Timezone: »
Oral
Chi-square Generative Adversarial Network
Chenyang Tao · Liqun Chen · Ricardo Henao · Jianfeng Feng · Lawrence Carin
To assess the difference between real and synthetic data, Generative Adversarial Networks (GANs) are trained using a distribution discrepancy measure. Three widely employed measures are information-theoretic divergences, integral probability metrics, and Hilbert space discrepancy metrics. We elucidate the theoretical connections between these three popular GAN training criteria and propose a novel procedure, called $\chi^2$ (Chi-square) GAN, that is conceptually simple, stable at training and resistant to mode collapse. Our procedure naturally generalizes to address the problem of simultaneous matching of multiple distributions. Further, we propose a resampling strategy that significantly improves sample quality, by repurposing the trained critic function via an importance weighting mechanism. Experiments show that the proposed procedure improves stability and convergence, and yields state-of-art results on a wide range of generative modeling tasks.
Author Information
Chenyang Tao (Duke University)
Liqun Chen (Duke University)
Ricardo Henao (Duke University)
Jianfeng Feng (Fudan University)
Lawrence Carin (Duke)
Related Events (a corresponding poster, oral, or spotlight)
-
2018 Poster: Chi-square Generative Adversarial Network »
Wed. Jul 11th 04:15 -- 07:00 PM Room Hall B #113
More from the Same Authors
-
2021 : Hölder Bounds for Sensitivity Analysis in Causal Reasoning »
Serge Assaad · Shuxi Zeng · Henry Pfister · Fan Li · Lawrence Carin -
2023 Poster: An Effective Meaningful Way to Evaluate Survival Models »
Shi-ang Qi · Neeraj Kumar · Mahtab Farrokh · Weijie Sun · Li-Hao Kuan · Rajesh Ranganath · Ricardo Henao · Russell Greiner -
2020 Poster: Learning Autoencoders with Relational Regularization »
Hongteng Xu · Dixin Luo · Ricardo Henao · Svati Shah · Lawrence Carin -
2020 Poster: Graph Optimal Transport for Cross-Domain Alignment »
Liqun Chen · Zhe Gan · Yu Cheng · Linjie Li · Lawrence Carin · Jingjing Liu -
2020 Poster: On Leveraging Pretrained GANs for Generation with Limited Data »
Miaoyun Zhao · Yulai Cong · Lawrence Carin -
2020 Poster: CLUB: A Contrastive Log-ratio Upper Bound of Mutual Information »
Pengyu Cheng · Weituo Hao · Shuyang Dai · Jiachang Liu · Zhe Gan · Lawrence Carin -
2019 Poster: Gromov-Wasserstein Learning for Graph Matching and Node Embedding »
Hongteng Xu · Dixin Luo · Hongyuan Zha · Lawrence Carin -
2019 Oral: Gromov-Wasserstein Learning for Graph Matching and Node Embedding »
Hongteng Xu · Dixin Luo · Hongyuan Zha · Lawrence Carin -
2019 Poster: Stochastic Blockmodels meet Graph Neural Networks »
Nikhil Mehta · Lawrence Carin · Piyush Rai -
2019 Poster: Variational Annealing of GANs: A Langevin Perspective »
Chenyang Tao · Shuyang Dai · Liqun Chen · Ke Bai · Junya Chen · Chang Liu · RUIYI (ROY) ZHANG · Georgiy Bobashev · Lawrence Carin -
2019 Oral: Stochastic Blockmodels meet Graph Neural Networks »
Nikhil Mehta · Lawrence Carin · Piyush Rai -
2019 Oral: Variational Annealing of GANs: A Langevin Perspective »
Chenyang Tao · Shuyang Dai · Liqun Chen · Ke Bai · Junya Chen · Chang Liu · RUIYI (ROY) ZHANG · Georgiy Bobashev · Lawrence Carin -
2018 Poster: Learning Registered Point Processes from Idiosyncratic Observations »
Hongteng Xu · Lawrence Carin · Hongyuan Zha -
2018 Poster: Policy Optimization as Wasserstein Gradient Flows »
RUIYI (ROY) ZHANG · Changyou Chen · Chunyuan Li · Lawrence Carin -
2018 Poster: JointGAN: Multi-Domain Joint Distribution Learning with Generative Adversarial Nets »
Yunchen Pu · Shuyang Dai · Zhe Gan · Weiyao Wang · Guoyin Wang · Yizhe Zhang · Ricardo Henao · Lawrence Carin -
2018 Oral: Policy Optimization as Wasserstein Gradient Flows »
RUIYI (ROY) ZHANG · Changyou Chen · Chunyuan Li · Lawrence Carin -
2018 Oral: JointGAN: Multi-Domain Joint Distribution Learning with Generative Adversarial Nets »
Yunchen Pu · Shuyang Dai · Zhe Gan · Weiyao Wang · Guoyin Wang · Yizhe Zhang · Ricardo Henao · Lawrence Carin -
2018 Oral: Learning Registered Point Processes from Idiosyncratic Observations »
Hongteng Xu · Lawrence Carin · Hongyuan Zha -
2018 Poster: Adversarial Time-to-Event Modeling »
Paidamoyo Chapfuwa · Chenyang Tao · Chunyuan Li · Courtney Page · Benjamin Goldstein · Lawrence Carin · Ricardo Henao -
2018 Oral: Adversarial Time-to-Event Modeling »
Paidamoyo Chapfuwa · Chenyang Tao · Chunyuan Li · Courtney Page · Benjamin Goldstein · Lawrence Carin · Ricardo Henao -
2018 Poster: Continuous-Time Flows for Efficient Inference and Density Estimation »
Changyou Chen · Chunyuan Li · Liquan Chen · Wenlin Wang · Yunchen Pu · Lawrence Carin -
2018 Poster: Variational Inference and Model Selection with Generalized Evidence Bounds »
Liqun Chen · Chenyang Tao · RUIYI (ROY) ZHANG · Ricardo Henao · Lawrence Carin -
2018 Oral: Continuous-Time Flows for Efficient Inference and Density Estimation »
Changyou Chen · Chunyuan Li · Liquan Chen · Wenlin Wang · Yunchen Pu · Lawrence Carin -
2018 Oral: Variational Inference and Model Selection with Generalized Evidence Bounds »
Liqun Chen · Chenyang Tao · RUIYI (ROY) ZHANG · Ricardo Henao · Lawrence Carin -
2017 Poster: Stochastic Gradient Monomial Gamma Sampler »
Yizhe Zhang · Changyou Chen · Zhe Gan · Ricardo Henao · Lawrence Carin -
2017 Poster: Adversarial Feature Matching for Text Generation »
Yizhe Zhang · Zhe Gan · Kai Fan · Zhi Chen · Ricardo Henao · Dinghan Shen · Lawrence Carin -
2017 Talk: Adversarial Feature Matching for Text Generation »
Yizhe Zhang · Zhe Gan · Kai Fan · Zhi Chen · Ricardo Henao · Dinghan Shen · Lawrence Carin -
2017 Talk: Stochastic Gradient Monomial Gamma Sampler »
Yizhe Zhang · Changyou Chen · Zhe Gan · Ricardo Henao · Lawrence Carin -
2017 Poster: Deep Generative Models for Relational Data with Side Information »
Changwei Hu · Piyush Rai · Lawrence Carin -
2017 Talk: Deep Generative Models for Relational Data with Side Information »
Changwei Hu · Piyush Rai · Lawrence Carin