Among likelihood-based approaches for deep generative modelling, variational autoencoders (VAEs) offer scalable amortized posterior inference and fast sampling. However, VAEs are also more and more outperformed by competing models such as normalizing flows (NFs), deep-energy models, or the new denoising diffusion probabilistic models (DDPMs). In this preliminary work, we improve VAEs by demonstrating how DDPMs can be used for modelling the prior distribution of the latent variables. The diffusion prior model improves upon Gaussian priors of classical VAEs and is competitive with NF-based priors. Finally, we hypothesize that hierarchical VAEs could similarly benefit from the enhanced capacity of diffusion priors.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Diffusion Priors In Variational Autoencoders


    Beteiligte:

    Erscheinungsdatum :

    2021


    Medientyp :

    Preprint


    Format :

    Elektronische Ressource


    Sprache :

    Englisch




    Item Recommendation with Variational Autoencoders and Heterogenous Priors

    Karamanolakis, Giannis / Cherian, Kevin Raji / Narayan, Ananth Ravi et al. | ArXiv | 2018

    Freier Zugriff

    Variational Autoencoders

    Ghojogh, Benyamin / Crowley, Mark / Karray, Fakhri et al. | Springer Verlag | 2022


    Deep Tracking Portfolios Using Autoencoders and Variational Autoencoders

    Urrego, Daniel Aragón / Nieto, Oscar Eduardo Reyes / Quimbayo, Carlos Andrés Zapata | Springer Verlag | 2024


    Certifiably Robust Variational Autoencoders

    Barrett, Ben / Camuto, Alexander / Willetts, Matthew et al. | ArXiv | 2021

    Freier Zugriff

    Mixed-curvature Variational Autoencoders

    Skopek, Ondrej / Ganea, Octavian-Eugen / Bécigneul, Gary | ArXiv | 2019

    Freier Zugriff