Research map / Probabilistic and scientific modeling / Variational and generative sampling
Monte Carlo Variational Auto-Encoders
Achille Thin, Nikita Kotelevskii, Arnaud Doucet, Alain Durmus, Eric Moulines, Maxim Panov
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 L-MCVAE with Langevin SIS transitions instead of VAE | VAE · CelebA, 100 epochs | NLL estimate: 10.11 ± 0.32 -> 8.24 ± 0.52 | Improved |
| Use L-MCVAE with Langevin SIS transitions instead of VAE | VAE · CIFAR-10, 100 epochs | NLL estimate: 67.9 ± 0.03 -> 67.5 ± 0.07 | Improved |
| Use L-MCVAE with Langevin SIS transitions instead of VAE | VAE · MNIST, 100 epochs | NLL estimate: 85.22 ± 0.07 -> 84.58 ± 0.39 | Mixed |
| Increase Langevin transition steps (ablation) | L-MCVAE · MNIST, CIFAR-10, and CelebA | NLL estimate: MNIST, K=5: 84.92 ± 0.24 -> K=10: 84.58 ± 0.39; CelebA, K=5: 8.72 ± 0.4 -> K=10: 8.24 ± 0.52; CIFAR-10, K=5: 67.5 ± 0.07 | Mixed |
| Use AIS-based A-MCVAE instead of L-MCVAE | Monte Carlo VAE · Probabilistic PCA on MNIST | ELBO tightness | Mixed |
| Add control variates to the A-MCVAE gradient estimator (ablation) | A-MCVAE · Probabilistic PCA on MNIST | gradient variance | Improved |
| Use learned neural backward kernels instead of approximate reversal kernels (ablation) | L-MCVAE · Probabilistic PCA on MNIST | estimator variance | Hurt |
| Use learned neural backward kernels instead of approximate reversal kernels (ablation) | L-MCVAE · Probabilistic PCA on MNIST | gradient variance | Hurt |
| Increase the number of Langevin steps (ablation) | L-MCVAE · MNIST training | held-out loglikelihood | Improved |
| Use learned rather than fixed temperatures (ablation) | L-MCVAE and A-MCVAE · MNIST, CelebA, and CIFAR-10 | NLL estimate | Mixed |
Closest papers by meaning
- Hamiltonian Variational Auto-Encoder (NeurIPS 2018 · similarity 0.92)
- Consistency Regularization for Variational Auto-Encoders (NeurIPS 2021 · similarity 0.90)
- Simple and Effective VAE Training with Calibrated Decoders (ICML 2021 · similarity 0.90)
- Cauchy–Schwarz Regularized Autoencoder (JMLR 2022 · similarity 0.90)
- Variational Autoencoder with Implicit Optimal Priors (AAAI 2019 · similarity 0.89)
- Forward $\chi^2$ Divergence Based Variational Importance Sampling (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.
