AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Iterative Ensemble Training with Anti-Gradient Control for Mitigating Memorization in Diffusion Models

Xiao Liu, Xiaoliu Guan, Yu Wu, Jiaxu Miao

ECCV 2024 · Image generation › Fast diffusion sampling · approach: Diffusion models as perception tools · 8 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 IET-AGC training to DDPMDDPM · CIFAR-10MQ0.4: 111 -> 14Improved
add IET-AGC training to DDPMDDPM · CIFAR-10FID: 8.81 -> 8.34Improved
train on images with added Gaussian noiseDDPM · CIFAR-10FID: 8.81 -> 94.61Hurt
use Ambient Diffusion instead of IET-AGCDDPM · CIFAR-10FID: 8.81 -> 11.7Hurt
add iterative ensemble training (IET) with IID shards (ablation)DDPM · CIFAR-10MQ0.5 and FID: MQ0.5 465 -> 382; FID 8.81 -> 7.61Improved
split data in a non-IID way for IET (ablation)IET · CIFAR-10FID: 7.61 -> 12.5Hurt
add anti-gradient control (AGC) alone (ablation)DDPM · CIFAR-10MQ0.5 and FID: MQ0.5 465 -> 154; FID 8.81 -> 11.36Mixed
combine IET with AGC (ablation)AGC · CIFAR-10FID and MQ0.5: FID 11.36 -> 8.34; MQ0.5 154 -> 117Improved
increase skipping threshold lambda from 0.5 to 0.8 (ablation)IET-AGC · CIFAR-10MQ0.5 and FID: MQ0.5 117 -> 45; FID 8.34 -> 10.43Mixed
increase number of shards from 10 to 15 (ablation)IET-AGC · CIFAR-10MQ0.5 and FID: MQ0.5 117 -> 169; FID 8.34 -> 9.1Hurt
apply IET-AGC when fine-tuning Stable DiffusionStable Diffusion · LAION-10ksimilarity score: 0.64 -> 0.37Improved
apply IET-AGC when fine-tuning a pre-trained DDPM for 2 epochspre-trained DDPM · CIFAR-10MQ0.4 and FID: MQ0.4 111 -> 83; FID 8.81 -> 7.93Improved

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