Research map / Probabilistic and scientific modeling / Molecules and proteins
Topographic VAEs learn Equivariant Capsules
T. Anderson Keller, Max Welling
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 |
|---|---|---|---|
| add temporal coherence (L=S/2) vs plain VAE | TVAE · MNIST | equivariance error: 13274 -> 574 | Improved |
| add temporal coherence with L=5S/36 | TVAE · MNIST | log-likelihood: -189.0 -> -186.0 | Improved |
| remove temporal coherence (L=0) from TVAE (ablation) | TVAE · MNIST | log-likelihood: -186.0 -> -218.5 | Hurt |
| TVAE L=0 vs VAE | TVAE · MNIST | equivariance error: 13274 -> 3217 | Improved |
| replace TVAE with BubbleVAE at L=5S/36 (ablation) | BubbleVAE · MNIST | log-likelihood: -186.0 -> -191.4 | Hurt |
| reduce L from S/3 to S/6 (ablation) | TVAE · dSprites | CapCorr X: 1.0 -> 0.67 | Hurt |
| reduce L from S/6 to 0 (ablation) | TVAE · dSprites | CapCorr X: 0.67 -> 0.17 | Hurt |
| TVAE L=S/3 vs BubbleVAE | TVAE · dSprites | equivariance error: 1951 -> 1034 | Improved |
| L=0 TVAE and BubbleVAE vs VAE | TVAE L=0 · dSprites | CapCorr: 0.18 -> 0.17 | No clear effect |
| increase K from 3 to 9 at L=5S/36 (ablation) | TVAE · MNIST | equivariance error: 3246.6 -> 2606.3 | Improved |
| increase K from 3 to 9 at L=5S/36 (ablation) | TVAE · MNIST | log-likelihood: -186.0 -> -190.6 | Hurt |
| increase K from 9 to 18 at L=0 (ablation) | TVAE · MNIST | equivariance error: 3216.6 -> 1090.3 | Improved |
Closest papers by meaning
- Flow Equivariant Recurrent Neural Networks (NeurIPS 2025 · similarity 0.88)
- Approximately Equivariant Neural Processes (NeurIPS 2024 · similarity 0.88)
- A Probabilistic Approach to Learning the Degree of Equivariance in Steerable CNNs (ICML 2024 · similarity 0.87)
- Variational Partial Group Convolutions for Input-Aware Partial Equivariance of Rotations and Color-Shifts (ICML 2024 · similarity 0.87)
- EQ-VAE: Equivariance Regularized Latent Space for Improved Generative Image Modeling (ICML 2025 · similarity 0.87)
- Lie Group Decompositions for Equivariant Neural Networks (ICLR 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.
