Timezone: »
Many generative models have to combat missing modes. The conventional wisdom to this end is by reducing through training a statistical distance (such as f -divergence) between the generated distribution and provided data distribution. But this is more of a heuristic than a guarantee. The statistical distance measures a global, but not local, similarity between two distributions. Even if it is small, it does not imply a plausible mode coverage. Rethinking this problem from a game-theoretic perspective, we show that a complete mode coverage is firmly attainable. If a generative model can approximate a data distribution moderately well under a global statistical distance measure, then we will be able to find a mixture of generators that collectively covers every data point and thus every mode, with a lower-bounded generation probability. Constructing the generator mixture has a connection to the multiplicative weights update rule, upon which we propose our algorithm. We prove that our algorithm guarantees complete mode coverage. And our experiments on real and synthetic datasets confirm better mode coverage over recent approaches, ones that also use generator mixtures but rely on global statistical distances.
Author Information
Peilin Zhong (Columbia University)
Yuchen Mo (Columbia University)
Chang Xiao (Columbia University)
Pengyu Chen (Columbia University)
Changxi Zheng (Columbia University)
More from the Same Authors
-
2020 Poster: Listening to Sounds of Silence for Speech Denoising »
Ruilin Xu · Rundi Wu · Yuko Ishiwaka · Carl Vondrick · Changxi Zheng -
2020 Poster: Planning with General Objective Functions: Going Beyond Total Rewards »
Ruosong Wang · Peilin Zhong · Simon Du · Russ Salakhutdinov · Lin Yang -
2019 Poster: Average Case Column Subset Selection for Entrywise $\ell_1$-Norm Loss »
Zhao Song · David Woodruff · Peilin Zhong -
2019 Poster: Efficient Symmetric Norm Regression via Linear Sketching »
Zhao Song · Ruosong Wang · Lin Yang · Hongyang Zhang · Peilin Zhong -
2019 Poster: Towards a Zero-One Law for Column Subset Selection »
Zhao Song · David Woodruff · Peilin Zhong -
2018 Poster: BourGAN: Generative Networks with Metric Embeddings »
Chang Xiao · Peilin Zhong · Changxi Zheng -
2018 Spotlight: BourGAN: Generative Networks with Metric Embeddings »
Chang Xiao · Peilin Zhong · Changxi Zheng