Research map / Label-efficient and robust learning / Knowledge distillation
Deep Mutual Learning
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 mutual peer learning to ResNet-32 | ResNet-32 on CIFAR-100 · CIFAR-100 | Top-1 accuracy (%): 1.20 | Helped |
| Add mutual peer learning to MobileNet | MobileNet on CIFAR-100 · CIFAR-100 | Top-1 accuracy (%): 2.56 | Helped |
| Add mutual peer learning to WRN-28-10 | WRN-28-10 on CIFAR-100 · CIFAR-100 | Top-1 accuracy (%): 1.59 | Helped |
| Add peer mutual learning | MobileNet on Market-1501 without ImageNet pretraining · Market-1501 single-query | mAP: 46.07 -> 52.15 | Helped |
| Add peer mutual learning | ImageNet-pretrained MobileNet on Market-1501 · Market-1501 single-query | mAP: 60.68 -> 68.83 | Helped |
| Replace fixed teacher distillation with mutual learning | ResNet-32 on CIFAR-100, using WRN-28-10 as teacher · CIFAR-100 | Top-1 accuracy (%): 69.48 -> 70.73 | Helped |
| Use a fixed MobileNet teacher rather than independent training | MobileNet student on Market-1501 · Market-1501 | mAP: 46.07 -> 45.16 | Hurt |
| Increase the number of MobileNet peers in the DML cohort | MobileNet on Market-1501 · Market-1501 | mAP | Helped |
| Replace individual peer targets with an averaged ensemble teacher (ablation) | DML cohort on Market-1501 · Market-1501 | mAP | Hurt |
| Use DML instead of independent training | MobileNet on Market-1501 · Market-1501 | mAP: 46.07 -> 52.15 | Helped |
Closest papers by meaning
- Improved Knowledge Distillation via Teacher Assistant (AAAI 2020 · similarity 0.90)
- Learning Student-Friendly Teacher Networks for Knowledge Distillation (NeurIPS 2021 · similarity 0.89)
- Weighted Mutual Learning with Diversity-Driven Model Compression (NeurIPS 2022 · similarity 0.89)
- Hybrid Data-Free Knowledge Distillation (AAAI 2025 · similarity 0.88)
- Cooperative Knowledge Distillation: A Learner Agnostic Approach (AAAI 2024 · similarity 0.87)
- A Kernel Perspective on Distillation-based Collaborative Learning (NeurIPS 2024 · similarity 0.87)
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.