Asymmetric Tri-training for Unsupervised Domain Adaptation

Abstract

Deep-layered models trained on a large number of labeled samples boost the accuracy of many tasks. It is important to apply such models to different domains because collecting many labeled samples in various domains is expensive. In unsupervised domain adaptation, one needs to train a classifier that works well on a target domain when provided with labeled source samples and unlabeled target samples. Although many methods aim to match the distributions of source and target samples, simply matching the distribution cannot ensure accuracy on the target domain. To learn discriminative representations for the target domain, we assume that artificially labeling target samples can result in a good representation. Tri-training leverages three classifiers equally to give pseudo-labels to unlabeled samples, but the method does not assume labeling samples generated from a different domain.In this paper, we propose an asymmetric tri-training method for unsupervised domain adaptation, where we assign pseudo-labels to unlabeled samples and train neural networks as if they are true labels. In our work, we use three networks asymmetrically. By asymmetric, we mean that two networks are used to label unlabeled target samples and one network is trained by the samples to obtain target-discriminative representations. We evaluate our method on digit recognition and sentiment analysis datasets. Our proposed method achieves state-of-the-art performance on the benchmark digit recognition datasets of domain adaptation.

TBA on ICML2017

References in corpus (4)

• Learning Transferable Features with Deep Adaptation Networks
• Unsupervised Domain Adaptation by Backpropagation
• Domain Separation Networks
• Deep Reconstruction-Classification Networks for Unsupervised Domain Adaptation

Cited by in corpus (138)

