Data Learning on Manifolds and Future Challenges

Informazioni relative al progetto

DEDALE

ID dell’accordo di sovvenzione: 665044

DOI 10.3030/665044

Progetto chiuso

Data della firma CE 26 Maggio 2015

Data di avvio 1 Ottobre 2015

Data di completamento 30 Settembre 2018

Finanziato da

• EXCELLENT SCIENCE - Future and Emerging Technologies (FET)

Costo totale € 2 702 397,50

Contributo UE € 2 702 397,50

Coordinato da

COMMISSARIAT A L ENERGIE ATOMIQUE ET AUX ENERGIES ALTERNATIVES
France

CORDIS fornisce collegamenti ai risultati finali pubblici e alle pubblicazioni dei progetti ORIZZONTE.

I link ai risultati e alle pubblicazioni dei progetti del 7° PQ, così come i link ad alcuni tipi di risultati specifici come dataset e software, sono recuperati dinamicamente da .OpenAIRE .

Risultati finali

Documents, reports (9)

Super-resolution and interpolation of the Euclid PSF

Report on applying dictionary learning on manifolds developed in DEDALE in combination with interpolation methods which are used extensively by the UCL group (e.g neural networks and Gaussian processes) to build a model for the Euclid PSF.

Optimization for manifold-valued signal restoration.

Report on optimization methods for signal restoration with manifold-valued representations: designing and implementing algorithms to solve linear inverse problems with non-linear signal representations; The case of non-linear and potentially non-convex problems will be discussed.

Adaptive transforms for manifold-valued data.

Report on the development of adapted multiscale transforms. The final learned dictionary is restricted to a class of dictionaries generated from a structured dictionary such as shearlet. Existence of fast transform/reconstruction will be discussed.

Non-linear learning on complex imaging data.

Report on Non-linear learning on complex imaging data: learning representations for data lying on unknown low-dimensional manifolds. The use of deep learning architectures like stacked sparse autoencoders will be particularly studied in this task.

Large-scale learning schemes.

Evaluation of cutting edge distributed processing platforms, such as GraphLab, Mahout and MLI, for benchmarking large-scale test sets for machine learning. Real time parallel processing considerations will be actively taken under account into this task.

Optimizations for non-linear learning.

Report on building upon proximal methods and problem splitting techniques to design highly-parallelizable sparse solvers (e.g. sparse/low-rank multivariate signal decompositions.)

Learning-based photometric and spectroscopic redshift estimation

Report on the development of a dictionary learning based method for spectroscopic and photometric redshift estimation of Euclid data.

Dictionary learning for multivariate/multispectral data.

Report on Dictionary Learning on multi-valued data. The case of multi-channel polarized data on the sphere will be considered.

Linear inverse problems with sparsity constraints.

Report on the development of dedicated solvers for the recovery of multivariate signals with adapted sparse priors, either in fixed representations from Task 2.1 or learnt representations from task 2.2. Convergence and Computation time will be discussed.

Demonstrators, pilots, prototypes (2)

Numerical toolbox and benchmarking platform.

Development of a numerical toolbox and benchmark test set. The code will be in Python and/orC++. The algorithms related to tasks 3.1 and 3.2 will be included in the toolbox, and benchmark data will also be joined to the toolbox for test and evaluation.

Toolbox and benchmarking platform for large scale learning.

Toolbox for parallel linear and non-linear sparsity based learning architectures.

Open Research Data Pilot (1)

Evaluation/validation of the mass mapping algorithms

Report on using dictionary methods for 2D and 3D mass mapping from weak lensing data. Data available in the framework of the Open Research Data Pilot concept.

Websites, patent fillings, videos etc. (1)

Project Website & Factsheet

Realization of web site, contains informations about the project (publications, technical notes, etc).

