Embracing Domain Gradient Conflicts: Domain Generalization Using Domain Gradient Equilibrium
Abstract: Single domain generalization (SDG) aims to learn a generalizable model from only one source domain available to unseen target domains. Existing SDG techniques rely on data or feature augmentation to generate distributions that complement the source domain. However, these approaches fail to address the challenge where gradient conflicts from synthesized domains impede the learning of domain-invariant representation. Inspired by the concept of mechanical equilibrium in physics, we propose a novel conflict-aware approach named domain gradient equilibrium for SDG. Unlike prior conflict-aware SDG methods that alleviate the gradient conflicts by setting them to zero or random values, the proposed domain gradient equilibrium method first decouples gradients into domain-invariant and domain-specific components. The domain-specific gradients are then adjusted and reweighted to achieve equilibrium, steering the model optimization toward a domain-invariant direction to enhance generalization capability. We conduct comprehensive experiments on four image recognition benchmarks, and our method achieves an accuracy improvement of 2.94% in the PACS dataset over existing state-of-the-art approaches, demonstrating the effectiveness of our proposed approach.
Primary Subject Area: [Content] Media Interpretation
Relevance To Conference: Our approach contributes to multimedia/multimodal processing by addressing the challenge of domain generalization in single-domain settings, which is crucial for developing robust and scalable multimedia systems. The proposed domain gradient equilibrium method enhances the learning of domain-invariant representations by decoupling gradients into domain-invariant and domain-specific components and adjusting domain-specific gradients to achieve equilibrium. This process steers the model optimization towards a domain-invariant direction, improving the generalization capability of multimedia models. Comprehensive experiments on four image recognition benchmarks demonstrate the effectiveness of the proposed approach, with a significant accuracy improvement over state-of-the-art methods. This performance gain highlights the potential of the domain gradient equilibrium method in enhancing the robustness and adaptability of multimedia systems when dealing with data from various domains.
Submission Number: 2715
Loading