Research map / Probabilistic and scientific modeling / Variational and generative sampling
Disentangling Factors of Variation with Cycle-Consistent Variational Auto-Encoders
Ananya Harsh Jha, Saket Anand, Maneesh Singh, VSR Veeravasarapu
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 |
|---|---|---|---|
| use reverse cycle-consistency VAE instead of Mathieu et al. | specified/unspecified disentangling · MNIST (z dim 16, s dim 16) | z-space test accuracy (lower is better): 66.83 -> 17.56 | Improved |
| use reverse cycle-consistency VAE instead of Mathieu et al. | specified/unspecified disentangling · MNIST (z dim 64, s dim 64) | z-space test accuracy (lower is better): 72.20 -> 26.55 | Improved |
| use reverse cycle-consistency VAE instead of Szabo et al. | specified/unspecified disentangling · MNIST (z dim 16, s dim 16) | z-space test accuracy (lower is better): 96.08 -> 17.56 | Improved |
| use reverse cycle-consistency VAE instead of Mathieu et al. | specified/unspecified disentangling · LineMod (z dim 64, s dim 256) | z-space test accuracy (lower is better): 89.17 -> 57.17 | Improved |
| use reverse cycle-consistency VAE instead of Mathieu et al. | specified/unspecified disentangling · LineMod (z dim 256, s dim 512) | z-space test accuracy (lower is better): 86.46 -> 57.70 | Improved |
| use reverse cycle-consistency VAE instead of Mathieu et al. | specified/unspecified disentangling · 2D Sprites (z dim 512, s dim 64) | z-space test accuracy (lower is better): 11.98 -> 11.47 | No clear effect |
| use reverse cycle-consistency VAE instead of Mathieu et al. | specified/unspecified disentangling · 2D Sprites (z dim 1024, s dim 512) | z-space test accuracy (lower is better): 12.25 -> 11.61 | No clear effect |
| use reverse cycle-consistency VAE instead of Szabo et al. | specified/unspecified disentangling · 2D Sprites (z dim 512, s dim 64) | z-space test accuracy (lower is better): 99.63 -> 11.47 | Improved |
| increase latent dimensions from 16/16 to 64/64 | Ours · MNIST | z-space test accuracy (lower is better): 17.56 -> 26.55 | Hurt |
| increase latent dimensions from 16/16 to 64/64 | Szabo et al. · MNIST | z-space test accuracy (lower is better): 96.08 -> 98.14 | Hurt |
| use pairwise reverse-cycle loss instead of absolute distance to sampled z (ablation) | Ours reverse cycle loss · MNIST/2D Sprites/LineMod | z-space test accuracy (lower is better) | Improved |
Closest papers by meaning
- Disentangling Disentanglement in Variational Autoencoders (ICML 2019 · similarity 0.92)
- Learning Disentangled Representations with Latent Variation Predictability (ECCV 2020 · similarity 0.92)
- Orthogonality-Enforced Latent Space in Autoencoders: An Approach to Learning Disentangled Representations (ICML 2023 · similarity 0.92)
- Learning Disentangled Joint Continuous and Discrete Representations (NeurIPS 2018 · similarity 0.91)
- NashAE: Disentangling Representations through Adversarial Covariance Minimization (ECCV 2022 · similarity 0.91)
- Overcoming the Disentanglement vs Reconstruction Trade-off via Jacobian Supervision (ICLR 2019 · similarity 0.90)
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.
