Domain Adaptation with Variational Domain-Agnostic Feature Replay
• Joint Contrastive Learning for Unsupervised Domain Adaptation

Cited by in corpus (4)

• Delving into the Continuous Domain Adaptation
• Unsupervised domain adaptation via coarse-to-fine feature alignment method using contrastive learning
• Layerwise Optimization by Gradient Decomposition for Continual Learning
• Look, Cast and Mold: Learning 3D Shape Manifold from Single-view Synthetic Data