• Deep Visual Domain Adaptation: A Survey
• Deep Subdomain Adaptation Network for Image Classification
• Model Adaptation: Unsupervised Domain Adaptation without Source Data
• A DIRT-T Approach to Unsupervised Domain Adaptation
• An Overview of Deep Semi-Supervised Learning
• A Comprehensive Survey on Transfer Learning
• Collaborative Unsupervised Domain Adaptation for Medical Image Diagnosis
• Unsupervised Domain Adaptation through Self-Supervision
• Universal Domain Adaptation through Self Supervision
• Heterogeneous domain adaptation: An unsupervised approach
• Maximum Classifier Discrepancy for Unsupervised Domain Adaptation
• Multi-Source Unsupervised Domain Adaptation via Pseudo Target Domain
• Improve Unsupervised Domain Adaptation with Mixup Training
• Learning to cluster in order to transfer across domains and tasks
• Self-Training: A Survey
• Theoretical Analysis of Self-Training with Deep Networks on Unlabeled Data
• Pseudo-Labeling Curriculum for Unsupervised Domain Adaptation
• Domain Adaptation on Point Clouds via Geometry-Aware Implicits
• Learning Transferable Parameters for Unsupervised Domain Adaptation
• FDA: Fourier Domain Adaptation for Semantic Segmentation
• TimeMatch: Unsupervised Cross-Region Adaptation by Temporal Shift Estimation
• Domain Adaptation for Object Detection via Style Consistency
• Self-ensembling for visual domain adaptation
• Deep Cocktail Network: Multi-source Unsupervised Domain Adaptation with Category Shift
• Dynamic Instance Domain Adaptation
• Transferrable Prototypical Networks for Unsupervised Domain Adaptation
• Virtual Mixup Training for Unsupervised Domain Adaptation
• Larger Norm More Transferable: An Adaptive Feature Norm Approach for Unsupervised Domain Adaptation
• Adversarial Domain Adaptation with Prototype-Based Normalized Output Conditioner
• Relation Matters: Foreground-aware Graph-based Relational Reasoning for Domain Adaptive Object Detection
• Augmented Cyclic Adversarial Learning for Low Resource Domain Adaptation
• Domain-Symmetric Networks for Adversarial Domain Adaptation
• GALA: Graph Diffusion-based Alignment with Jigsaw for Source-free Domain Adaptation
• Inferring Latent Domains for Unsupervised Deep Domain Adaptation
• Confidence Regularized Self-Training
• Racial Faces in-the-Wild: Reducing Racial Bias by Information Maximization Adaptation Network
• Self-Training and Adversarial Background Regularization for Unsupervised Domain Adaptive One-Stage Object Detection
• Exploring Object Relation in Mean Teacher for Cross-Domain Detection
• Domain-Specific Batch Normalization for Unsupervised Domain Adaptation
• Knowledge Distillation Methods for Efficient Unsupervised Adaptation Across Multiple Domains
• Source Data-absent Unsupervised Domain Adaptation through Hypothesis Transfer and Labeling Transfer
• Progressive Feature Alignment for Unsupervised Domain Adaptation
• Open Set Domain Adaptation by Backpropagation
• SALUDA: Surface-based Automotive Lidar Unsupervised Domain Adaptation
• Joint Contrastive Learning for Unsupervised Domain Adaptation
• Cluster Alignment with a Teacher for Unsupervised Domain Adaptation
• MALA: Cross-Domain Dialogue Generation with Action Learning
• Towards Shape Biased Unsupervised Representation Learning for Domain Generalization
• Cross-domain Object Detection through Coarse-to-Fine Feature Adaptation
• DiDA: Disentangled Synthesis for Domain Adaptation
• Vicinal and categorical domain adaptation
• DUSA: Decoupled Unsupervised Sim2Real Adaptation for Vehicle-to-Everything Collaborative Perception
• Meta Discovery: Learning to Discover Novel Classes given Very Limited Data
• Source Free Unsupervised Graph Domain Adaptation
• Dual Mixup Regularized Learning for Adversarial Domain Adaptation
• Improving Domain Adaptation Through Class Aware Frequency Transformation
• Entropy Minimization vs. Diversity Maximization for Domain Adaptation
• Rethinking Importance Weighting for Deep Learning under Distribution Shift
• Cross-View Regularization for Domain Adaptive Panoptic Segmentation
• Attention Regularized Laplace Graph for Domain Adaptation
• Discriminative and Geometry Aware Unsupervised Domain Adaptation
• Domain Alignment with Triplets
• Improving Unsupervised Domain Adaptation with Variational Information Bottleneck
• Foreground-Aware Stylization and Consensus Pseudo-Labeling for Domain Adaptation of First-Person Hand Segmentation
• Learning from a Complementary-label Source Domain: Theory and Algorithms
• ST3D: Self-training for Unsupervised Domain Adaptation on 3D Object Detection
• Known-class Aware Self-ensemble for Open Set Domain Adaptation
• Adversarial Cross-Domain Action Recognition with Co-Attention
• Effective Label Propagation for Discriminative Semi-Supervised Domain Adaptation
• Contrastively Smoothed Class Alignment for Unsupervised Domain Adaptation
• Disjoint Label Space Transfer Learning with Common Factorised Space
• Class-imbalanced Domain Adaptation: An Empirical Odyssey
