Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variational Lossy Autoencoder
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 |
|---|---|---|---|
| replace IAF posterior with AF prior (factorized decoder) | VAE on statically binarized MNIST · Statically Binarized MNIST | NLL Test: 79.88 -> 79.30 | Helped |
| add autoregressive PixelCNN decoder on top of AF prior VAE | VLAE · Statically Binarized MNIST | NLL Test: 79.30 -> 79.03 | Helped |
| add AF prior to PixelCNN decoder VAE (ablation) | PixelCNN decoder VAE · Dynamically binarized MNIST | NLL Test: 79.48 -> 78.94 | Helped |
| add VAE latent code to PixelCNN decoder (vs unconditional PixelCNN) (ablation) | PixelCNN decoder · Dynamically binarized MNIST | NLL Test: 87.55 -> 79.48 | Helped |
| add VAE latent code to PixelCNN decoder (vs unconditional decoder) (ablation) | VLAE · OMNIGLOT | NLL Test: 95.02 -> 90.98 | Helped |
| add VAE latent code to PixelCNN decoder (vs unconditional decoder) (ablation) | VLAE · Caltech-101 Silhouettes | NLL Test: 89.26 -> 77.36 | Helped |
| per-dataset hyperparameter tuning | VLAE · OMNIGLOT | NLL Test: 90.98 -> 89.83 | Helped |
| use small receptive field PixelCNN decoder (lossy code) vs factorized decoder | VLAE lossy compression · Statically Binarized MNIST | average latent code bits: 37.3 -> 19.2 bits | Helped |
Closest papers by meaning
- Variational Autoencoders Pursue PCA Directions (by Accident) (CVPR 2019 · similarity 0.90)
- MAE: Mutual Posterior-Divergence Regularization for Variational AutoEncoders (ICLR 2019 · similarity 0.90)
- PixelVAE: A Latent Variable Model for Natural Images (ICLR 2017 · similarity 0.90)
- Consistency Regularization for Variational Auto-Encoders (NeurIPS 2021 · similarity 0.90)
- Trading Information between Latents in Hierarchical Variational Autoencoders (ICLR 2023 · similarity 0.90)
- A Statistical Analysis of Wasserstein Autoencoders for Intrinsically Low-dimensional Data (ICLR 2024 · similarity 0.89)
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.