Research map / Label-efficient and robust learning / Domain adaptation and generalization
Heterogeneous Domain Adaptation With Adversarial Neural Representation Learning: Experiments on E-Commerce and Cybersecurity
Mohammadreza Reza Ebrahimi, Yidong Chai, Hao Helen Zhang, Hsinchun Chen
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 shared dictionary learning (ablation) | HANDA · Office31-Caltech256 SURF to DeCAF (average) | accuracy: 23.6 -> 97.2 | Improved |
| add adversarial kernel matching (ablation) | HANDA · Office31-Caltech256 SURF to DeCAF (average) | accuracy: 93.1 -> 97.2 | Improved |
| use adversarial kernel matching loss instead of entropy-based alignment (ablation) | HANDA · Office31-Caltech256 SURF to DeCAF (average) | accuracy: 96.3 -> 97.2 | Improved |
| add shared dictionary learning (ablation) | HANDA · Reuters multilingual EN→SP (10 labeled/class) | accuracy: 23.4 -> 74.5 | Improved |
| add adversarial kernel matching (ablation) | HANDA · Reuters multilingual EN→SP (10 labeled/class) | accuracy: 68.0 -> 74.5 | Improved |
| add adversarial kernel matching (ablation) | HANDA · Reuters multilingual EN→SP (20 labeled/class) | accuracy: 75.7 -> 79.8 | Improved |
| use adversarial kernel matching loss instead of entropy-based alignment (ablation) | HANDA · Reuters multilingual EN→SP (10 labeled/class) | accuracy: 73.7 -> 74.5 | Improved |
| train shared dictionary and adversarial matching jointly rather than sequentially (ablation) | HANDA · Reuters multilingual EN→SP (10 labeled/class) | accuracy: 71.1 -> 74.5 | Improved |
| use 2 hidden layers instead of 1 (ablation) | HANDA feature extractor · Reuters multilingual EN→SP (10 labeled/class) | accuracy: 73.0 -> 74.5 | Improved |
| use 2 hidden layers instead of 5 (ablation) | HANDA feature extractor · Reuters multilingual EN→SP (10 labeled/class) | accuracy: 72.3 -> 74.5 | Improved |
| use 3 hidden layers instead of 2 (ablation) | HANDA feature extractor · Reuters multilingual EN→SP (10 and 20 labels) | accuracy: 74.5 -> 74.4; 79.8 -> 80.0 | No clear effect |
| adapt English dark web markets to Russian with HANDA vs no-DA neural network | HANDA · Dark web cyber threat detection (Russian) | AUC: 0.96 -> 0.98 | Improved |
Closest papers by meaning
- Domain-Adversarial Training of Neural Networks (JMLR 2016 · similarity 0.87)
- Partial Adversarial Domain Adaptation (ECCV 2018 · similarity 0.86)
- Divergence-Agnostic Unsupervised Domain Adaptation by Adversarial Attacks (TPAMI 2021 · similarity 0.86)
- Multi-Adversarial Domain Adaptation (AAAI 2018 · similarity 0.86)
- A DIRT-T Approach to Unsupervised Domain Adaptation (ICLR 2018 · similarity 0.86)
- Open-Set Heterogeneous Domain Adaptation: Theoretical Analysis and Algorithm (AAAI 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.
