Research map / Label-efficient and robust learning / Recognition and long-tail
DROCC: Deep Robust One-Class Classification
Sachin Goyal, Aditi Raghunathan, Moksh Jain, Harsha Vardhan Simhadri, Prateek Jain
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 DROCC's manifold-based adversarial negative generation instead of DeepSVDD | One-class anomaly detection · Arrhythmia | F1-Score: 0.54 ± 0.01 -> 0.69 ± 0.02 | Improved |
| Use DROCC's manifold-based adversarial negative generation instead of DeepSVDD | One-class anomaly detection · Thyroid | F1-Score: 0.73 ± 0.00 -> 0.78 ± 0.03 | Improved |
| Use DROCC's manifold-based adversarial negative generation instead of DeepSVDD | One-class anomaly detection · Abalone | F1-Score: 0.62 ± 0.01 -> 0.68 ± 0.02 | Improved |
| Use DROCC instead of DeepSVDD | One-class anomaly detection · ImageNet-10 French Horn | AUC: 53.06 ± 6.52 -> 76.97 ± 1.67 | Improved |
| Add a Mahalanobis distance learned from feature influence to the DROCC-LF objective | DROCC-LF · MNIST 0 vs. 1 with half-zero negatives | Recall at 3% FPR: 98.16% -> 99.4% | Improved |
| Use outlier exposure with DROCC instead of DeepSAD | Limited-negative classification · MNIST 0 vs. 1 with half-zero negatives | Recall at 3% FPR: 90.91% -> 98.16% | Improved |
| Sample synthetic negatives randomly rather than generating them by gradient ascent (ablation) | DROCC · CIFAR-10 Deer | AUC: 73.62 ± 2.00 -> 56.09 ± 1.19 | Hurt |
| Sample synthetic negatives randomly rather than generating them by gradient ascent (ablation) | DROCC · CIFAR-10 Cat | AUC: 67.13 ± 1.51 -> 67.33 ± 0.72 | No clear effect |
| Sample synthetic negatives randomly rather than generating them by gradient ascent (ablation) | DROCC · CIFAR-10 Frog | AUC: 74.43 ± 0.92 -> 74.82 ± 1.77 | No clear effect |
| Sample synthetic negatives randomly rather than generating them by gradient ascent (ablation) | DROCC · CIFAR-10 Airplane | AUC: 81.66 ± 0.22 -> 79.67 ± 2.09 | Hurt |
| Augment CIFAR-10 images with flips and small rotations (ablation) | GEOM · CIFAR-10 Deer | AUC: 87.8 ± 0.2 -> 65.7 ± 1.7 | Hurt |
| Augment CIFAR-10 images with flips and small rotations (ablation) | DROCC · CIFAR-10 Deer | AUC: 73.6 ± 2.0 -> 71.1 ± 2.9 | No clear effect |
Later papers that compared with this method
3 later accepted papers report DROCC as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Deep Orthogonal Hypersphere Compression for Anomaly Detection (ICLR 2024 · better on 2 of 3 shared results)
- Dense Projection for Anomaly Detection (AAAI 2024 · better on 1 of 1 shared result)
- Perturbation Learning Based Anomaly Detection (NeurIPS 2022 · better on 6 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning and Evaluating Representations for Deep One-Class Classification (ICLR 2021 · similarity 0.85)
- Deep One-Class Classification via Interpolated Gaussian Descriptor (AAAI 2022 · similarity 0.85)
- Robust One-Class Classification with Signed Distance Function using 1-Lipschitz Neural Networks (ICML 2023 · similarity 0.85)
- Few-Shot One-Class Classification via Meta-Learning (AAAI 2021 · similarity 0.84)
- Generalized One-Class Learning Using Pairs of Complementary Classifiers (TPAMI 2021 · similarity 0.84)
- Robust Visual Recognition with Class-Imbalanced Open-World Noisy Data (AAAI 2024 · similarity 0.81)
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.
