Research map / Label-efficient and robust learning / Recognition and long-tail
Your Dissimilarities Define You: Complementary Learning Exploiting Class Diversities
Dimitrios Katsikas, Nikolaos Passalis, Anastasios Tefas
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 CDLuni to cross-entropy | ResNet-50 · ImageNet val | top-1 error (%): 23.72 -> 22.61 | Improved |
| add CDLuni to cross-entropy | ResNet-101 · ImageNet val | top-1 error (%): 22.78 -> 21.46 | Improved |
| add CDLuni to MaxSup | ResNet-50 · ImageNet val | top-1 error (%): 22.88 -> 22.35 | Improved |
| add CDLuni to MaxSup | ResNet-101 · ImageNet val | top-1 error (%): 21.39 -> 21.52 | Hurt |
| add CDLself to OLS | ResNet-50 · ImageNet val | top-1 error (%): 22.51 -> 22.31 | Improved |
| add CDLuni to CE on DeiT-Small without MixUp/CutMix | DeiT-Small · ImageNet val | top-1 error (%): 25.61 -> 23.12 | Improved |
| add CDLuni to all baseline losses | ResNet-18 and DenseNet-121 · CIFAR-100, TinyImageNet | error rate (%): 24.98 -> 23.92 (CE, RN18 CIFAR-100) | Improved |
| add CDLuni to MaxSup | DenseNet-121 · TinyImageNet | error rate (%): 36.32 -> 35.04 | Improved |
| use CE + CDLself instead of other self-KD methods | ResNet-34 · CIFAR-100 (self-KD) | error rate (%): 22.62 -> 22.44 (vs OLS) | Improved |
| use CE + CDLself instead of OLS on MobileNetV2 TinyImageNet | MobileNetV2 · TinyImageNet (self-KD) | error rate (%): 39.11 -> 38.88 | Improved |
| add CDLself to CE for open-set recognition | ResNet-18 · Open-set CIFAR-100 | OSCR: 64.49 -> 66.82 | Improved |
| add confusing samples with uniform target to CDL (ablation) | CE + CDLself · Open-set CIFAR-100 | OSCR: 66.82 -> 68.15 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MaxSup — MaxSup: Overcoming Representation Collapse in Label Smoothing (NeurIPS 2025) · better on 3 of 3 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
- Which Features are Learnt by Contrastive Learning? On the Role of Simplicity Bias in Class Collapse and Feature Suppression (ICML 2023 · similarity 0.85)
- Rank-Transformed Dissimilarity Profiles for High-Dimensional Classification (NeurIPS 2026 · similarity 0.85)
- Targeted Representation Alignment for Open-World Semi-Supervised Learning (CVPR 2024 · similarity 0.84)
- Neural Collapse for Cross-entropy Class-Imbalanced Learning with Unconstrained ReLU Features Model (ICML 2024 · similarity 0.84)
- HYPO: Hyperspherical Out-Of-Distribution Generalization (ICLR 2024 · similarity 0.84)
- Navigate Beyond Shortcuts: Debiased Learning Through the Lens of Neural Collapse (CVPR 2024 · similarity 0.84)
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.
