Research map / Label-efficient and robust learning / Recognition and long-tail
Effective End-to-end Unsupervised Outlier Detection via Inlier Priority of Discriminative Network
Siqi Wang, Yijie Zeng, Xinwang Liu, En Zhu, Jianping Yin, Chuanfu Xu, Marius Kloft
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 |
|---|---|---|---|
| use end-to-end negative-entropy score instead of feeding SSD representations to isolation forest (ablation) | E3Outlier · CIFAR10 10% | AUROC: 64.0 -> 83.5 | Improved |
| use end-to-end score instead of decoupled isolation forest (ablation) | E3Outlier · MNIST 10% | AUROC: 93.8 -> 94.1 | No clear effect |
| replace CAE with surrogate-supervision discriminative network (both with isolation forest) (ablation) | SSD-IF · CIFAR10 10% | AUROC: 54.1 -> 64.0 | Improved |
| replace CAE-based UOD with E3Outlier | E3Outlier · CIFAR10 10% | AUROC: 55.9 -> 83.5 | Improved |
| train CAE on the same K-times augmented data (ablation) | CAE · MNIST 10% | AUROC: 68.0 -> 55.5 | Hurt |
| tune CAE latent dimension to 64 (ablation) | CAE · CIFAR10/CIFAR100 10% | AUROC: 56.3/56.1 | No clear effect |
| add irregular affine and patch re-arranging operations to regular affine set (ablation) | E3Outlier · five benchmarks 10% | AUROC | Improved |
| use negative entropy score instead of PL/MP (ablation) | E3Outlier · five benchmarks 10% | AUROC: up to 2.3% AUC | Improved |
| use maximum-probability score instead of pseudo-label score (ablation) | E3Outlier · five benchmarks 10% | AUROC | Improved |
| use ResNet-50/DenseNet-40 instead of WRN-10-4 (ablation) | E3Outlier · CIFAR10/SVHN/CIFAR100 | AUROC | Mixed |
| train longer than 3 epochs (ablation) | E3Outlier · five benchmarks 10% | AUROC | No clear effect |
| increase outlier ratio (ablation) | E3Outlier · five benchmarks | AUROC | Hurt |
Closest papers by meaning
- ODIM: Outlier Detection via Likelihood of Under-Fitted Generative Models (ICML 2024 · similarity 0.89)
- ALTBI: Constructing Improved Outlier Detection Models via Optimization of Inlier-Memorization Effect (AAAI 2025 · similarity 0.89)
- DOS: Diverse Outlier Sampling for Out-of-Distribution Detection (ICLR 2024 · similarity 0.87)
- RODEO: Robust Outlier Detection via Exposing Adaptive Out-of-Distribution Samples (ICML 2024 · similarity 0.87)
- Envisioning Outlier Exposure by Large Language Models for Out-of-Distribution Detection (ICML 2024 · similarity 0.86)
- Quantile-Based Maximum Likelihood Training for Outlier Detection (AAAI 2024 · 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.
