Search

Events

VADOR / Events / SWM Colloquium : Optimizing my distributions and proving convergence, or how to look into the mirror, Pierre-Cyril Aubin, TU Wien
15. May 2024, 16:30 until 18:30
SWM Colloquium : Optimizing my distributions and proving convergence, or how to look into the mirror, Pierre-Cyril Aubin, TU Wien
Many problems in machine learning and applied statistics can be formulated as optimizing a functional over the space of probability measures, e.g. the Kullback–Leibler (KL) divergence. But can we guarantee that we have a converging algorithm? Starting from Expectation-Maximization (EM), I will show that it can always be written as a mirror descent and present two cases, 1) the joint distribution is an exponential family and 2) we have a non-parametric distribution, but only over the latent space. In these cases, EM only involves convex functions and we have a (sub)linear convergence rate. Moving to variational inference in disguise, namely entropic optimal transport, I will then focus on the convergence of Sinkhorn's algorithm, a.k.a IPFP or RAS, outlining the similarities with EM.

Finally, I will show that both these algorithms fall within a general majorize-minimize framework for which we prove novel rates of convergence based on a five-point property introduced by Csiszár and Tusnády (in 1984).

The talk is based on joint works with Anna Korba (ENSAE, France) and Flavien Léger (INRIA Paris), see https://arxiv.org/abs/2305.04917 , Sections 1, 4.7 and 4.8 for an overview.

About Cookies

Our website uses cookies to ensure you get the best experience on our website, for analytical purposes, to provide social media features, and for targeted advertising. This it is necessary in order to pass information on to respective service providers. If you would like additional information about cookies on this website, please see our Data Protection Declaration.

  • These cookies are required to help our website run smoothly.

    Name Purpose Lifetime Type Provider
    wordpress_test_cookie Testing-Cookie to check whether cookies are allowed. 1 Year HTTP Homepage TUW
    PHPSESSID Used by WordPress to retain the state of your current user session for all page requests. Session HTTP Homepage TUW
    wordpress_logged_in_{hash} Used by Wordpress to keep users logged in. {hash} represents an unique user token. 1 Year HTTP Homepage TUW
    wp-settings-time-{id} Used to customize your view of admin interface, and possibly also the main site interface. 1 Year HTTP Homepage TUW
    wordpress_sec_{hash} This cookie is used to store your authentication details. Its use is limited to the admin console area. {hash} represents an unique user token. 1 Year HTTP Homepage TUW
    wp-settings-{id} Used to customize your view of admin interface, and possibly also the main site interface. 1 Year HTTP Homepage TUW
    wp-wpml_current_language Stores the current language. This cookie is enabled by default on sites that use the Language filtering for AJAX operations feature. 1 Day HTTP Homepage TUW
    wp-wpml_current_admin_language_{hash} Stores the current WordPress administration area language. {hash} represents an unique user token. 1 Day HTTP Homepage TUW
    CookieConsent_b92907 Saves your settings for the use of cookies on this website. 1 Year HTML Homepage TUW

  • These cookies help us to continuously improve our services and adapt our website to your needs. We statistically evaluate the pseudonymized data collected from our website.

    Name Purpose Lifetime Type Provider
    _pk_id.136.56ce Used to store a few details about the user such as the unique visitor ID. 13 months HTML Matomo TUW
    _pk_ref Is used to store the information of the users home website. 6 months HTML Matomo TUW
    _pk_ses.136.56ce Is needed to store temporary data of the visit. 30 minutes HTML Matomo TUW