GAN Architecture
Generative Adversarial Networks (GANs) are a class of deep learning models designed to generate new data instances that resemble a training dataset. Current research focuses on improving GAN architectures to address challenges like training instability, mode collapse, and the need for large datasets, exploring variations such as conditional GANs (cGANs), and incorporating techniques like hypernetworks and downstream feedback. These advancements are significantly impacting diverse fields, enabling applications ranging from medical image enhancement and data augmentation to synthetic data generation for improving model fairness and efficiency in computationally intensive simulations.
Papers
October 28, 2024
October 8, 2024
September 25, 2024
August 28, 2024
June 26, 2024
June 23, 2024
June 5, 2024
May 25, 2024
March 27, 2024
March 25, 2024
February 26, 2024
February 2, 2024
January 17, 2024
November 15, 2023
September 5, 2023
June 23, 2023
June 15, 2023
May 28, 2023