Research map / Label-efficient and robust learning / Domain adaptation and generalization
f-Domain Adversarial Learning: Theory and Algorithms
Read the paperCheck 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 DANN's constant source-classifier assumption with f-DAL's per-category domain classifier | DANN / f-DAL (JS) · Digits, Amazon Reviews, Office-31, Office-Home | accuracy (%): 93.3 -> 96.6; 76.3 -> 80.1; 82.2 -> 88.8; 57.6 -> 66.8 | Helped |
| Use Pearson χ² divergence instead of JS divergence (ablation) | f-DAL · Amazon Reviews | accuracy (%): 80.1 -> 81.6 | Helped |
| Use Pearson χ² divergence instead of JS divergence (ablation) | f-DAL · Office-Home | accuracy (%): 66.8 -> 68.3 | Helped |
| Use Pearson χ² divergence instead of JS divergence (ablation) | f-DAL · Digits | accuracy (%): 96.6 -> 96.3 | Hurt |
| Use Pearson χ² divergence instead of JS divergence (ablation) | f-DAL · Office-31 | accuracy (%): 88.8 -> 89.2 | Mixed |
| Add sampling-based label alignment (ablation) | f-DAL (Pearson χ²) · Office-Home | accuracy (%): 68.3 -> 70.0 | Helped |
| Add sampling-based label alignment (ablation) | f-DAL (Pearson χ²) · Office-31 | accuracy (%): 89.2 -> 89.2 | No clear effect |
| Use γ-weighted JS instead of JS (ablation) | f-DAL · Office-31 | accuracy (%): 88.8 -> 88.9 | No clear effect |
| Vary γ for shifted Pearson χ² (ablation) | f-DAL · Digits average | accuracy (%): 96.3; 96.2; 96.4; 96.3 | No clear effect |
| Add spectral normalization to the two classifiers (ablation) | f-DAL · Visual tasks | transfer performance | No clear effect |
| Remove the masking scheme when combining alignment with f-DAL-Pearson (ablation) | f-DAL (Pearson χ²) + Alignment · Office-31 and Office-Home | accuracy (%) | No clear effect |
Closest papers by meaning
- Adversarial Discriminative Domain Adaptation (CVPR 2017 · similarity 0.92)
- Improving the Generalization of Adversarial Training with Domain Adaptation (ICLR 2019 · similarity 0.91)
- Domain Adaptation with Adversarial Training on Penultimate Activations (AAAI 2023 · similarity 0.91)
- Bridging Theory and Algorithm for Domain Adaptation (ICML 2019 · similarity 0.91)
- On $f$-Divergence Principled Domain Adaptation: An Improved Framework (NeurIPS 2024 · similarity 0.90)
- Kernel Learning with Adversarial Features: Numerical Efficiency and Adaptive Regularization (NeurIPS 2025 · similarity 0.90)
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.