• Butterfly: One-step Approach towards Wildly Unsupervised Domain Adaptation
• Unsupervised Domain Adaptation Based on Source-guided Discrepancy
• A Sample Selection Approach for Universal Domain Adaptation
• HoMM: Higher-order Moment Matching for Unsupervised Domain Adaptation
• An Adversarial Perturbation Oriented Domain Adaptation Approach for Semantic Segmentation
• Unsupervised Visual Domain Adaptation: A Deep Max-Margin Gaussian Process Approach
• MetaMixUp: Learning Adaptive Interpolation Policy of MixUp with Meta-Learning
• Learning Smooth Representation for Unsupervised Domain Adaptation
• Learning a Domain-Invariant Embedding for Unsupervised Domain Adaptation Using Class-Conditioned Distribution Alignment
• Clarinet: A One-step Approach Towards Budget-friendly Unsupervised Domain Adaptation
• Feature Adaptation of Pre-Trained Language Models across Languages and Domains with Robust Self-Training
• Zero-Shot Deep Domain Adaptation
• Online Knowledge Distillation via Multi-branch Diversity Enhancement
• Identity Preserving Generative Adversarial Network for Cross-Domain Person Re-identification
• Joint Semantic Domain Alignment and Target Classifier Learning for Unsupervised Domain Adaptation
• Gradual Domain Adaptation in the Wild:When Intermediate Distributions are Absent
• Prototypical Cross-domain Knowledge Transfer for Cervical Dysplasia Visual Inspection
• Asymmetric Tri-training for Debiasing Missing-Not-At-Random Explicit Feedback
• Preserving Semantic Consistency in Unsupervised Domain Adaptation Using Generative Adversarial Networks
• Contradictory Structure Learning for Semi-supervised Domain Adaptation
• Fine-grained Domain Adaptive Crowd Counting via Point-derived Segmentation
• GradMix: Multi-source Transfer across Domains and Tasks
• Domain Adaptation for Sentiment Analysis Using Increased Intraclass Separation
• CUDA: Contradistinguisher for Unsupervised Domain Adaptation
• Data Priming Network for Automatic Check-Out
• Self-adaptive Re-weighted Adversarial Domain Adaptation
• A General Upper Bound for Unsupervised Domain Adaptation
• Semi-supervised Learning for Few-shot Image-to-Image Translation
• Weakly Labeled Sound Event Detection Using Tri-training and Adversarial Learning
• AdaFlow: Domain-Adaptive Density Estimator with Application to Anomaly Detection and Unpaired Cross-Domain Translation
• Conditional Coupled Generative Adversarial Networks for Zero-Shot Domain Adaptation
• Transporting Causal Mechanisms for Unsupervised Domain Adaptation
• Annotation Cost Reduction of Stream-based Active Learning by Automated Weak Labeling using a Robot Arm
• Domain Impression: A Source Data Free Domain Adaptation Method
• Unsupervised Domain Adaptation on Reading Comprehension
• Sequential Learning for Domain Generalization
• Learning to Contextually Aggregate Multi-Source Supervision for Sequence Labeling
• Towards Accurate and Robust Domain Adaptation under Noisy Environments
• Disentanglement Then Reconstruction: Learning Compact Features for Unsupervised Domain Adaptation
• Progressive Ensemble Networks for Zero-Shot Recognition
• A Fully Convolutional Tri-branch Network (FCTN) for Domain Adaptation
• Attract or Distract: Exploit the Margin of Open Set
• A Dictionary Approach to Domain-Invariant Learning in Deep Networks
• Learning Classifiers for Domain Adaptation, Zero and Few-Shot Recognition Based on Learning Latent Semantic Parts
• Divergence Optimization for Noisy Universal Domain Adaptation
• Towards Recognizing New Semantic Concepts in New Visual Domains
• Efficient Pre-trained Features and Recurrent Pseudo-Labeling in Unsupervised Domain Adaptation
• Resource and data efficient self supervised learning
• Signature-Graph Networks
• Enlarging Discriminative Power by Adding an Extra Class in Unsupervised Domain Adaptation
• Pseudo-labels for Supervised Learning on Dynamic Vision Sensor Data, Applied to Object Detection under Ego-motion
• Adversarial Learning for Zero-shot Domain Adaptation
• Contrastive Vicinal Space for Unsupervised Domain Adaptation
• Deep Spherical Manifold Gaussian Kernel for Unsupervised Domain Adaptation
• ADeLA: Automatic Dense Labeling with Attention for Viewpoint Adaptation in Semantic Segmentation
• Unsupervised Domain Adaptive Object Detection using Forward-Backward Cyclic Adaptation
• Unsupervised Adaptive Semantic Segmentation with Local Lipschitz Constraint
• Cross Domain Image Matching in Presence of Outliers
• Unsupervised Domain Adaptation: A Reality Check
• Semi-Supervised Text Classification via Self-Pretraining
• Adaptive Pseudo-Label Refinement by Negative Ensemble Learning for Source-Free Unsupervised Domain Adaptation
• iFAN: Image-Instance Full Alignment Networks for Adaptive Object Detection
• Hard Class Rectification for Domain Adaptation
• Learning to see across Domains and Modalities
• Teacher-Student Competition for Unsupervised Domain Adaptation
• Deep Domain Adaptation under Deep Label Scarcity