• HomePage

  • The FMJH supports

  • Research

  • PGMO


    PGMO Courses 2023

    Gabriele Steidl

    (Technische Universität Berlin)

    February tuesday 7th and thursday 9th, 2023

    "Splitting Algorithms and Generalized Normalizing Flows"

    Variational methods are a quite universal and flexible approach for solving inverse problems, in particular in imaging sciences. Taking into account their specific structure as sum of several different terms, splitting algorithms provide a canonical tool for their efficient solution. Their strength consists in the splitting of the original problem into a sequence of smaller proximal problems which are easy and fast to compute.
    Operator splitting methods were first applied to linear, single-valued operators for solving partial differential equations in the 60th of the last century. More than 20 years later these methods were generalized in the convex analysis community to the solution of inclusion problems and again  more than 20 years they became popular in image processing and machine learning. Nowadays they are accomplished by so-called Plug-and-Play techniques, where a proximal denoising step is substituted by another denoiser. Popular denoisers were BM3D or MMSE methods which are based on (nonlocal) image patches.
    Meanwhile certain learned neural network do a better job. However, convergence of the PnP splitting algorithms is still an issue. Normalizing flows are special generative neural networks which are invertible.
    We demonstrate how they can be used as regularizers in inverse problems for learning from few images by using e.g. image patches. Unfortunately, normalizing flows suffer from a limited expressivity. This can be improved by applying generalized normalizing flows consisting of a forward and a backward Markov chain. Such Markov chains may in particular contain Langevin layers.

    We will also consider Wasserstein-2 spaces and Wasserstein gradient flows, where the above Langevin flow appears as a special instance. We will discuss recent developments for estimating Wasserstein gradient flows by neural networks.

    More informations



    PGMO Courses History

      PGMO Courses 2022

      published the March 7, 2022
      PGMO Courses 2022 François Delarue (Université Côte d'Azur) "Mean field games" March 7-8, 2022 "Mean field games" The objective of this course is to give an overview of the theory of mean field games, initiated by J.M. Lasry et P.L. …

      PGMO Courses 2021

      published the March 8, 2022
      PGMO Course 2021 February 3rd and 10th, 2021 "DYNAMICS IN GAMES: ALGORITHMS AND LEARNING" Sylvain Sorin Game theory studies interactions between agents with specific aims, be they rational actors, genes, or computers. This course is intended to provide the main …

      PGMO Courses 2020

      published the February 6, 2022
      PGMO Course 2020 September 10-11, 2020 PGMO Lecture "Algorithmic game theory : from multi-agent optimization to online learning" Roberto Cominetti (Universidad Adolfo Ibáñez, Santiago) Biography and Panayotis Mertikopoulos (LIG, Grenoble) Biography A satellite event of the SMAI-MODE 2020 conference organized …

      PGMO Courses 2019

      published the February 5, 2022
      PGMO Course 2019 January 14-15, 2019 - " A VU-point of view of nonsmooth optimization " Claudia SAGASTIZABAL (Unicamp, Brazil) - Ensta Abstract Slide 1 Slide 2 PGMO Course 2019 March 20-21, 2019 - STOCHASTIC PROGRAMMING Alexander SHAPIRO (Georgia Tech) …

      PGMO Courses 2018

      published the February 4, 2022
      PGMO Course 2018 January 24-216, 2018 - "Interior Point Methods for Very Large Scale Optimization" Jacek GONDZIO (University of Edinburgh) Biography Abstract Lecture 1 Lecture 2 Lecture 3 Lecture 4 Lecture 5 PGMO Course 2018 March, 13-14, 2018 - "BANDIT …

      PGMO Courses 2017

      published the February 3, 2022
      PGMO Course 2017 January, 25-26, 2017 - "First Order Optimization Methods" Marc TEBOULLE (Tel Aviv University) Abstract Poster PGMO Course 2017 February 21st, 2017 March 7th, 2017 - "Discrete convexity and applications : from stable matchings to quantum groups" Gleb …

      PGMO Courses 2016

      published the February 2, 2022
      PGMO Course 2016 December 13, 2016 - "The technician routing problem with conventional and electric vehicles" Jose A. Montoya Abstract PGMO Course 2016 October 19, 2016 - "Smart Grids and Optimization: A Winning Combination" Miguel F. Anjos Abstract PGMO Course …

      Pgmo Courses 2013-2014

      published the February 1, 2022
      PGMO Course 2014 May 5th, 2014, Ensta - "Bi-level stochastic programming" Georg Pflug (University of Vienna) Quite often, the result of our decisions depend on the action of others. A typical example is the setting of contracts, where the price …