AI Paper Atlas

Research map / Image generation / GANs

Least Squares Generative Adversarial Networks

Xudong Mao, Qing Li, Haoran Xie, Raymond Y.K. Lau, Zhen Wang, Stephen Paul Smolley

ICCV 2017 · Image generation › GANs · approach: Stabilized GAN training · 5,579 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
replace sigmoid cross entropy loss with least squares lossDCGAN without BN in G, Adam · LSUN bedroomsuccessful training runs: 5 of 10 succeed vs never successfulImproved
replace sigmoid cross entropy loss with least squares lossDCGAN without BN in G and D, RMSProp · LSUN bedroomimage quality / mode collapseImproved
replace sigmoid cross entropy loss with least squares lossDCGAN without BN in G, RMSProp · LSUN bedroomimage qualityNo clear effect
replace sigmoid cross entropy loss with least squares lossDCGAN without BN in G and D, Adam · LSUN bedroommode collapseNo clear effect
replace sigmoid cross entropy loss with least squares loss3-layer FC GAN · 2D mixture of 8 Gaussiansmode coverage: regular GAN collapses from step 15k; LSGAN learns the mixtureImproved
use RMSProp instead of Adam (ablation)regular GAN without BN in G · LSUN bedroomsuccessful learningImproved
set a-b-c coding to b-c=1,b-a=2 versus 0-1 coding (c=b) (ablation)LSGAN · LSUNimage qualityNo clear effect

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