Research map / Label-efficient and robust learning / Self-supervised and contrastive
Implicit variance regularization in non-contrastive SSL
Manu Srinath Halvagal, Axel Laborieux, Friedemann Zenke
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 |
|---|---|---|---|
| remove EMA target network (ablation) | IsoLoss alpha=0.5 · CIFAR-10 | accuracy (%): 91.5 -> 91.5 | No clear effect |
| remove EMA target network (ablation) | IsoLoss alpha=0.5 · CIFAR-100 | accuracy (%): 69.0 -> 64.3 | Hurt |
| remove EMA target network (ablation) | IsoLoss alpha=0.5 · STL-10 | accuracy (%): 89.0 -> 87.4 | Hurt |
| remove EMA target network (ablation) | IsoLoss alpha=0.5 · TinyImageNet | accuracy (%): 44.8 -> 40.4 | Hurt |
| remove EMA target network (ablation) | DirectPred alpha=0.5 · CIFAR-10 | accuracy (%): 92.0 -> 12.1 | Hurt |
| remove EMA target network (ablation) | DirectPred alpha=0.5 · CIFAR-100 | accuracy (%): 66.6 -> 1.6 | Hurt |
| remove EMA target network (ablation) | DirectCopy alpha=1 · CIFAR-10 | accuracy (%): 91.3 -> 12.1 | Hurt |
| remove EMA target network (ablation) | IsoLoss alpha=1 · CIFAR-100 | accuracy (%): 69.3 -> 63.0 | Hurt |
| replace DirectPred loss with IsoLoss (with EMA) | closed-form predictor · TinyImageNet | accuracy (%): 40.1 -> 44.8 | Improved |
| replace DirectPred loss with IsoLoss (with EMA) | closed-form predictor · CIFAR-100 | accuracy (%): 66.6 -> 69.0 | Improved |
| replace DirectPred loss with IsoLoss (with EMA) | closed-form predictor · CIFAR-10 | accuracy (%): 92.0 -> 91.5 | No clear effect |
| use nonlinear predictor BYOL instead of IsoLoss (with EMA) | IsoLoss · CIFAR-10 | accuracy (%): 91.5 -> 92.6 | Mixed |
Closest papers by meaning
- PEIRA: Learning Predictive Encoders through Inter-View Regressor Alignment (NeurIPS 2026 · similarity 0.86)
- Explicit Eigenvalue Regularization Improves Sharpness-Aware Minimization (NeurIPS 2024 · similarity 0.85)
- When Can We Approximate Wide Contrastive Models with Neural Tangent Kernels and Principal Component Analysis? (AAAI 2025 · similarity 0.85)
- Preventing Dimensional Collapse in Self-Supervised Learning via Orthogonality Regularization (NeurIPS 2024 · similarity 0.85)
- Variational Learning Finds Flatter Solutions at the Edge of Stability (NeurIPS 2025 · similarity 0.85)
- Drive vs. Decay: On the Training Dynamics of Joint-Embedding Predictive Architectures (NeurIPS 2026 · similarity 0.85)
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.
