Research map / Probabilistic and scientific modeling / Variational and generative sampling
From Variational to Deterministic Autoencoders
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 |
|---|---|---|---|
| Fit a 10-component GMM to the learned latent codes instead of sampling from the fitted single Gaussian | RAE · MNIST | FID: 23.92 -> 9.81 | Helped |
| Fit a 10-component GMM to the learned latent codes instead of sampling from the fitted single Gaussian | RAE · CIFAR-10 | FID: 83.87 -> 76.28 | Helped |
| Fit a 10-component GMM to the learned latent codes instead of sampling from the fitted single Gaussian | RAE · CelebA | FID: 48.20 -> 44.68 | Helped |
| Use a 10-component GMM rather than the fixed isotropic Gaussian prior | WAE · MNIST | FID: 20.42 -> 9.39 | Helped |
| Use a 10-component GMM rather than the fixed isotropic Gaussian prior | WAE · CelebA | FID: 53.67 -> 42.73 | Helped |
| Use decoder L2 regularization instead of gradient penalty | RAE · MNIST samples, 10-component GMM | FID: 11.54 -> 8.69 | Helped |
| Use decoder spectral normalization instead of gradient penalty | RAE · CelebA samples, 10-component GMM | FID: 45.63 -> 40.95 | Helped |
| Choose among gradient penalty, L2 regularization, or spectral normalization | RAE · Image datasets | FID | Mixed |
| Remove latent-code regularization and explicit decoder regularization (ablation) | RAE · MNIST samples, fitted Gaussian | FID: 23.92 -> 58.73 | Hurt |
| Remove latent-code regularization and explicit decoder regularization (ablation) | RAE · MNIST samples, 10-component GMM | FID: 9.81 -> 10.66 | Hurt |
| Remove latent-code regularization and explicit decoder regularization (ablation) | RAE · CelebA samples, fitted Gaussian | FID: 48.20 -> 127.85 | Hurt |
| Remove latent-code regularization but retain decoder L2 regularization (ablation) | RAE · MNIST samples, fitted Gaussian | FID: 22.22 -> 315.15 | Hurt |
Closest papers by meaning
- Shape your Space: A Gaussian Mixture Regularization Approach to Deterministic Autoencoders (NeurIPS 2021 · similarity 0.93)
- MAE: Mutual Posterior-Divergence Regularization for Variational AutoEncoders (ICLR 2019 · similarity 0.92)
- On Implicit Regularization in $β$-VAEs (ICML 2020 · similarity 0.92)
- Variational Autoencoder with Implicit Optimal Priors (AAAI 2019 · similarity 0.92)
- Trading Information between Latents in Hierarchical Variational Autoencoders (ICLR 2023 · similarity 0.92)
- Unscented Autoencoder (ICML 2023 · similarity 0.92)
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.