Research map / Probabilistic and scientific modeling / Variational and generative sampling
Arbitrary Conditional Distributions with Energy
Ryan Strauss, Junier Oliva
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 energy network likelihood instead of proposal-only likelihood (ablation) | ACE · UCI Power, missing rate 0.0 | conditional log-likelihood: 0.583 -> 0.631 | Improved |
| use energy network likelihood instead of proposal-only likelihood (ablation) | ACE · UCI Gas, missing rate 0.0 | conditional log-likelihood: 9.484 -> 9.643 | Improved |
| use energy network likelihood instead of proposal-only likelihood (ablation) | ACE · UCI Hepmass, missing rate 0.0 | conditional log-likelihood: -4.417 -> -3.859 | Improved |
| use energy network likelihood instead of proposal-only likelihood (ablation) | ACE · UCI Miniboone, missing rate 0.5 | conditional log-likelihood: -9.169 -> -5.701 | Improved |
| use energy network likelihood instead of proposal-only likelihood (ablation) | ACE · UCI BSDS, missing rate 0.5 | conditional log-likelihood: 75.767 -> 80.613 | Improved |
| replace ACFlow with ACE proposal network (fully-connected, mixture of Gaussians) | conditional density model · UCI Gas, missing rate 0.0 | conditional log-likelihood: 8.086 -> 9.484 | Improved |
| replace ACFlow with ACE | conditional density model · UCI Hepmass, missing rate 0.0 | conditional log-likelihood: -8.197 -> -3.859 | Improved |
| replace VAEAC with ACE | conditional density model · UCI Power, missing rate 0.0 | conditional log-likelihood: -0.042 -> 0.631 | Improved |
| increase training missing rate from 0.0 to 0.5 | ACE · UCI Hepmass | conditional log-likelihood: -3.859 -> -8.133 | Hurt |
| replace ACFlow with ACE | arbitrary marginal estimation · UCI Miniboone, marginal over 10 dims | marginal log-likelihood: -10.36 -> -7.94 | Improved |
| compare ACE with TAN trained per marginal | ACE · UCI BSDS, marginal over 10 dims | marginal log-likelihood: 20.31 vs 20.44 | No clear effect |
| consistency fine-tuning with variance coefficient 2.0 instead of 0.0 (ablation) | ACE · UCI Hepmass | conditional log-likelihood: -3.555 -> -4.090 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ACFlow — ACFlow: Flow Models for Arbitrary Conditional Likelihoods (ICML 2020) · better on 15 of 15 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
- A Likelihood Based Approach to Distribution Regression Using Conditional Deep Generative Models (ICML 2025 · similarity 0.85)
- Generative Conditional Distributions by Neural (Entropic) Optimal Transport (ICML 2024 · similarity 0.84)
- Partition Tree: Conditional Density Estimation over General Outcome Spaces (NeurIPS 2026 · similarity 0.84)
- Conditional Diffusion Models are Minimax-Optimal and Manifold-Adaptive for Conditional Distribution Estimation (ICLR 2025 · similarity 0.84)
- Density Ratio Estimation with Conditional Probability Paths (ICML 2025 · similarity 0.84)
- Learning normalized image densities via dual score matching (NeurIPS 2025 · similarity 0.84)
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.
