Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-supervised learning of Split Invariant Equivariant representations
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 a transformation-conditioned hypernetwork predictor instead of the original predictor | EquiMod · 3DIEBench validation | rotation prediction R^2: 0.47 -> 0.60 | Helped |
| Add a transformation-conditioned hypernetwork predictor | Only Equivariance · 3DIEBench validation | rotation prediction R^2: 0.51 -> 0.60 | Helped |
| Add an invariant-equivariant split with separate projection heads (ablation) | SIE / VICReg-EquiMod comparison · 3DIEBench validation | rotation prediction R^2: 0.67 -> 0.73 | Helped |
| Add an invariant-equivariant split with separate projection heads (ablation) | SIE / VICReg-EquiMod comparison · 3DIEBench validation | top-1 accuracy: 84.10 -> 82.94 | Hurt |
| Replace the hypernetwork predictor with a linear predictor (ablation) | SIE · 3DIEBench validation | rotation prediction R^2: 0.73 -> 0.38 | Hurt |
| Replace the hypernetwork predictor with an MLP predictor (ablation) | SIE · 3DIEBench validation | rotation prediction R^2: 0.73 -> 0.38 | Hurt |
| Add a hypernetwork predictor instead of an MLP predictor (ablation) | SIE · 3DIEBench validation | top-1 accuracy: 81.38 -> 82.94 | No clear effect |
| Remove the invariant part and use only equivariance (ablation) | SIE · 3DIEBench validation | rotation prediction R^2: 0.73 -> 0.60 | Hurt |
| Add the invariant part to the equivariant representation (ablation) | SIE · 3DIEBench validation | top-1 accuracy: 80.98 -> 82.94 | Helped |
| Train for more epochs (ablation) | SIE · 3DIEBench validation | rotation prediction R^2: 0.68 -> 0.72 -> 0.73 -> 0.73 | Mixed |
| Train for more epochs (ablation) | SIE · 3DIEBench validation | top-1 accuracy: 77.59 -> 81.05 -> 82.12 -> 82.94 | Helped |
| Evaluate embeddings after the projection head rather than representations before it (ablation) | SIE · 3DIEBench validation | rotation prediction R^2: 0.73 -> 0.60 | Hurt |
Closest papers by meaning
- Interpreting Equivariant Representations (ICML 2024 · similarity 0.89)
- In-Context Symmetries: Self-Supervised Learning through Contextual World Models (NeurIPS 2024 · similarity 0.89)
- Neural Isometries: Taming Transformations for Equivariant ML (NeurIPS 2024 · similarity 0.89)
- From Bricks to Bridges: Product of Invariances to Enhance Latent Space Communication (ICLR 2024 · similarity 0.87)
- Improving Equivariant Networks with Probabilistic Symmetry Breaking (ICLR 2025 · similarity 0.87)
- Understanding the Role of Equivariance in Self-supervised 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.