Since their inception, Variational Autoencoders (VAEs) have become central in machine learning. Despite their widespread use, numerous questions regarding their theoretical properties remain open. Using PAC-Bayesian theory, this work develops statistical guarantees for VAEs. First, we derive the first PAC-Bayesian bound for posterior distributions conditioned on individual samples from the data-generating distribution. Then, we utilize this result to develop generalization guarantees for the VAE's reconstruction loss, as well as upper bounds on the distance between the input and the regenerated distributions. More importantly, we provide upper bounds on the Wasserstein distance between the input distribution and the distribution defined by the VAE's generative model.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Statistical Guarantees for Variational Autoencoders using PAC-Bayesian Theory


    Beteiligte:

    Erscheinungsdatum :

    2023


    Anmerkungen:

    Spotlight Paper at NeurIPS 2023


    Medientyp :

    Preprint


    Format :

    Elektronische Ressource


    Sprache :

    Englisch




    Theoretical Convergence Guarantees for Variational Autoencoders

    Surendran, Sobihan / Godichon-Baggioni, Antoine / Corff, Sylvain Le | ArXiv | 2024

    Freier Zugriff

    Deep Tracking Portfolios Using Autoencoders and Variational Autoencoders

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


    Variational Autoencoders

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



    Certifiably Robust Variational Autoencoders

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

    Freier Zugriff