AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Semi-Implicit Denoising Diffusion Models (SIDDMs)

yanwu xu, Mingming Gong, Shaoan Xie, Wei Wei, Matthias Grundmann, kayhan Batmanghelich, Tingbo Hou

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Denoising diffusion probabilistic models · 3 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
Decompose denoising matching into adversarial marginal matching and explicit AFD conditional matching (ablation)SIDDMs · CIFAR10FID: 77.15 -> 2.24Improved
Add adversarial marginal matching alongside the AFD term (ablation)SIDDMs · CIFAR10FID: 41.27 -> 2.24Improved
Add the discriminator reconstruction regularizer (ablation)SIDDMs · CIFAR10FID: 3.20 -> 2.24Improved
Use 2 diffusion steps instead of 1 (ablation)SIDDMs · CIFAR10FID: 27.39 -> 3.17Improved
Use 4 diffusion steps instead of 2 (ablation)SIDDMs · CIFAR10FID: 3.17 -> 2.24Improved
Use 16 diffusion steps instead of 4 (ablation)SIDDMs · CIFAR10FID: 2.24 -> 2.15Improved
Increase diffusion steps from 16 to 32 (ablation)SIDDMs · CIFAR10FID: 2.15 -> 2.46Hurt
Vary AFD weight while retaining both objective terms (ablation)SIDDMs · CIFAR10FID: 2.24; 2.63; 2.55No clear effect
Use the semi-implicit objective instead of DDGAN trainingSIDDMs and DDGANs · CIFAR10FID: 3.75 -> 2.24Improved
Use the semi-implicit objective instead of DDGAN trainingSIDDMs and DDGANs · CelebA-HQ-256FID: 7.64 -> 7.37Improved
Use the semi-implicit objective instead of DDGAN trainingSIDDMs and DDGANs · ImageNet1000-64FID: 20.63 -> 3.13Improved
Use SIDDMs rather than the best listed diffusion baselineSIDDMs and ADM · ImageNet1000-64FID: 2.07 -> 3.13Hurt

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