Flow-GAN: Combining Maximum Likelihood and Adversarial Learning in Generative Models
This addresses the problem of evaluating and improving generative models for researchers and practitioners, offering a hybrid approach that is incremental over existing GAN and maximum likelihood methods.
The paper tackled the challenge of combining adversarial and maximum likelihood training in generative models, proposing Flow-GANs to enable exact likelihood evaluation and hybrid training, resulting in high held-out likelihoods and visual fidelity on MNIST and CIFAR-10 datasets.
Adversarial learning of probabilistic models has recently emerged as a promising alternative to maximum likelihood. Implicit models such as generative adversarial networks (GAN) often generate better samples compared to explicit models trained by maximum likelihood. Yet, GANs sidestep the characterization of an explicit density which makes quantitative evaluations challenging. To bridge this gap, we propose Flow-GANs, a generative adversarial network for which we can perform exact likelihood evaluation, thus supporting both adversarial and maximum likelihood training. When trained adversarially, Flow-GANs generate high-quality samples but attain extremely poor log-likelihood scores, inferior even to a mixture model memorizing the training data; the opposite is true when trained by maximum likelihood. Results on MNIST and CIFAR-10 demonstrate that hybrid training can attain high held-out likelihoods while retaining visual fidelity in the generated samples.