AI Paper Atlas

Research map / Image generation / GANs

Learning Conditional Generative Models for Phase Retrieval

Tobias Uelwer, Sebastian Konietzny, Alexander Oberstrass, Stefan Harmeling

JMLR 2023 · Image generation › GANs · approach: Generative adversarial network theory

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 latent optimization to the CGANPRCGAN-D · MNIST Fourier phase retrievalMSE: 0.0185 -> 0.0009Improved
add latent optimization to the CGANPRCGAN-D · FMNIST Fourier phase retrievalMSE: 0.0149 -> 0.0084Improved
add weight optimization to the CGANPRCGAN-D · CelebA Fourier phase retrievalMSE: 0.0155 -> 0.0115Improved
latent optimization on CIFAR-10PRCGAN-D · CIFAR-10 Fourier phase retrievalMSE: 0.0402 -> 0.0489Hurt
replace E2E regression with PRCGAN-D adversarial trainingE2E · MNIST Fourier phase retrievalMSE: 0.0164 -> 0.0185No clear effect
add adversarial loss (ablation)PRCGAN-L · MNIST Fourier phase retrievalMSE: 0.0289 -> 0.0009Improved
condition on magnitudes rather than labels (ablation)PRCGAN-L · MNIST Fourier phase retrievalMSE: 0.0160 -> 0.0009Improved
feed correct magnitudes to the generator (ablation)PRCGAN-L · MNIST Fourier phase retrievalMSE: 0.0244 -> 0.0009Improved
use LPIPS instead of MAE reconstruction loss (ablation)PRCGAN-D · CelebA Fourier phase retrievalMSE: 0.0155 -> 0.0143Improved
use LPIPS instead of MAE reconstruction loss (ablation)PRCGAN-L · CelebA Fourier phase retrievalMSE: 0.0093 -> 0.0096No clear effect
use MSE instead of MAE reconstruction loss (ablation)PRCGAN-L · CelebA Fourier phase retrievalMSE: 0.0093 -> 0.0125Hurt
add intermediate fully connected layers (ablation)PRCGAN-D · CelebA Fourier phase retrievalMSE: 0.0260 -> 0.0155Improved

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