Two Generator Game: Learning to Sample via Linear Goodness-of-Fit Test
Lizhong Ding · Mengyang Yu · Li Liu · Fan Zhu · Yong Liu · Yu Li · Ling Shao

Thu Dec 12th 05:00 -- 07:00 PM @ East Exhibition Hall B + C #11

Learning the probability distribution of high-dimensional data is a challenging problem. To solve this problem, we formulate a deep energy adversarial network (DEAN), which casts the energy model learned from real data into an optimization of a goodness-of-fit (GOF) test statistic. DEAN can be interpreted as a GOF game between two generative networks, where one explicit generative network learns an energy-based distribution that fits the real data, and the other implicit generative network is trained by minimizing a GOF test statistic between the energy-based distribution and the generated data, such that the underlying distribution of the generated data is close to the energy-based distribution. We design a two-level alternative optimization procedure to train the explicit and implicit generative networks, such that the hyper-parameters can also be automatically learned. Experimental results show that DEAN achieves high quality generations compared to the state-of-the-art approaches.

Author Information

Lizhong Ding (Inception Institute of Artificial Intelligence)
Mengyang Yu (Inception Institute of Artificial Intelligence)
Li Liu (Inception Institute of Artificial Intelligence)
Fan Zhu (Inception Institute of Artificial Intelligence)
Yong Liu (Institute of Information Engineering, CAS)
Yu Li (King Abdullah University of Science and Technology)
Ling Shao (Inception Institute of Artificial Intelligence)

More from the Same Authors