AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Smooth Normalizing Flows

Jonas Köhler, Andreas Krämer, Frank Noe

NeurIPS 2021 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Normalizing flow training and design · 17 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use smooth mixture transformations instead of neural spline transformationsAlanine dipeptide flow with pure density estimation · Alanine dipeptide test setFME × 10^4: 12.13 -> 1.04Improved
Use smooth mixture transformations instead of neural spline transformationsAlanine dipeptide flow with pure density estimation · Alanine dipeptide test setNLL: -210.32 -> -211.04No clear effect
Add force matching loss (ablation)Smooth flow · Alanine dipeptide test set; ωk=0FME × 10^4: 1.04 -> 0.32Improved
Add force matching loss (ablation)Smooth flow · Alanine dipeptide test set; ωk=0NLL: -211.04 -> -211.40Improved
Add reverse KLD loss (ablation)Smooth flow · Alanine dipeptide test setNLL: -211.04 -> -206.33Hurt
Add reverse KLD loss (ablation)Smooth flow · Alanine dipeptide test setKLD: 193.91 -> 219.03Mixed
Add reverse KLD and force matching losses (ablation)Smooth flow · Alanine dipeptide test setFME × 10^4: 1.04 -> 0.48Mixed
Train the spline flow with force matching loss (ablation)Spline flow · Alanine dipeptide training runs; ωk=0successful runs: 10/10 failed runsHurt
Train the spline flow with force matching loss (ablation)Spline flow · Alanine dipeptide training runs; ωk=0.1successful runs: 7/10 failed runsHurt
Use smooth transformations as the flow potential rather than spline transformationsAlanine dipeptide molecular dynamics · Microcanonical MDenergy stability: 8×10^{-4} kJ/mol vs 6×10^{-4} kJ/molImproved
Replace analytic inverse evaluation with multi-bin root findingNeural spline flow, dim 2 · Runtime benchmarkinversion runtime factor: 2.1Hurt
Replace analytic inverse evaluation with multi-bin root findingNeural spline flow, dim 2048 · Runtime benchmarkinversion runtime factor: 13.4Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy