AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

One-Line-of-Code Data Mollification Improves Optimization of Likelihood-based Generative Models

Ba-Hien Tran, Giulio Franzese, Pietro Michiardi, Maurizio Filippone

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Distributional control of diffusion · 0 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
Add annealed Gaussian data mollification during trainingLikelihood-based generative models · CIFAR10 and CelebA 64FID: 131.15 -> 121.75; 74.62 -> 64.87; 191.98 -> 155.13; 81.25 -> 79.68; 97.59 -> 70.91; 80.19 -> 72.97Improved
Add Gaussian mollificationNVAE · CIFAR10FID: 53.64 -> 52.26Improved
Add Gaussian mollificationMAF · UCI test setstest log-likelihood: -16.32 -> -11.51; -14.87 -> -11.96; -8.27 -> -6.17; -13.03 -> -11.65Improved
Replace Gaussian mollification with blurring mollification (ablation)Real-NVP · CIFAR10 and CelebA 64FID: 121.75 -> 120.88; 79.68 -> 85.40Mixed
Replace Gaussian mollification with blurring mollification (ablation)Glow · CIFAR10 and CelebA 64FID: 64.87 -> 66.70; 70.91 -> 74.74Hurt
Replace Gaussian mollification with blurring mollification (ablation)VAE · CIFAR10 and CelebA 64FID: 155.13 -> 175.40; 72.97 -> 77.29Hurt
Use cosine instead of sigmoid noise schedule (ablation)Real-NVP · CIFAR10FID: 121.75 -> 118.71Improved
Use linear instead of sigmoid noise schedule (ablation)Real-NVP · CIFAR10FID: 121.75 -> 123.93Hurt
Use cosine or linear instead of sigmoid noise schedule (ablation)Glow · CIFAR10FID: 64.87 -> 71.90; 64.87 -> 74.36Hurt
Use cosine instead of sigmoid noise schedule (ablation)VAE, Beta-VAE, and IWAE · CIFAR10FID: 155.13 -> 154.71; 93.90 -> 92.86; 146.70 -> 146.49Improved
Use linear instead of sigmoid noise schedule (ablation)VAE, Beta-VAE, and IWAE · CIFAR10FID: 155.13 -> 156.47; 93.90 -> 93.14; 146.70 -> 149.16Hurt
Gradually reduce mollification rather than use a fixed small amount of noiseLikelihood-based generative models · Optimizationoptimization performanceImproved

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