Research map / Probabilistic and scientific modeling / Variational and generative sampling
Autoregressive Energy Machines
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 the ResMADE-MoG proposal density with AEM energy conditionals | AEM · POWER | log likelihood (nats): 0.61 \pm 0.01 -> 0.70 \pm 0.01 | Helped |
| Replace the ResMADE-MoG proposal density with AEM energy conditionals | AEM · GAS | log likelihood (nats): 12.80 \pm 0.01 -> 13.03 \pm 0.01 | Helped |
| Replace the ResMADE-MoG proposal density with AEM energy conditionals | AEM · HEPMASS | log likelihood (nats): -13.42 \pm 0.01 -> -12.85 \pm 0.01 | Helped |
| Replace the ResMADE-MoG proposal density with AEM energy conditionals | AEM · MINIBOONE | log likelihood (nats): -11.01 \pm 0.23 -> -10.17 \pm 0.26 | Hurt |
| Replace the ResMADE-MoG proposal density with AEM energy conditionals | AEM · BSDS300 | log likelihood (nats): 157.41 \pm 0.14 -> 158.7 1 \pm 0.14 | Hurt |
| Approximate the AEM with a KDE over proposal samples (ablation) | AEM · POWER | log likelihood (nats): 0.70 \pm 0.01 -> 0.65 \pm 0.01 | Hurt |
| Approximate the AEM with a KDE over proposal samples (ablation) | AEM · GAS | log likelihood (nats): 13.03 \pm 0.01 -> 12.89 \pm 0.01 | Hurt |
| Approximate the AEM with a KDE over proposal samples (ablation) | AEM · HEPMASS | log likelihood (nats): -12.85 \pm 0.01 -> -12.87 \pm 0.01 | No clear effect |
| Approximate the AEM with a KDE over proposal samples (ablation) | AEM · MINIBOONE | log likelihood (nats): -10.17 \pm 0.26 -> -10.33 \pm 0.22 | Hurt |
| Use an AEM prior instead of a ResMADE-VAE prior (ablation) | VAE · Dynamically-binarized MNIST | ELBO: -82.96 \pm 0.23 -> -82.92 \pm 0.23 | No clear effect |
| Use an AEM prior instead of a ResMADE-VAE prior (ablation) | VAE · Dynamically-binarized MNIST | log p(x): -79.89 \pm 0.21 -> -79.87 \pm 0.21 | No clear effect |
Closest papers by meaning
- Learning normalized image densities via dual score matching (NeurIPS 2025 · similarity 0.85)
- Energy-Based Modelling for Discrete and Mixed Data via Heat Equations on Structured Spaces (NeurIPS 2024 · similarity 0.84)
- Joint Learning of Energy-based Models and their Partition Function (ICML 2025 · similarity 0.83)
- Free energy Estimation on Any State Space (NeurIPS 2026 · similarity 0.83)
- BARNN: A Bayesian Autoregressive and Recurrent Neural Network (ICML 2025 · similarity 0.83)
- Energy-based Automated Model Evaluation (ICLR 2024 · similarity 0.82)
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.