Research map / Label-efficient and robust learning / Domain adaptation and generalization
Multi-class Heterogeneous Domain Adaptation
Joey Tianyi Zhou, Ivor W. Tsang, Sinno Jialin Pan, Mingkui Tan
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 nonnegative constraint on G (SHFR-NNLS vs SHFR-LASSO) (ablation) | SHFR · Multilingual Reuters (English source) | classification accuracy (%): 67.08 -> 71.47 | Improved |
| add l1-norm regularization to NNLS (SHFR vs SHFR-NNLS) (ablation) | SHFR · Multilingual Reuters (English source) | classification accuracy (%): 71.47 -> 72.82 | Improved |
| add l1-norm regularization to NNLS (SHFR vs SHFR-NNLS) (ablation) | SHFR · Cross-lingual Sentiment (English source, Japanese target) | classification accuracy (%): 61.51 -> 62.58 | Improved |
| use ECOC coding instead of one-vs-one for SHFR | SHFR · Multilingual Reuters (English source, Spanish target) | classification accuracy (%): 69.45 -> 72.82 | Improved |
| use ECOC coding instead of one-vs-all to build dictionary | SHFR dictionary D · Multilingual Reuters (English source) | coherence sum |rho| (lower is better): 0.8532 -> 0.1349 | Improved |
| learn global transformation G (SHFR) instead of SVM-T alone on binary task 5 | SHFR · Multilingual Reuters binary task 5 | binary accuracy (%): 43.32 -> 80.76 | Improved |
| learn global transformation G (SHFR) instead of SVM-T alone on binary task 1 | SHFR · Multilingual Reuters binary task 1 | binary accuracy (%): 74.42 -> 65.39 | Hurt |
| learn global G, average binary accuracy gain over SVM-T on BBC and Sentiment | SHFR · BBC and Sentiment | binary accuracy (%): +26.53% on BBC, +3.01% on Sentiment | Improved |
| remove negative alignments (SHFR-RN) in binary tasks (ablation) | SHFR · Multilingual Reuters binary task 1 | binary accuracy (%): 65.39 -> 85.39 | Improved |
| remove negative alignments when many dichotomizers are used (ablation) | SHFR · Multilingual Reuters multi-class | classification accuracy (%) | No clear effect |
| increase number of dichotomizers beyond 31 | SHFR (ECOC) · Multilingual Reuters | classification accuracy (%) | No clear effect |
| use few labeled target instances (fewer than 10) with SHFR vs baselines | SHFR · Multilingual Reuters (English->Spanish) | classification accuracy (%) | Improved |
Closest papers by meaning
- Open-Set Heterogeneous Domain Adaptation: Theoretical Analysis and Algorithm (AAAI 2025 · similarity 0.87)
- Semi-Supervised Heterogeneous Domain Adaptation: Theory and Algorithms (TPAMI 2022 · similarity 0.87)
- Source-Free Adaptation to Measurement Shift via Bottom-Up Feature Restoration (ICLR 2022 · similarity 0.86)
- A Unified Framework for Heterogeneous Semi-supervised Learning (CVPR 2025 · similarity 0.86)
- Learning Cross-Domain Landmarks for Heterogeneous Domain Adaptation (CVPR 2016 · similarity 0.86)
- Instance Level Affinity-Based Transfer for Unsupervised Domain Adaptation (CVPR 2021 · similarity 0.85)
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.
