AdaGAN: Boosting Generative Models

Generative Adversarial Networks (GAN) are an effective method for training generative models of complex data such as natural images. However, they are notoriously hard to train and can suffer from the problem of missing modes where the model is not able to produce examples in certain regions of the space. We propose an iterative procedure, called AdaGAN, where at every step we add a new component into a mixture model by running a GAN algorithm on a re-weighted sample. This is inspired by boosting algorithms, where many potentially weak individual predictors are greedily aggregated to form a strong composite predictor. We prove analytically that such an incremental procedure leads to convergence to the true distribution in a finite number of steps if each step is optimal, and convergence at an exponential rate otherwise. We also illustrate experimentally that this procedure addresses the problem of missing modes.

[1]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[2]  Yoav Freund,et al.  A decision-theoretic generalization of on-line learning and an application to boosting , 1995, EuroCOLT.

[3]  Andrew R. Barron,et al.  Mixture Density Estimation , 1999, NIPS.

[4]  Radford M. Neal.  Annealed importance sampling , 1998, Stat. Comput..

[5]  Geoffrey E. Hinton,et al.  Self Supervised Boosting , 2002, NIPS.

[6]  Saharon Rosset,et al.  Boosting Density Estimation , 2002, NIPS.

[7]  Flemming Topsøe,et al.  Jensen-Shannon divergence and Hilbert space embedding , 2004, International Symposium onInformation Theory, 2004. ISIT 2004. Proceedings..

[8]  Matthias Hein,et al.  Hilbertian Metrics and Positive Definite Kernels on Probability Measures , 2005, AISTATS.

[9]  Zhuowen Tu,et al.  Learning Generative Models via Discriminative Approaches , 2007, 2007 IEEE Conference on Computer Vision and Pattern Recognition.

[10]  Mark D. Reid,et al.  Information, Divergence and Risk for Binary Experiments , 2009, J. Mach. Learn. Res..

[11]  Yoshua Bengio,et al.  Generative Adversarial Nets , 2014, NIPS.

[12]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[13]  W. H. Pun,et al.  Statistical Decision Theory , 2014 .

[14]  Pieter Abbeel,et al.  InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets , 2016, NIPS.

[15]  Wojciech Zaremba,et al.  Improved Techniques for Training GANs , 2016, NIPS.

[16]  Soumith Chintala,et al.  Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks , 2015, ICLR.

[17]  Joost van de Weijer,et al.  Ensembles of Generative Adversarial Networks , 2016, ArXiv.

[18]  Sebastian Nowozin,et al.  f-GAN: Training Generative Neural Samplers using Variational Divergence Minimization , 2016, NIPS.

[19]  David Pfau,et al.  Unrolled Generative Adversarial Networks , 2016, ICLR.

[20]  Ruslan Salakhutdinov,et al.  On the Quantitative Analysis of Decoder-Based Generative Models , 2016, ICLR.

[21]  Yoshua Bengio,et al.  Mode Regularized Generative Adversarial Networks , 2016, ICLR.

[22]  Stefano Ermon,et al.  Boosted Generative Models , 2016, AAAI.