Research map / Label-efficient and robust learning / Domain adaptation and generalization
Contradistinguisher: A Vapnik's Imperative to Unsupervised Domain Adaptation
Sourabh Balgi, Ambedkar Dukkipati
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add target unsupervised contradistinguish loss to source supervised training (ablation) | CUDA · Office-31 single-source, fixed ResNet-50 | mean accuracy: 82.1 -> 88.7 | Improved |
| add target unsupervised loss (ablation) | CUDA · Office-31 single-source, fixed ResNet-152 | mean accuracy: 84.6 -> 90.0 | Improved |
| add target unsupervised loss (ablation) | CUDA · Office-31 multi-source | mean accuracy: 87.6 -> 90.4 | Improved |
| add source unsupervised loss (ablation) | CUDA · Office-31 single-source, fixed ResNet-50 | mean accuracy: 88.7 -> 88.5 | No clear effect |
| add target adversarial regularization (ablation) | CUDA · Office-31 multi-source | mean accuracy: 89.5 -> 90.8 | Improved |
| add target adversarial regularization (ablation) | CUDA · Office-31 single-source, fixed ResNet-152 | mean accuracy: 89.7 -> 90.7 | Improved |
| add target adversarial regularization (ablation) | CUDA · Office-31 single-source, fixed ResNet-50 | mean accuracy: 88.5 -> 88.0 | Hurt |
| add source adversarial regularization (ablation) | CUDA · Office-31 single-source, fixed ResNet-50 | mean accuracy: 88.0 -> 88.9 | Improved |
| fix pretrained ResNet-50 instead of fine-tuning for source-only training (ablation) | source supervised baseline · Office-31 single-source | mean accuracy: 50.6 -> 82.1 | Improved |
| use ResNet-152 instead of ResNet-50 encoder | CUDA · Office-31 single-source | mean accuracy: 89.8 -> 91.1 | Improved |
| add data augmentation and target clustering like CAN (CUDA*) (ablation) | CUDA · VisDA-2017 test | accuracy: 78.1 -> 85.8 | Improved |
| add data augmentation and target clustering like CAN (CUDA*) (ablation) | CUDA · VisDA-2017 validation | accuracy: 79.6 -> 84.2 | Improved |
Closest papers by meaning
- On Balancing Bias and Variance in Unsupervised Multi-Source-Free Domain Adaptation (ICML 2023 · similarity 0.90)
- U$^3$CF: Unbiased, Unconfounding, and Unified Causal Framework for Multi-Target Domain Adaptation (ICML 2026 · similarity 0.89)
- Co-regularized Alignment for Unsupervised Domain Adaptation (NeurIPS 2018 · similarity 0.89)
- Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning (NeurIPS 2022 · similarity 0.89)
- OSAN: A One-Stage Alignment Network To Unify Multimodal Alignment and Unsupervised Domain Adaptation (CVPR 2023 · similarity 0.89)
- SimProF: A Simple Probabilistic Framework for Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