Pubblicazioni

Peer reviewed articles (21)

Cosmic microwave background reconstruction from WMAP and Planck PR2 data

Autori: J. Bobin, F. Sureau, J.-L. Starck
Pubblicato in: Astronomy & Astrophysics, Numero 591, 2016, Pagina/e A50, ISSN 0004-6361
Editore: Springer Verlag
DOI: 10.1051/0004-6361/201527822

Sparsity and inverse problems in astrophysics

Autori: Jean-Luc Starck
Pubblicato in: Journal of Physics: Conference Series, Numero 699, 2016, Pagina/e 012010, ISSN 1742-6588
Editore: Institute of Physics
DOI: 10.1088/1742-6596/699/1/012010

High resolution weak lensing mass mapping combining shear and flexion

Autori: F. Lanusse, J.-L. Starck, A. Leonard, S. Pires
Pubblicato in: Astronomy & Astrophysics, Numero 591, 2016, Pagina/e A2, ISSN 0004-6361
Editore: Springer Verlag
DOI: 10.1051/0004-6361/201628278

Point Spread Function Field Learning Based on Optimal Transport Distances

Autori: F. Ngolé, J.-L. Starck
Pubblicato in: SIAM Journal on Imaging Sciences, Numero 10/3, 2017, Pagina/e 1549-1578, ISSN 1936-4954
Editore: Society for Industrial and Applied Mathematics
DOI: 10.1137/16M1093677

α -Molecules

Autori: Philipp Grohs, Sandra Keiper, Gitta Kutyniok, Martin Schäfer
Pubblicato in: Applied and Computational Harmonic Analysis, Numero 41/1, 2016, Pagina/e 297-336, ISSN 1063-5203
Editore: Academic Press
DOI: 10.1016/j.acha.2015.10.009

Sparse Reconstruction of the Merging A520 Cluster System

Autori: Austin Peel, François Lanusse, Jean-Luc Starck
Pubblicato in: The Astrophysical Journal, Numero 847/1, 2017, Pagina/e 23, ISSN 1538-4357
Editore: IOPScience
DOI: 10.3847/1538-4357/aa850d

Space variant deconvolution of galaxy survey images

Autori: S. Farrens, F. M. Ngolè Mboula, J.-L. Starck
Pubblicato in: Astronomy & Astrophysics, Numero 601, 2017, Pagina/e A66, ISSN 0004-6361
Editore: Springer Verlag
DOI: 10.1051/0004-6361/201629709

Singular spectrum-based matrix completion for time series recovery and prediction

Autori: Grigorios Tsagkatakis, Baltasar Beferull-Lozano, Panagiotis Tsakalides
Pubblicato in: EURASIP Journal on Advances in Signal Processing, Numero 2016/1, 2016, ISSN 1687-6180
Editore: springer
DOI: 10.1186/s13634-016-0360-0

Land Classification Using Remotely Sensed Data: Going Multilabel

Autori: Konstantinos Karalas, Grigorios Tsagkatakis, Michael Zervakis, Panagiotis Tsakalides
Pubblicato in: IEEE Transactions on Geoscience and Remote Sensing, Numero 54/6, 2016, Pagina/e 3548-3563, ISSN 0196-2892
Editore: Institute of Electrical and Electronics Engineers
DOI: 10.1109/TGRS.2016.2520203

Matrix and Tensor Completion on a Human Activity Recognition Framework

Autori: Sofia Savvaki, Grigorios Tsagkatakis, Athanasia Panousopoulou, Panagiotis Tsakalides
Pubblicato in: IEEE Journal of Biomedical and Health Informatics, Numero 21/6, 2017, Pagina/e 1554-1561, ISSN 2168-2194
Editore: Institute of Electrical and Electronics Engineers Inc.
DOI: 10.1109/JBHI.2017.2716112

Compressed sensing for finite-valued signals

Autori: Sandra Keiper, Gitta Kutyniok, Dae Gwan Lee, Götz E. Pfander
Pubblicato in: Linear Algebra and its Applications, Numero 532, 2017, Pagina/e 570-613, ISSN 0024-3795
Editore: Elsevier BV
DOI: 10.1016/j.laa.2017.07.006

Wasserstein Dictionary Learning: Optimal Transport-Based Unsupervised Nonlinear Dictionary Learning

Autori: Morgan A. Schmitz, Matthieu Heitz, Nicolas Bonneel, Fred Ngolè, David Coeurjolly, Marco Cuturi, Gabriel Peyré, Jean-Luc Starck
Pubblicato in: SIAM Journal on Imaging Sciences, Numero 11/1, 2018, Pagina/e 643-678, ISSN 1936-4954
Editore: Society for Industrial and Applied Mathematics
DOI: 10.1137/17M1140431

