Research map / Label-efficient and robust learning / Domain adaptation and generalization
PCL: Proxy-Based Contrastive Learning for Domain Generalization
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace positive sample-to-sample alignment with proxy-to-sample positive relations and in-domain negative pairs | Softmax CE training for domain generalization · PACS, Office-Home, TerraIncognita, and DomainNet | average accuracy: 88.1 -> 88.7; 70.6 -> 71.6; 50.0 -> 52.1; 46.5 -> 47.7 | Helped |
| add positive sample-to-sample alignment to Softmax CE (ablation) | domain-generalization training · PACS, Office-Home, and TerraIncognita | accuracy: 88.1 -> 86.7; 70.6 -> 70.1; 50.0 -> 48.5 | Hurt |
| replace PCL with supervised contrastive learning (ablation) | Office-Home domain generalization · Office-Home | average accuracy: 68.92 -> 65.98 | Hurt |
| replace PCL with Proxy-Anchor (ablation) | Office-Home domain generalization · Office-Home | average accuracy: 68.92 -> 62.48 | Hurt |
| add PCL to Softmax CE (ablation) | AlexNet trained without ImageNet pretraining · CIFAR-100, batch-size 128 | top1 accuracy: 44.11 -> 48.16 | Helped |
| add PCL to Softmax CE (ablation) | AlexNet trained without ImageNet pretraining · CIFAR-100, batch-size 256 | top1 accuracy: 46.28 -> 49.12 | Helped |
| increase embedding size (ablation) | PCL · Office-Home | accuracy | Mixed |
| vary batch size (ablation) | PCL · Office-Home | accuracy | No clear effect |
| use no ImageNet-pretrained weights | PCL · Office-Home, PACS, and TerraIncognita | average accuracy: 23.69 -> 26.05; 47.84 -> 51.51; 24.25 -> 25.13; 20.50 -> 22.02; 21.60 -> 22.02 | Mixed |
Closest papers by meaning
- DomCLP: Domain-wise Contrastive Learning with Prototype Mixup for Unsupervised Domain Generalization (AAAI 2025 · similarity 0.89)
- Domain Generalization via Conditional Invariant Representations (AAAI 2018 · similarity 0.88)
- Connect, Not Collapse: Explaining Contrastive Learning for Unsupervised Domain Adaptation (ICML 2022 · similarity 0.88)
- Robust Domain Generalization under Divergent Marginal and Conditional Distributions (NeurIPS 2026 · similarity 0.87)
- Mixup-Induced Domain Extrapolation for Domain Generalization (AAAI 2024 · similarity 0.87)
- Generalization Analysis for Supervised Contrastive Representation Learning under Non-IID Settings (ICML 2025 · similarity 0.86)
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.