Research map / Probabilistic and scientific modeling / Variational and generative sampling
A-NICE-MC: Adversarial Training for MCMC
Jiaming Song, Shengjia Zhao, Stefano Ermon
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 A-NICE-MC instead of HMC | MCMC sampling · mog2, mog6, ring5 | ESS: 1.00 -> 355.39; 1.00 -> 320.03; 0.43 -> 155.57 | Improved |
| Use A-NICE-MC instead of HMC | MCMC sampling · ring | ESS: 1000.00 -> 1000.00 | No clear effect |
| Use A-NICE-MC instead of HMC | MCMC sampling · ring, mog2, mog6, ring5 | ESS/s: 121212 -> 128205; 78 -> 50409; 39 -> 40768; 29 -> 19325 | Improved |
| Use A-NICE-MC instead of HMC | Bayesian logistic regression sampling · german, heart, australian | ESS: 2178.00 -> 926.49; 5000.00 -> 1251.16; 1345.82 -> 1015.75 | Hurt |
| Use A-NICE-MC instead of HMC | Bayesian logistic regression sampling · german, heart, australian | ESS/s: 216.17 -> 1289.03; 1005.03 -> 3204.00; 289.11 -> 1857.37 | Improved |
| Use A-NICE-MC instead of HMC | MCMC sampling · ring5 | R hat: 1.26 -> 1.002 | Improved |
| Add a pairwise discriminator (ablation) | A-NICE-MC image-chain training · CelebA | autocorrelation | Improved |
| Add shortcut connections (ablation) | MGAN image-chain model · CelebA | mode-transition speed | Hurt |
| Train for more iterations and bootstrap rounds | A-NICE-MC · synthetic energy functions | ESS | Improved |
Later papers that compared with this method
2 later accepted papers report A-NICE-MC as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Ai-sampler: Adversarial Learning of Markov kernels with involutive maps (ICML 2024 · better on 0 of 1 shared result)
- Involutive MCMC: a Unifying Framework (ICML 2020 · better on 6 of 8 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Ai-sampler: Adversarial Learning of Markov kernels with involutive maps (ICML 2024 · similarity 0.89)
- Markovian Flow Matching: Accelerating MCMC with Continuous Normalizing Flows (NeurIPS 2024 · similarity 0.87)
- AutoStep: Locally adaptive involutive MCMC (ICML 2025 · similarity 0.87)
- Parameter Expanded Stochastic Gradient Markov Chain Monte Carlo (ICLR 2025 · similarity 0.87)
- Learning Extrapolative Sequence Transformations from Markov Chains (ICML 2025 · similarity 0.87)
- Microcanonical Langevin Ensembles: Advancing the Sampling of Bayesian Neural Networks (ICLR 2025 · similarity 0.87)
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.