A Haar wavelet-based perceptual similarity index for image quality assessment

Autori: Rafael Reisenhofer, Sebastian Bosse, Gitta Kutyniok, Thomas Wiegand
Pubblicato in: Signal Processing: Image Communication, Numero 61, 2018, Pagina/e 33-43, ISSN 0923-5965
Editore: Elsevier BV
DOI: 10.1016/j.image.2017.11.001

Unsupervised feature-learning for galaxy SEDs with denoising autoencoders

Autori: J. Frontera-Pons, F. Sureau, J. Bobin, E. Le Floc’h
Pubblicato in: Astronomy & Astrophysics, Numero 603, 2017, Pagina/e A60, ISSN 0004-6361
Editore: Springer Verlag
DOI: 10.1051/0004-6361/201630240

Optimal approximation of piecewise smooth functions using deep ReLU neural networks

Autori: Philipp Petersen, Felix Voigtlaender
Pubblicato in: Neural Networks, Numero 108, 2018, Pagina/e 296-330, ISSN 0893-6080
Editore: Pergamon Press Ltd.
DOI: 10.1016/j.neunet.2018.08.019

Constraint matrix factorization for space variant PSFs field restoration

Autori: F Ngolè, J-L Starck, K Okumura, J Amiaux, P Hudelot
Pubblicato in: Inverse Problems, Numero 32/12, 2016, Pagina/e 124001, ISSN 0266-5611
Editore: Institute of Physics Publishing
DOI: 10.1088/0266-5611/32/12/124001

Breaking degeneracies in modified gravity with higher (than 2nd) order weak-lensing statistics

Autori: Austin Peel, Valeria Pettorino, Carlo Giocoli, Jean-Luc Starck, Marco Baldi
Pubblicato in: Astronomy & Astrophysics, Numero 619, 2018, Pagina/e A38, ISSN 0004-6361
Editore: Springer Verlag
DOI: 10.1051/0004-6361/201833481

Cosmological constraints with weak-lensing peak counts and second-order statistics in a large-field survey

Autori: Austin Peel, Chieh-An Lin, François Lanusse, Adrienne Leonard, Jean-Luc Starck, Martin Kilbinger
Pubblicato in: Astronomy & Astrophysics, Numero 599, 2017, Pagina/e A79, ISSN 0004-6361
Editore: Springer Verlag
DOI: 10.1051/0004-6361/201629928

Degradation analysis in the estimation of photometric redshifts from non-representative training sets

Autori: J D Rivera, B Moraes, A I Merson, S Jouvel, F B Abdalla, M C B Abdalla
Pubblicato in: Monthly Notices of the Royal Astronomical Society, Numero 477/4, 2018, Pagina/e 4330-4347, ISSN 0035-8711
Editore: Blackwell Publishing Inc.
DOI: 10.1093/mnras/sty880

Improving weak lensing mass map reconstructions using Gaussian and sparsity priors: application to DES SV

Autori: N Jeffrey, F B Abdalla, O Lahav, F Lanusse, J-L Starck, A Leonard, D Kirk, C Chang, E Baxter, T Kacprzak, S Seitz, V Vikram, L Whiteway, T M C Abbott, S Allam, S Avila, E Bertin, D Brooks, A Carnero Rosell, M Carrasco Kind, J Carretero, F J Castander, M Crocce, C E Cunha, C B D’Andrea, L N da Costa, C Davis, J De Vicente, S Desai, P Doel, T F Eifler, A E Evrard, B Flaugher, P Fosalba, J Frie
Pubblicato in: Monthly Notices of the Royal Astronomical Society, Numero 479/3, 2018, Pagina/e 2871-2888, ISSN 0035-8711
Editore: Blackwell Publishing Inc.
DOI: 10.1093/mnras/sty1252

Multi-band morpho-Spectral Component Analysis Deblending Tool (MuSCADeT): Deblending colourful objects

