Deep generative models have achieved impressive success in recent years. Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), as emerging families for generative model learning, have largely been considered as two distinct paradigms and received extensive independent studies respectively. This paper aims to establish formal connections between GANs and VAEs through a new formulation of them. We interpret sample generation in GANs as performing posterior inference, and show that GANs and VAEs involve minimizing KL divergences of respective posterior and inference distributions with opposite directions, extending the two learning phases of classic wake-sleep algorithm, respectively. The unified view provides a powerful tool to analyze a diverse set of existing model variants, and enables to transfer techniques across research lines in a principled way. For example, we apply the importance weighting method in VAE literatures for improved GAN learning, and enhance VAEs with an adversarial mechanism that leverages generated samples. Experiments show generality and effectiveness of the transferred techniques.


    Access

    Download


    Export, share and cite



    Title :

    On Unifying Deep Generative Models


    Contributors:

    Publication date :

    2017


    Remarks:

    Polished and extended content over the ICLR conference version: https://openreview.net/pdf?id=rylSzl-R-


    Type of media :

    Preprint


    Type of material :

    Electronic Resource


    Language :

    English




    Winning Lottery Tickets in Deep Generative Models

    Kalibhat, Neha Mukund / Balaji, Yogesh / Feizi, Soheil | ArXiv | 2020

    Free access

    Deep Generative Design Models for Improved Door Frame Performance

    Puthuvayil, Nirmal / Zaman, Thoheer / K, Arunkumar et al. | British Library Conference Proceedings | 2021


    Deep Generative Design Models for Improved Door Frame Performance

    K, Arunkumar / Puthuvayil, Nirmal / Cheyadri, Ramesh et al. | SAE Technical Papers | 2021


    Unifying HPC

    Bales, Bryan | NTRS | 2017


    Unifying the Validation of Ambient Solar Wind Models

    Martin A Reiss / Karin Muglach / Richard Mullinix et al. | NTRS | 2022