Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning Disentangled Representations with Latent Variation Predictability
Xinqi Zhu, Chang Xu, Dacheng Tao
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 the VP loss to CascadeVAE | CascadeVAE · Dsprites | VP score: 59.2 -> 65.5 | Improved |
| add the VP loss to CascadeVAE | CascadeVAE · Dsprites | FactorVAE score: 91.3 -> 91.7 | No clear effect |
| add the VP loss to CascadeVAE | CascadeVAE · 3DShapes | VP score: 62.3 -> 66.4 | Improved |
| add the VP loss to CascadeVAE | CascadeVAE · 3DShapes | FactorVAE score: 94.7 -> 95.6 | No clear effect |
| increase alpha from 0.001 to 0.1 (ablation) | VPGAN-flat · CelebA | VP score: 58.4 -> 64.5 | Improved |
| increase alpha from 0.001 to 0.1 (ablation) | VPGAN-flat · CelebA | FID: 22.3 -> 32.8 | Hurt |
| increase alpha from 0.001 to 0.1 (ablation) | VPGAN-hierar · CelebA | VP score: 64.6 -> 70.3 | Improved |
| add hierarchical latent inputs instead of flat (ablation) | VPGAN · CelebA | VP score: 64.5 -> 70.3 | Improved |
| add hierarchical latent inputs instead of flat (ablation) | VPGAN · CelebA | FID: 32.8 -> 56.9 | Hurt |
| replace InfoGAN with VPGAN-hierar | GAN · CelebA | VP score: 34.5 -> 70.3 | Improved |
| replace InfoGAN with VPGAN | GAN · 3DChairs | VP score: 36.7 -> 42.0 | Improved |
| replace InfoGAN with VPGAN | GAN · 3DChairs | FID: 30.4 -> 32.1 | Hurt |
Closest papers by meaning
- Disentangling Factors of Variation with Cycle-Consistent Variational Auto-Encoders (ECCV 2018 · similarity 0.92)
- Overcoming the Disentanglement vs Reconstruction Trade-off via Jacobian Supervision (ICLR 2019 · similarity 0.90)
- Learning Disentangled Representation by Exploiting Pretrained Generative Models: A Contrastive Learning View (ICLR 2022 · similarity 0.90)
- Synergy Between Sufficient Changes and Sparse Mixing Procedure for Disentangled Representation Learning (ICLR 2025 · similarity 0.88)
- Disentangled Representation Learning via Flow Matching (NeurIPS 2026 · similarity 0.88)
- $\alpha$TC-VAE: On the relationship between Disentanglement and Diversity (ICLR 2024 · similarity 0.88)
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.
