TAROT: Towards Essentially Domain-Invariant Robustness with Theoretical Justification

2025/05/10 by Dong‐Yoon Yang, Yang, Dongyoon, Jihu Lee +3
Computer Science · #Adversarial Robustness in Machine Learning #Artificial Intelligence (cs.AI) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (stat.ML) #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2505.06580

openalex publication_date 2025/05/10 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Robust domain adaptation against adversarial attacks is a critical research area that aims to develop models capable of maintaining consistent performance across diverse and challenging domains. In this paper, we derive a new generalization bound for robust risk on the target domain using a novel divergence measure specifically designed for robust domain adaptation. Building upon this, we propose a new algorithm named TAROT, which is designed to enhance both domain adaptability and robustness. Through extensive experiments, TAROT not only surpasses state-of-the-art methods in accuracy and robustness but also significantly enhances domain generalization and scalability by effectively learning domain-invariant features. In particular, TAROT achieves superior performance on the challenging DomainNet dataset, demonstrating its ability to learn domain-invariant representations that generalize well across different domains, including unseen ones. These results highlight the broader applicability of our approach in real-world domain adaptation scenarios.

Citations

• Reliable and Efficient Concept Erasure of Text-to-Image Diffusion Models
• Robust CLIP: Unsupervised Adversarial Fine-Tuning of Vision Embeddings for Robust Large Vision-Language Models
• Enhancing Adversarial Robustness in Low-Label Regime via Adaptively Weighted Regularization and Knowledge Distillation
• Understanding Zero-Shot Adversarial Robustness for Large-Scale Models
• SRoUDA: Meta Self-training for Robust Unsupervised Domain Adaptation
• Efficiently Computing Local Lipschitz Constants of Neural Networks via Bound Propagation
• Rethinking Lipschitz Neural Networks and Certified Robustness: A Boolean Function Perspective
• Improving Adversarial Robustness by Putting More Regularizations on Less Robust Samples
• The robust way to stack and bag: the local Lipschitz way
• Exploring Adversarially Robust Training for Unsupervised Domain Adaptation
• Two Wrongs Don't Make a Right: Combating Confirmation Bias in Learning with Label Noise
• Adversarial Robustness for Unsupervised Domain Adaptation
• Latent Space Regularization for Unsupervised Domain Adaptation in Semantic Segmentation
• Do Wider Neural Networks Really Help Adversarial Robustness?
• Do Adversarially Robust ImageNet Models Transfer Better?
• Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks
• Exactly Computing the Local Lipschitz Constant of ReLU Networks
• Pseudo-Labeling and Confirmation Bias in Deep Semi-Supervised Learning
• Are Labels Required for Improving Adversarial Robustness?
• Bridging Theory and Algorithm for Domain Adaptation
• Support and Invertibility in Domain-Invariant Representations
• Theoretically Principled Trade-off between Robustness and Accuracy
• Moment Matching for Multi-Source Domain Adaptation
• Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examples
• Spectrally-normalized margin bounds for neural networks
• Deep Hashing Network for Unsupervised Domain Adaptation
• Towards Deep Learning Models Resistant to Adversarial Attacks
• Parseval Networks: Improving Robustness to Adversarial Examples
• MultiNet: Real-time Joint Semantic Reasoning for Autonomous Driving
• Unsupervised Domain Adaptation by Backpropagation
• Intriguing properties of neural networks

Related

• Unsupervised Robust Domain Adaptation: Paradigm, Theory and Algorithm
• Towards Trustworthy Unsupervised Domain Adaptation: A Representation Learning Perspective for Enhancing Robustness, Discrimination, and Generalization
• RDA: Robust Domain Adaptation via Fourier Adversarial Attacking
• DART: A Principled Approach to Adversarially Robust Unsupervised Domain Adaptation
• Robustified Domain Adaptation