Research map / Label-efficient and robust learning / Self-supervised and contrastive
Intermediate Layers Matter in Momentum Contrastive Self Supervised 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 intermediate-layer MSE loss | MoCo · NIH Chest X-ray, 1% labels, E2E | AUC: 59.2 -> 61.4 | Helped |
| Add intermediate-layer Barlow Twins loss | MoCo · NIH Chest X-ray, 1% labels, E2E | AUC: 59.2 -> 62.9 | Helped |
| Add intermediate-layer MSE loss | MoCo · Diabetic Retinopathy, 1% labels, E2E | AUC: 88.1 -> 93.6 | Helped |
| Add intermediate-layer Barlow Twins loss | MoCo · Diabetic Retinopathy, 1% labels, E2E | AUC: 88.1 -> 92.5 | Helped |
| Add intermediate-layer MSE loss | MoCo · Breast Cancer Histopathology, 1% labels, E2E | F1-score: 82.8 -> 85.1 | Helped |
| Add intermediate-layer Barlow Twins loss | MoCo · Breast Cancer Histopathology, 1% labels, E2E | F1-score: 82.8 -> 84.4 | Helped |
| Add intermediate-layer MSE loss | MoCo · NIH Chest X-ray, 100% labels, E2E | AUC: 82.4 -> 81.5 | No clear effect |
| Add intermediate-layer loss | MoCo · NIH Chest X-ray, 100% labels, linear probing | AUC: MoCo 74.4; MoCo + MSE 74.8; MoCo + Barlow Twins 73.5 | Mixed |
| Apply MSE loss to blocks 1-2-3-4 rather than blocks 2-3-4 (ablation) | MoCo + MSE · NIH Chest X-ray, 100% labels | AUC: 79.5 -> 81.5 | Helped |
| Apply MSE loss to blocks 1-2-3-4 rather than blocks 3-4 (ablation) | MoCo + MSE · NIH Chest X-ray, 100% labels | AUC: 80.8 -> 81.5 | Helped |
| Apply MSE loss to blocks 1-2-3-4 rather than block 4 only (ablation) | MoCo + MSE · NIH Chest X-ray, 100% labels | AUC: 80.1 -> 81.5 | Helped |
| Apply Barlow Twins loss to blocks 1-2-3-4 rather than blocks 2-3-4 (ablation) | MoCo + Barlow Twins · NIH Chest X-ray, 100% labels | AUC: 78.9 -> 80 | Helped |
Closest papers by meaning
- Momentum Contrast for Unsupervised Visual Representation Learning (CVPR 2020 · similarity 0.88)
- Self-Contrastive Learning: Single-Viewed Supervised Contrastive Framework Using Sub-network (AAAI 2023 · similarity 0.87)
- Contrastive learning of global and local features for medical image segmentation with limited annotations (NeurIPS 2020 · similarity 0.86)
- Revisiting Contrastive Methods for Unsupervised Learning of Visual Representations (NeurIPS 2021 · similarity 0.86)
- Self-Supervised Contrastive Learning is Approximately Supervised Contrastive Learning (NeurIPS 2025 · similarity 0.85)
- Systematic Comparison of Semi-supervised and Self-supervised Learning for Medical Image Classification (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.