AI Paper Atlas

Research map / Image generation / GANs

EigenGAN: Layer-Wise Eigen-Learning for GANs

Zhenliang He, Meina Kan, Shiguang Shan

ICCV 2021 · Image generation › GANs · approach: Stabilized GAN training · 57 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 layer-wise subspace models (ablation)EigenGAN generator · CelebA face generation, role of layer-wise latent variablesshare of major variation captured by layer-wise latentsImproved
add bottom noise besides layer-wise latentsEigenGAN generator · CelebA face generationcaptured variationImproved
increase data rank from 5 to 20 (subspace rank 5 or more)Linear EigenGAN · toy datasets, Gaussian datacosine similarity with PCA: 1.00 -> 0.78 (KL-f-GAN, 5→1 vs 20→5)Hurt
change GAN loss among KL-f-GAN, Vanilla, WGAN, LSGAN, HingeLinear EigenGAN · toy datasetscosine similarity with PCA: 0.99 -> 1.00 (10→1, WGAN vs Vanilla GAN)No clear effect
change data distribution from Gaussian to uniformLinear EigenGAN · toy datasets, 20→1 KL-f-GANcosine similarity with PCA: 0.97 -> 0.89Hurt
change data distribution from Gaussian to uniformLinear EigenGAN · toy datasets, 10→3 WGANcosine similarity with PCA: 0.89 -> 0.93No 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