Abstract:Unsupervised Domain Adaptation (UDA) aims to learn a classifier for the unlabeled target domain by leveraging knowledge from a labeled source domain with a different but related distribution. Many existing approaches typically learn a domain-invariant representation space by directly matching the marginal distributions of the two domains. However, they ignore exploring the underlying discriminative features of the target data and align the cross-domain discriminative features, which may lead to suboptimal performance. To tackle these two issues simultaneously, this paper presents a Joint Clustering and Discriminative Feature Alignment (JCDFA) approach for UDA, which is capable of naturally unifying the mining of discriminative features and the alignment of class-discriminative features into one single framework. Specifically, in order to mine the intrinsic discriminative information of the unlabeled target data, JCDFA jointly learns a shared encoding representation for two tasks: supervised classification of labeled source data, and discriminative clustering of unlabeled target data, where the classification of the source domain can guide the clustering learning of the target domain to locate the object category. We then conduct the cross-domain discriminative feature alignment by separately optimizing two new metrics: 1) an extended supervised contrastive learning, i.e., semi-supervised contrastive learning 2) an extended Maximum Mean Discrepancy (MMD), i.e., conditional MMD, explicitly minimizing the intra-class dispersion and maximizing the inter-class compactness. When these two procedures, i.e., discriminative features mining and alignment are integrated into one framework, they tend to benefit from each other to enhance the final performance from a cooperative learning perspective. Experiments are conducted on four real-world benchmarks (e.g., Office-31, ImageCLEF-DA, Office-Home and VisDA-C)- All the results demonstrate that our JCDFA can obtain remarkable margins over state-of-the-art domain adaptation methods. Comprehensive ablation studies also verify the importance of each key component of our proposed algorithm and the effectiveness of combining two learning strategies into a framework.

SPA: A Graph Spectral Alignment Perspective for Domain Adaptation

Attention-based Cross-Layer Domain Alignment for Unsupervised Domain Adaptation

Deep Joint Two-Stream Wasserstein Auto-Encoder and Selective Attention Alignment for Unsupervised Domain Adaptation

Revisiting Deep Subspace Alignment for Unsupervised Domain Adaptation

Learning Distinctive Margin Toward Active Domain Adaptation

Domain-Agnostic Prior for Transfer Semantic Segmentation.

Domain-Aware Graph Network for Bridging Multi-Source Domain Adaptation

Spectral Unsupervised Domain Adaptation for Visual Recognition

Visual Domain Adaptation Using Weighted Subspace Alignment

Self-Guided Adaptation: Progressive Representation Alignment for Domain Adaptive Object Detection

Discriminative and Geometry-Aware Unsupervised Domain Adaptation

GrabDAE: An Innovative Framework for Unsupervised Domain Adaptation Utilizing Grab-Mask and Denoise Auto-Encoder

Deep joint subdomain alignment for unsupervised domain adaptation

Enhanced Unsupervised Domain Adaptation with Dual-Attention Between Classification and Domain Alignment.

Semantic Concentration for Domain Adaptation

Graph-Relational Domain Adaptation

When Unsupervised Domain Adaptation Meets Tensor Representations.

GLA-DA: Global-Local Alignment Domain Adaptation for Multivariate Time Series

Towards Uncovering the Intrinsic Data Structures for Unsupervised Domain Adaptation using Structurally Regularized Deep Clustering

Joint Clustering and Discriminative Feature Alignment for Unsupervised Domain Adaptation

Multi-source Distilling Domain Adaptation