Research map / Probabilistic and scientific modeling / Variational and generative sampling
Look Ma, No Latent Variables: Accurate Cutset Networks via Compilation
Tahrima Rahman, Shasha Jin, Vibhav Gogate
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 |
|---|---|---|---|
| learn cutset network with latent-model compilation (CNxD) instead of from data alone (CN) | cutset networks · 20 density datasets | average test log-likelihood: -57.05 -> -55.40 | Improved |
| learned structure (CNxD) instead of random deep structure (CNR) (ablation) | cutset networks · 20 density datasets | average test log-likelihood: -62.81 -> -55.40 | Improved |
| CNxD versus ACBN | CNxD · 20 density datasets | test log-likelihood: -56.10 -> -55.40 | Improved |
| CNxD versus ACMN | CNxD · 20 density datasets | test log-likelihood: -55.78 -> -55.40 | Mixed |
| CNxD versus PSDD | CNxD · 20 density datasets | test log-likelihood: -59.06 -> -55.40 | Improved |
| CNxD versus latent models MT and BCN | CNxD · 20 density datasets | average test log-likelihood: -55.33 -> -55.40 ; BCN -53.91 | Hurt |
| CNxD versus CN for MAP at 20% evidence | CNxD · MAP completion | avg oracle log-likelihood: -26.48 -> -25.65 | Improved |
| CNxD versus CN for MAP at 50% evidence | CNxD · MAP completion | avg oracle log-likelihood: -38.77 -> -37.91 | Improved |
| CNxD versus CN for MAP at 80% evidence | CNxD · MAP completion | avg oracle log-likelihood: -44.19 -> -43.65 | Improved |
| exact MAP on CNxD versus approximate marginal MAP on MT | CNxD · MAP completion 20% evidence | avg oracle log-likelihood: -27.82 -> -25.65 | Improved |
| exact MAP on CNxD versus approximate marginal MAP on BCN | CNxD · MAP completion 50% evidence | avg oracle log-likelihood: -39.36 -> -37.91 | Improved |
| increase depth of CNxD (ablation) | CNxD · density estimation vs running time | test log-likelihood | Improved |
Closest papers by meaning
- Learning Tractable Probabilistic Models from Inconsistent Local Estimates (NeurIPS 2022 · similarity 0.89)
- Learning Ensembles of Cutset Networks (AAAI 2016 · similarity 0.88)
- Identifying Metric Structures of Deep Latent Variable Models (ICML 2025 · similarity 0.87)
- Online but Accurate Inference for Latent Variable Models with Local Gibbs Sampling (JMLR 2017 · similarity 0.86)
- Efficient Marginalization of Discrete and Structured Latent Variables via Sparsity (NeurIPS 2020 · similarity 0.86)
- Lagging Inference Networks and Posterior Collapse in Variational Autoencoders (ICLR 2019 · similarity 0.86)
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.
