Our understanding of GAN (generative adversarial net) training is still very limited since it is a non-convex-non-concave min-max optimization. As a result, most recent studies focused on local analysis. In this talk, we discuss how to perform a global analysis of GANs. We prove that the original JS-GAN has exponentially many bad strict local minima which are perceived as mode-collapse. We show that a 2-line modification to JS-GAN called relativistic standard GAN (RS-GAN) eliminates all bad basins. We extend the two results to a large class of losses as well: for separable GANs (including JS-GAN, hinge-GAN, LS-GAN) exponentially many bad basins exist, while for R-GANs no bad basins exist. The effectiveness of R-GANs were verified by a few empirical works before (e.g. ESR-GAN in super resolution). Based on theory, we predict that R-GANs has a bigger advantage for narrower neural nets, and our experiments verify that R-GANs (e.g. RS-GAN) can beat their separable counter-parts (e.g. JS-GAN) by 5-10 FID scores in narrower nets. Our theory also implies that the advantage is larger for higher-dimensional images; we show that for high-resolution images like LSUN, while JS-GAN only generates noise, RS-GAN can generate quite good images.
8月21日
10:30am - 11:30am
地點
https://hkust.zoom.us/j/5616960008
講者/表演者
Prof. Ruoyu Sun
University of Illinois at Urbana-Champaign
主辦單位
Department of Mathematics
聯絡方法
mathseminar@ust.hk
付款詳情
對象
Alumni, Faculty and Staff, PG Students, UG Students
語言
英語
其他活動
7月14日
研討會, 演講, 講座
IAS / School of Science Joint Lecture - Boron Clusters
Abstract The study of carbon clusters led to the discoveries of fullerenes, carbon nanotubes, and graphene. Are there other elements that can form similar nanostructures? To answer this questio...
5月15日
研討會, 演講, 講座
IAS / School of Science Joint Lecture - Laser Spectroscopy of Computable Atoms and Molecules with Unprecedented Accuracy
Abstract Precision spectroscopy of the hydrogen atom, a fundamental two-body system, has been instrumental in shaping quantum mechanics. Today, advances in theory and experiment allow us to ext...