In this article, we mathematically study several GAN related topics, including Inception score, label smoothing, gradient vanishing and the -log(D(x)) alternative. --- An advanced version is included in arXiv:1703.02000 "Activation Maximization Generative Adversarial Nets". Please refer Section 6 in 1703.02000 for detailed analysis on Inception Score, and refer its appendix for the discussions on Label Smoothing, Gradient Vanishing and -log(D(x)) Alternative.
Cite
@article{arxiv.1708.01729,
title = {Inception Score, Label Smoothing, Gradient Vanishing and -log(D(x)) Alternative},
author = {Zhiming Zhou and Weinan Zhang and Jun Wang},
journal= {arXiv preprint arXiv:1708.01729},
year = {2018}
}
Comments
An advanced version is included in arXiv:1703.02000 "Activation Maximization Generative Adversarial Nets"