Speaker
Peter Maass, University of Bremen
Time
2021-06-29 15:00 ~ 16:00, in 2 days
Venue
VooV Meeting Online
VooV Meeting Info

    VooV Meeting download website: https://voovmeeting.com/index.html
    Conference ID: 874 299 937
    Password: 210625
    Conference Link: https://meeting.tencent.com/s/gT394dDXm1Ke

Abstract

We start with a basic introduction on deep learning approaches to inverse problems. We then focus on the learned ISTA concept and describe it as a method for learning a data dependent optimized Tikhonov functional.

The main part of the talk is on learning with few data. In particular we investigate deep prior networks for solving inverse problems. Using the LISTA architecture in a deep prior network allows to proof equivalences to classical regularization schemes.

On the experimental side we focus on low dose CT reconstructions. We present a standardized data set and perform a numerical comparison of different deep learning concepts. The comparison is in terms of accuracy but also in terms of the amount of test data needed for training.

This is a joint work with Johannes Leuschner, Maximilian Schmidt, Soren Dittmer, Daniel Otero Baguer.
Bio

Peter Maass is professor for Applied Mathematics since 1999 and was the director of the Center for Industrial Mathematics at University of Bremen from 2009 to 2020. His main research areas include inverse problems, machine learning, parameter identification, wavelet analysis and since several years deep learning based on neural networks. Prof. Maass studied mathematics in Karlsruhe, Cambridge and Heidelberg and obtained his doctorate in 1988 from TU Berlin and his habilitation in mathematics from Saarland University in 1993. Peter Maass was a full professor for Numerical Analysis at University of Potsdam from 1993 -1999.

Peter Maass was awarded an honorary doctorate by the University of Saarland, Germany in 2018. Prof. Maass is taking a leading role in numerous research projects, e.g. he is the speaker of the DFG-funded Research Training Group 2224 ‘Parameter Identification – Analysis, Algorithms, Applications’ and was the deputy speaker of the DFG-SFB 747 ‘Micro Cold Forming’. Peter Maass is a member of the Advisory Board of the Interdisciplinary Center for Scientific Computing (IWR), Heidelberg and Member of the Executive Board of EU-Maths-In. Prof. Maass is (co)-author of more than 100 publications in peer-reviewed literature, three monographs and 27 book chapters. Peter Maass holds seven patents and patent applications.
--
FROM 202.121.181.*