Research map / Label-efficient and robust learning / Domain adaptation and generalization
LEARNING FACTORIZED REPRESENTATIONS FOR OPEN-SET DOMAIN ADAPTATION
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 |
|---|---|---|---|
| Add a source-label classifier loss to FRODA (D-FRODA) (ablation) | FRODA · BCIS | recognition accuracy: FRODA→D-FRODA: B→C 73.8±6.1→74.6±5.5; B→I 71.0±2.0→71.4±2.0; B→S 54.7 ± 2.9→55.4±2.7; C→B 67.5±1.4→67.6±1.2; C→I 74.5±1. | Mixed |
| Add a source-label classifier loss to FRODA (D-FRODA) (ablation) | FRODA · Office | recognition accuracy: A→D 88.0→87.4; A→W 78.7→78.1; W→A 76.5→77.1; W→D 98.0→98.5; D→A 73.7→73.6; D→W 94.6→94.4 | Mixed |
| Use the learned linear classifier W instead of linear SVM (ablation) | D-FRODA · BCIS | recognition accuracy: B→C 74.6±5.5→61.2±1.2; B→I 71.4±2.0→59.3±1.1; C→I 75.0±1.8→66.2±1.7; I→C 80.5±1.6→70.5±2.5 | Hurt |
| Use 3-nearest neighbors instead of linear SVM (ablation) | FRODA · BCIS | recognition accuracy: B→C 73.8±6.1→67.7±2.8; B→I 71.0±2.0→64.1±2.4; I→B 66.0±1.9→60.9±3.7 | Hurt |
| Add auxiliary unknown-class source samples (ablation) | D-FRODA-U · BCIS | recognition accuracy: D-FRODA-SVM→D-FRODA-U-SVM: B→C 74.6±5.5→71.9±3.8; B→I 71.4±2.0→69.2±2.7; B→S 55.4±2.7→56.7±3.3; C→B 67.6±1.2→64.8±1.8; I | Mixed |
| Increase the unknown-detection threshold ε once it is sufficiently large (ablation) | FRODA · BCIS average across 12 pairs | recognition accuracy | No clear effect |
| Factor target data into shared known-class and private unknown-class subspaces with group sparsity | FRODA · BCIS | recognition accuracy: BCIS S→B: ATI 23.2±3.2; FRODA 55.7±2.5 | Helped |
| Use factorized shared/private representations rather than the ATI approach | FRODA · Office | runtime: roughly 32 seconds→roughly 2.5 seconds | Helped |
Closest papers by meaning
- Towards Inheritable Models for Open-Set Domain Adaptation (CVPR 2020 · similarity 0.91)
- Learning an Invariant Hilbert Space for Domain Adaptation (CVPR 2017 · similarity 0.91)
- Multivariate Regression on the Grassmannian for Predicting Novel Domains (CVPR 2016 · similarity 0.89)
- Progressive Graph Learning for Open-Set Domain Adaptation (ICML 2020 · similarity 0.89)
- Diversify and Match: A Domain Adaptive Representation Learning Paradigm for Object Detection (CVPR 2019 · similarity 0.89)
- Open-Set Domain Adaptation for Semantic Segmentation (CVPR 2024 · 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.