ISyE Seminar Series: Sewoong Oh

"The Power of Two Samples in Generative Adversarial Networks"

Presentation by Professor Sewoong Oh
Paul G. Allen School of Computer Science and Engineering
University of Washington

Wednesday, September 18
3:15pm - Refreshments, Lind Hall 305
3:30pm - Graduate Seminar, Lind Hall 305



Prof. Oh brings the tools from Blackwell’s seminal result on comparing two stochastic experiments from 1953, to shine a new light on a modern application of great interest: Generative Adversarial Networks (GAN). Binary hypothesis testing is at the center of training GANs, where a trained neural network (called a critic) determines whether a given samples from the real data or the generated (fake) data. By jointly training the generator and the critic, the hope is that eventually the trained generator will generate realistic samples. One of the major challenges in GAN is known as “mode collapse”; the lack of diversity in the samples generated by thus trained generators. Oh proposes a new training framework, where the critic is fed with multiple samples jointly (which Oh calls packing), as opposed to each samples separately as done in standard GAN training. With this simple but fundamental departure from standard GANs, experimental results show that the diversity of the generated samples improve significantly. Oh analyzes this practical gain by first providing a formal mathematical definition of mode collapse and making a fundamental connection between the idea of packing and the intensity of mode collapse. Precisely, Oh shows that the packed critic naturally penalizes mode collapse, thus encouraging generators with less mode collapse. The analyses critically rely on operational interpretation of hypothesis testing and corresponding data processing inequalities, which lead to sharp analyses with simple proofs. For this talk, Oh will assume no prior background on GANs. 



Sewoong Oh is an Associate Professor of Paul G. Allen School of Computer Science and Engineering at University of Washington. He received his PhD from the department of Electrical Engineering at Stanford University. Following his PhD, he worked as a postdoctoral researcher at Laboratory for Information and Decision Systems (LIDS) at MIT. His research interest is in theoretical machine learning, including generative adversarial networks and saddle-point problems, and privacy and blockchains. He was co-awarded the best paper award at SIGMETRICS in 2015, NSF CAREER award in 2016 and GOOGLE Faculty Research Award.


Seminar Video:

Start date
Wednesday, Sept. 18, 2019, 3:15 p.m.

Lind Hall
Room 305