Core Concepts
Optimal Transport theory enhances stability and performance in generative modeling.
Abstract
Optimal Transport (OT) theory bridges distributions efficiently.
OT-based generative models improve stability and performance.
UOTM-SD method gradually adjusts divergence terms for enhanced robustness.
Theoretical analysis and experimental validation support the effectiveness of OT-based GANs.
UOTM-SD outperforms other OT-based models in image generation tasks.
Stats
우리의 방법은 CIFAR-10에서 FID 점수를 2.51로 향상시켰습니다.
UOTM-SD는 CelebA-HQ에서 FID 점수를 5.99로 개선했습니다.
Quotes
"Our UOTM-SD method gradually adjusts divergence terms for enhanced robustness."
"Theoretical analysis and experimental validation support the effectiveness of OT-based GANs."