Autori: R. Joseph, F. Courbin, J.-L. Starck
Pubblicato in: Astronomy & Astrophysics, Numero 589, 2016, Pagina/e A2, ISSN 0004-6361
Editore: Springer Verlag
DOI: 10.1051/0004-6361/201527923

Other (15)

A Mathematical Framework for Feature Selection from Real-World Data with Non-Linear Observations

Autori: Genzel, Martin; Kutyniok, Gitta
Pubblicato in: Numero 25, 2016
Editore: ArXiv

Optimal Approximation with Sparsely Connected Deep Neural Networks

Autori: Helmut Bölcskei, Philipp Grohs, Gitta Kutyniok, Philipp Petersen
Pubblicato in: 2017
Editore: ArXiv

Robust 1-Bit Compressed Sensing via Hinge Loss Minimization

Autori: Martin Genzel, Alexander Stollenwerk
Pubblicato in: 2018
Editore: arXiv

Recovering Structured Data From Superimposed Non-Linear Measurements

Autori: Martin Genzel, Peter Jung
Pubblicato in: 2017
Editore: ArXiv

Compressed Sensing for Analog Signals

Autori: Bernard G. Bodmann, Axel Flinth, Gitta Kutyniok
Pubblicato in: 2018
Editore: arXiv

Anisotropic Multiscale Systems on Bounded Domains

Autori: Grohs, Philipp; Kutyniok, Gitta; Ma, Jackie; Petersen, Philipp; Raslan, Mones
Pubblicato in: Numero 33, 2015
Editore: ArXiv

ℓ1-Analysis Minimization and Generalized (Co-)Sparsity: When Does Recovery Succeed?

Autori: Martin Genzel, Gitta Kutyniok, Maximilian März
Pubblicato in: 2017
Editore: ArXiv

The Mismatch Principle: Statistical Learning Under Large Model Uncertainties

Autori: Martin Genzel, Gitta Kutyniok
Pubblicato in: 2018
Editore: arXiv

Structured, compactly supported Banach frame decompositions of decomposition spaces

Autori: Voigtlaender, Felix
Pubblicato in: Numero 21, 2016
Editore: ArXiv

Learning sparse representations on the sphere

Autori: Florent Sureau, Felix Voigtlaender, Malte Wust, Jean-Luc Starck, Gitta Kutyniok
Pubblicato in: 2018
Editore: arXiv

A Distributed Learning Architecture for Scientific Imaging Problems

Autori: A. Panousopoulou, S. Farrens, K. Fotiadou, A. Woiselle, G. Tsagkatakis, , J.-L. Starck, P. Tsakalides
Pubblicato in: 2018
Editore: arXiv

Parameter inference and model comparison using theoretical predictions from noisy simulations

Autori: Niall Jeffrey, Filipe B. Abdalla
Pubblicato in: 2018
Editore: arXiv

Convolutional Neural Networks for Spectroscopic Redshift Estimation on Euclid Data

Autori: Radamanthys Stivaktakis, Grigorios Tsagkatakis, Bruno Moraes, Filipe Abdalla, Jean-Luc Starck, Panagiotis Tsakalides
Pubblicato in: 2018
Editore: arXiv

ZXCorr: Cosmological Measurements from Angular Power Spectra Analysis of BOSS DR12 Tomography

Autori: Arthur Loureiro, Bruno Moraes, Filipe B. Abdalla, Andrei Cuceu, Michael McLeod, Lorne Whiteway, Sreekumar T. Balan, Aurélien Benoit-Lévy, Ofer Lahav, Marc Manera, Richard Rollins, Henrique S. Xavier
Pubblicato in: 2018
Editore: arXiv

Analysis vs. synthesis sparsity for α-shearlets

Autori: Felix Voigtlaender, Anne Pein
Pubblicato in: 2017
Editore: arXiv

Conference proceedings (1)

A distributed learning architecture for big imaging problems in astrophysics

Autori: Athanasia Panousopoulou, Sammuel Farrens, Yannis Mastorakis, Jean-Luck Starck, Panagiotis Tsakailides
Pubblicato in: 2017 25th European Signal Processing Conference (EUSIPCO), 2017, Pagina/e 1440-1444, ISBN 978-0-9928626-7-1
Editore: IEEE
DOI: 10.23919/eusipco.2017.8081447