Timezone: »
Poster
f-GANs in an Information Geometric Nutshell
Richard Nock · Zac Cranko · Aditya K Menon · Lizhen Qu · Robert Williamson
Nowozin \textit{et al} showed last year how to extend the GAN \textit{principle} to all $f$-divergences. The approach is elegant but falls short of a full description of the supervised game, and says little about the key player, the generator: for example, what does the generator actually converge to if solving the GAN game means convergence in some space of parameters? How does that provide hints on the generator's design and compare to the flourishing but almost exclusively experimental literature on the subject? In this paper, we unveil a broad class of distributions for which such convergence happens --- namely, deformed exponential families, a wide superset of exponential families ---. We show that current deep architectures are able to factorize a very large number of such densities using an especially compact design, hence displaying the power of deep architectures and their concinnity in the $f$-GAN game. This result holds given a sufficient condition on \textit{activation functions} --- which turns out to be satisfied by popular choices. The key to our results is a variational generalization of an old theorem that relates the KL divergence between regular exponential families and divergences between their natural parameters. We complete this picture with additional results and experimental insights on how these results may be used to ground further improvements of GAN architectures, via (i) a principled design of the activation functions in the generator and (ii) an explicit integration of proper composite losses' link function in the discriminator.
Author Information
Richard Nock (Data61, The Australian National University & The University of Sydney)
Zac Cranko (The Australian National University & Data61)
Aditya K Menon (Data61/CSIRO)
Lizhen Qu (Data61)
Robert Williamson (Australian National University & Data61)
Related Events (a corresponding poster, oral, or spotlight)
-
2017 Spotlight: f-GANs in an Information Geometric Nutshell »
Wed. Dec 6th 07:20 -- 07:25 PM Room Hall C
More from the Same Authors
-
2019 : Poster Session »
Gergely Flamich · Shashanka Ubaru · Charles Zheng · Josip Djolonga · Kristoffer Wickstrøm · Diego Granziol · Konstantinos Pitas · Jun Li · Robert Williamson · Sangwoong Yoon · Kwot Sin Lee · Julian Zilly · Linda Petrini · Ian Fischer · Zhe Dong · Alexander Alemi · Bao-Ngoc Nguyen · Rob Brekelmans · Tailin Wu · Aditya Mahajan · Alexander Li · Kirankumar Shiragur · Yair Carmon · Linara Adilova · SHIYU LIU · Bang An · Sanjeeb Dash · Oktay Gunluk · Arya Mazumdar · Mehul Motani · Julia Rosenzweig · Michael Kamp · Marton Havasi · Leighton P Barnes · Zhengqing Zhou · Yi Hao · Dylan Foster · Yuval Benjamini · Nati Srebro · Michael Tschannen · Paul Rubenstein · Sylvain Gelly · John Duchi · Aaron Sidford · Robin Ru · Stefan Zohren · Murtaza Dalal · Michael A Osborne · Stephen J Roberts · Moses Charikar · Jayakumar Subramanian · Xiaodi Fan · Max Schwarzer · Nicholas Roberts · Simon Lacoste-Julien · Vinay Prabhu · Aram Galstyan · Greg Ver Steeg · Lalitha Sankar · Yung-Kyun Noh · Gautam Dasarathy · Frank Park · Ngai-Man (Man) Cheung · Ngoc-Trung Tran · Linxiao Yang · Ben Poole · Andrea Censi · Tristan Sylvain · R Devon Hjelm · Bangjie Liu · Jose Gallego-Posada · Tyler Sypherd · Kai Yang · Jan Nikolas Morshuis -
2019 Poster: A Primal-Dual link between GANs and Autoencoders »
Hisham Husain · Richard Nock · Robert Williamson -
2018 Poster: A loss framework for calibrated anomaly detection »
Aditya Menon · Robert Williamson -
2018 Poster: Constant Regret, Generalized Mixability, and Mirror Descent »
Zakaria Mhammedi · Robert Williamson -
2018 Spotlight: Constant Regret, Generalized Mixability, and Mirror Descent »
Zakaria Mhammedi · Robert Williamson -
2018 Spotlight: A loss framework for calibrated anomaly detection »
Aditya Menon · Robert Williamson -
2016 Poster: A scaled Bregman theorem with applications »
Richard Nock · Aditya Menon · Cheng Soon Ong -
2016 Poster: On Regularizing Rademacher Observation Losses »
Richard Nock -
2015 Workshop: Learning and privacy with incomplete data and weak supervision »
Giorgio Patrini · Tony Jebara · Richard Nock · Dimitrios Kotzias · Felix Xinnan Yu -
2015 Poster: Learning with Symmetric Label Noise: The Importance of Being Unhinged »
Brendan van Rooyen · Aditya Menon · Robert Williamson -
2015 Spotlight: Learning with Symmetric Label Noise: The Importance of Being Unhinged »
Brendan van Rooyen · Aditya Menon · Robert Williamson -
2014 Poster: From Stochastic Mixability to Fast Rates »
Nishant Mehta · Robert Williamson -
2014 Poster: (Almost) No Label No Cry »
Giorgio Patrini · Richard Nock · Tiberio Caetano · Paul Rivera -
2014 Spotlight: (Almost) No Label No Cry »
Giorgio Patrini · Richard Nock · Tiberio Caetano · Paul Rivera -
2014 Oral: From Stochastic Mixability to Fast Rates »
Nishant Mehta · Robert Williamson -
2012 Poster: Mixability in Statistical Learning »
Tim van Erven · Peter Grünwald · Mark Reid · Robert Williamson -
2011 Workshop: Relations between machine learning problems - an approach to unify the field »
Robert Williamson · John Langford · Ulrike von Luxburg · Mark Reid · Jennifer Wortman Vaughan -
2011 Poster: Composite Multiclass Losses »
Elodie Vernet · Robert Williamson · Mark Reid -
2009 Workshop: Clustering: Science or art? Towards principled approaches »
Margareta Ackerman · Shai Ben-David · Avrim Blum · Isabelle Guyon · Ulrike von Luxburg · Robert Williamson · Reza Zadeh