SPA++: Generalized Graph Spectral Alignment for Versatile Domain Adaptation

2025/08/07 by Zhiqing Xiao, Xiao, Zhiqing, Haobo Wang +9 · 1 citation
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2508.05182

openalex publication_date 2025/08/07 · openalex created_date 2025/10/15 · openalex updated_date 2026/07/28

Abstract

Domain Adaptation (DA) aims to transfer knowledge from a labeled source domain to an unlabeled or sparsely labeled target domain under domain shifts. Most prior works focus on capturing the inter-domain transferability but largely overlook rich intra-domain structures, which empirically results in even worse discriminability. To tackle this tradeoff, we propose a generalized graph SPectral Alignment framework, SPA++. Its core is briefly condensed as follows: (1)-by casting the DA problem to graph primitives, it composes a coarse graph alignment mechanism with a novel spectral regularizer toward aligning the domain graphs in eigenspaces; (2)-we further develop a fine-grained neighbor-aware propagation mechanism for enhanced discriminability in the target domain; (3)-by incorporating data augmentation and consistency regularization, SPA++ can adapt to complex scenarios including most DA settings and even challenging distribution scenarios. Furthermore, we also provide theoretical analysis to support our method, including the generalization bound of graph-based DA and the role of spectral alignment and smoothing consistency. Extensive experiments on benchmark datasets demonstrate that SPA++ consistently outperforms existing cutting-edge methods, achieving superior robustness and adaptability across various challenging adaptation scenarios.

Citations

• SA-GDA: Spectral Augmentation for Graph Domain Adaptation
• SPA: A Graph Spectral Alignment Perspective for Domain Adaptation
• A Message Passing Perspective on Learning Dynamics of Contrastive Learning
• Refined Pseudo labeling for Source-free Domain Adaptive Object Detection
• Guiding Pseudo-labels with Uncertainty Estimation for Source-free Unsupervised Domain Adaptation
• Towards a Unified Theoretical Understanding of Non-contrastive Learning via Rank Differential Mechanism
• RLSbench: Domain Adaptation Under Relaxed Label Shift
• CoNMix for Source-free Single and Multi-target Domain Adaptation
• Subsidiary Prototype Alignment for Universal Domain Adaptation
• Multi-Domain Long-Tailed Learning by Augmenting Disentangled Representations
• Semi-Supervised Domain Adaptation with Auto-Encoder via Simultaneous Learning
• MemSAC: Memory Augmented Sample Consistency for Large Scale Unsupervised Domain Adaptation
• A Closer Look at Smoothness in Domain Adversarial Training
• Balancing Discriminability and Transferability for Source-Free Domain Adaptation
• Reusing the Task-specific Classifier as a Discriminator: Discriminator-free Adversarial Domain Adaptation
• Beyond Separability: Analyzing the Linear Transferability of Contrastive Representations to Related Subpopulations
• On Multi-Domain Long-Tailed Recognition, Imbalanced Domain Generalization and Beyond
• SIGMA: Semantic-complete Graph Matching for Domain Adaptive Object Detection
• Image Classification using Graph Neural Network and Multiscale Wavelet Superpixels
• Image classification using graph neural network and multiscale wavelet superpixels
• Improving Out-of-Distribution Robustness via Selective Augmentation
• Parametric Contrastive Learning
• Graph Domain Adaptation: A Generative View
• Multi-Target Domain Adaptation with Collaborative Consistency Learning
• Cross-Domain Adaptive Clustering for Semi-Supervised Domain Adaptation
• Domain Generalization with MixStyle
• Bi-Classifier Determinacy Maximization for Unsupervised Domain Adaptation
• In-N-Out: Pre-Training and Self-Training using Auxiliary Information for Out-of-Distribution Robustness
• FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation
• Improving Calibration for Long-Tailed Recognition
• Transfer Learning of Graph Neural Networks with Ego-graph Information Maximization
• Attract, Perturb, and Explore: Learning a Feature Alignment Network for Semi-supervised Domain Adaptation
• Learning to Combine: Knowledge Aggregation for Multi-Source Domain Adaptation
• Unsupervised Multi-Target Domain Adaptation Through Knowledge Distillation
• Domain Adaptation with Auxiliary Target Domain-Oriented Classifier
• Remix: Rebalanced Mixup
• Curriculum Manager for Source Selection in Multi-Source Domain Adaptation
• A survey on domain adaptation theory: learning bounds and theoretical guarantees
• Gradually Vanishing Bridge for Adversarial Domain Adaptation
• Meta Pseudo Labels
• RIDE: Rewarding Impact-Driven Exploration for Procedurally-Generated Environments
• Universal Domain Adaptation through Self Supervision
• Unifying Graph Convolutional Neural Networks and Label Propagation
• Drop to Adapt: Learning Discriminative Features for Unsupervised Domain Adaptation
• Pseudo-Labeling and Confirmation Bias in Deep Semi-Supervised Learning
• Invariant Risk Minimization
• Learning Imbalanced Datasets with Label-Distribution-Aware Margin Loss
• MixMatch: A Holistic Approach to Semi-Supervised Learning
• Semi-supervised Domain Adaptation via Minimax Entropy
• Label Propagation for Deep Semi-supervised Learning
• Relation-Aware Graph Attention Network for Visual Question Answering
• Moment Matching for Multi-Source Domain Adaptation
• Learning to Propagate Labels: Transductive Propagation Network for Few-shot Learning
• Deep Cocktail Network: Multi-source Unsupervised Domain Adaptation with Category Shift
• Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
• Graph Attention Networks
• Deep Hashing Network for Unsupervised Domain Adaptation
• On Calibration of Modern Neural Networks
• Adversarial Discriminative Domain Adaptation
• Temporal Ensembling for Semi-Supervised Learning
• Semi-Supervised Classification with Graph Convolutional Networks
• Learning the Roots of Visual Domain Shift
• Deep CORAL: Correlation Alignment for Deep Domain Adaptation
• Identity Mappings in Deep Residual Networks
• Deep Residual Learning for Image Recognition
• Rethinking the Inception Architecture for Computer Vision
• Learning Transferable Features with Deep Adaptation Networks
• Deep Visual-Semantic Alignments for Generating Image Descriptions
• Microsoft COCO: Common Objects in Context
• Spectral Networks and Locally Connected Networks on Graphs
• A tutorial on spectral clustering

Cited by

• Cross-Resolution Semantic Learning for Graph Domain Adaptation

Related

• SPA: A Graph Spectral Alignment Perspective for Domain Adaptation
• SA-GDA: Spectral Augmentation for Graph Domain Adaptation
• Domain Adaptation on Graphs by Learning Aligned Graph Bases
• Just DIAL: DomaIn Alignment Layers for Unsupervised Domain Adaptation
• Homophily Enhanced Graph Domain Adaptation