Abstract | ||
---|---|---|
Which Generative Adversarial Networks (GANs) generates the most plausible images? has been a frequently asked question among researchers. To address this problem, we first propose an emph{incomplete} U-statistics estimate of maximum mean discrepancy $mathrm{MMD}_{inc}$ to measure the distribution discrepancy between generated and real images. $mathrm{MMD}_{inc}$ enjoys the advantages of asymptotic normality, computation efficiency, and model agnosticity. We then propose a GANs analysis framework to select and test the best member in GANs family using the Post Selection Inference (PSI) with $mathrm{MMD}_{inc}$. In the experiments, we adopt the proposed framework on 7 GANs variants and compare their $mathrm{MMD}_{inc}$ scores. |
Year | Venue | Field |
---|---|---|
2018 | ICLR | Maximum mean discrepancy,Discrete mathematics,Mathematical optimization,Inference,Real image,Mathematics,Computation,Asymptotic distribution |
DocType | Volume | Citations |
Journal | abs/1802.05411 | 0 |
PageRank | References | Authors |
0.34 | 9 | 6 |
Name | Order | Citations | PageRank |
---|---|---|---|
Yao-Hung Hubert Tsai | 1 | 4 | 1.46 |
Makoto Yamada | 2 | 459 | 43.38 |
Denny C.-Y. Wu | 3 | 1 | 4.06 |
Ruslan Salakhutdinov | 4 | 12190 | 764.15 |
Ichiro Takeuchi | 5 | 132 | 23.25 |
kenji fukumizu | 6 | 1683 | 158.91 |