Graphic STI
logo EPFL
text EPFL
english only
Biomedical Imaging Group
Seminars
EPFL > BIG > SeminarsLogin
CONTENTS

Home page

News & Events

Seminars

People

Research

Publications

Tutorials & Reviews

Recent Talks

Demos

Download Algorithms

Jobs and Trainees

Teaching

Student Projects

Intranet

Order by     

Steer&Detect on Images 14 Nov 2017

Julien Fageot
EPFL STI LIB

Fundamental computational barriers in inverse problems and the mathematics of information27 Oct 2017

Alexander Bastounis
Cambridge University

Two of the most influential recent developments in applied mathematics are neural networks and compressed sensing. Compressed sensing (e.g. via basis pursuit or lasso) has seen considerable success at solving inverse problems and neural networks are rapidly becoming commonplace in everyday life with use cases ranging from self driving cars to automated music production. The observed success of these approaches would suggest that solving the underlying mathematical model on a computer is both well understood and computationally efficient. We will demonstrate that this is not the case. Instead, we show the following paradox: it is impossible to design algorithms that solve these problems to one significant figure when given inaccurate input data, even when the inaccuracies can be made arbitrarily small. This will occur even when the input data is in many senses well conditioned and shows that every existing algorithm will fail on some simple inputs. Further analysis of the situation for neural networks leads to the following additional ‘paradoxes of deep learning’: (1) One cannot guarantee the existence of algorithms for accurately training the neural network, and (2) one can have 100% success rate on arbitrarily many test cases, yet uncountably many misclassifications on elements that are arbitrarily close to the training set. Explaining the apparent contradiction of the observed success when applying compressed sensing, lasso and neural networks to real world examples given the aforementioned non existence result will require the development of new mathematical ideas and tools. We shall explain some of these ideas and give further information on all of the above paradoxes during the talk.

Variational use of B-splines and Kernel Based Functions27 Oct 2017

Christophe Rabut
INSA Toulouse

Deep learning based data manifold projection - a new regularization for inverse problems17 Oct 2017

Harshit Gupta
EPFL STI LIB

GlobalBioIm Lib - v2: new tools, more flexibility, and improved composition rules.03 Oct 2017

Emmanuel Soubies
EPFL STI LIB

Exact Discretization of Continuous-Domain Linear Inverse Problems with Generalized TV Regularization Using B-Splines​24 Aug 2017

Thomas Debarre
EPFL STI LIB

Fractional Integral transforms and Time-Frequency Representations02 Jun 2017

Prof. Ahmed I. Zayed
Department of Mathematical Sciences DePaul University

First steps toward fast PET reconstruction30 May 2017

Mike McCann
EPFL STI LIB

Lipid membranes and surface reconstruction - a biologically inspired method for 3D segmentation16 May 2017

Nicolas Chiaruttini
University of Geneva

Optical Diffraction Tomography: Principles and Algorithms09 May 2017

Thanh-an Pham
EPFL STI LIB

Compressed Sensing for Dose Reduction in STEM Tomography11 Apr 2017

Laurène Donati
EPFL STI LIB

Chasing Mycobacteria10 Apr 2017

Virginie Uhlmann
EPFL STI LIB

Multifractal analysis for signal and image classification23 Mar 2017

Stéphane Jaffard
UPEC

3D SIM and measurements time-reallocation for scanning based systems: Introduction and preliminary results on these two problems14 Mar 2017

Emmanuel Soubies
EPFL STI LIB

Inverse problems and multimodality for biological imaging28 Feb 2017

Denis Fortun
EPFL STI LIB

RKHS to find the Representer Theorem for regularization operators whose null space is not finite dimensional09 Feb 2017

Harshit Gupta
EPFL STI LIB

A unified reconstruction framework for coherent imaging24 Jan 2017

Ferréol Soulez
EPFL STI LIB

BPConvNet for compressed sensing recovery in bioimaging10 Jan 2017

Kyong Jin
EPFL STI LIB

Steerable template detection based on maximum correlation: preliminary results13 Dec 2016

Adrien Depeursinge
EPFL STI LIB

Opportunities in Computational Imaging for Biomicroscopy06 Dec 2016

Prof. Michael Leibling
Idap Research Institute

A multiple scattering approach to diffraction tomography30 Nov 2016

Luc Zeng
EPFL STI LIB

Learning Optimal Shrinkage Splines for ADMM Algorithms22 Nov 2016

Ha Nguyen
EPFL STI LIB

SIGGRAPH ASIA 201601 Nov 2016

Daniel Schmitter
EPFL STI LIB

Lévy's Persian summers18 Oct 2016

Julien Fageot
EPFL STI LIB

High-quality parallel-ray X-ray CT back projection using optimized interpolation11 Oct 2016

Mike McCann
EPFL STI LIB

Algorithmic Aspects of Compressive Sensing03 Oct 2016

Verner Vlacic
Cambridge University

ICIP 201620 Sep 2016

Anaïs Badoual
EPFL STI LIB

Machine Vision forum in Heidelberg17 Aug 2016

Virginie Uhlmann
EPFL STI LIB

K-space interpolation using CV(complex valued)-CNN & sparse and low-rank model of ALOHA09 Aug 2016

Kyong Jin
EPFL STI LIB

Complete Compressed Sensing Framework for STEM Tomography19 Jul 2016

Laurène Donati
EPFL STI LIB

Trainable shrinkage splines: inverse problems meet deep learning28 Jun 2016

Ha Nguyen
EPFL STI LIB

A reconstruction framework for coherent imaging31 May 2016

Ferréol Soulez
EPFL STI LIB

© 2010 EPFL • webmaster.big@epfl.ch • 26.01.2010