Research map / Label-efficient and robust learning / Self-supervised and contrastive
Neural Harmonics: Bridging Spectral Embedding and Matrix Completion in 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 |
|---|---|---|---|
| Use the RQmin trace-maximization objective instead of VICReg | ResNet-18 self-supervised learning · CIFAR-10 validation | top-1 accuracy: 91.15 ± 0.16 -> 91.19 ± 0.13 | No clear effect |
| Use the RQmin trace-maximization objective instead of VICReg | ResNet-18 self-supervised learning · CIFAR-100 validation | top-1 accuracy: 67.57 ± 0.20 -> 68.12 ± 0.26 | Helped |
| Use the RQmin trace-maximization objective instead of VICReg | ResNet-18 self-supervised learning · ImageNet-100 validation | top-1 accuracy: 78.89 ± 0.38 -> 78.98 ± 0.33 | No clear effect |
| Increase projection-head depth (ablation) | SSL models on ImageNet-100 · ImageNet-100 training-set representations | incoherence | Hurt |
| Use backbone representations rather than projection-head embeddings | Pretrained SSL models · ImageNet-100 downstream classification | accuracy | Helped |
Closest papers by meaning
- Contrastive and Non-Contrastive Self-Supervised Learning Recover Global and Local Spectral Embedding Methods (NeurIPS 2022 · similarity 0.90)
- Preventing Dimensional Collapse in Self-Supervised Learning via Orthogonality Regularization (NeurIPS 2024 · similarity 0.85)
- Learning to Discover Iterative Spectral Algorithms (NeurIPS 2026 · similarity 0.85)
- Matrix Information Theory for Self-Supervised Learning (ICML 2024 · similarity 0.84)
- On the Importance of Embedding Norms in Self-Supervised Learning (ICML 2025 · similarity 0.84)
- Understanding Augmentation-based Self-Supervised Representation Learning via RKHS Approximation and Regression (ICLR 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.