From Statistical physics to computer science and back...
Softwares from the WASP
(Widely Applied Statistical Physics)
This page lists the links to the different projects I am curently involved in the application of statisical physics ideas and methods for interdisciplinary purpose. They contains codes, papers and data on the algoritmhs we have been developing for problems such as constraint satisfaction and optimization, statistical inference, machine learning, modules detection in network and compressed sensing.
Applying Statistical Physics to Inference in Compressed Sensing.
The SWept Approximate Message Passing approach to generalized linear inverse sparse and structured problem. If your stucture is complex, you might want to try out our Neural Network version of the prior using a restricted Boltzmann machine.
MOdule DEtection in NETworks with belief propagation.
This code, written by Jean Barbier, contains in particular the AMP algorithm for error correction over the AWGN channel using Sparse Superposition Codes
This code, written by Marylou Gabrie and Eric Tramel, shows how one can use the TAP free energy in order to learn Restricted Boltzmann Machine neural networks that are generative models of the MNIST database.
This code, written by Alaa Saade in the Julia language, use a new spectral method for Matrix Completion. A matlab version is also provided.
This matlab code follows our work with Zdeborova and Lesieur.
Professor Florent Krzakala University Pierre et Marie Curie
Lab. de Physique Statistique
ENS and UMR CNRS 8550
Tel: +33 (0)1 44 32 25 82
Fax: +33 (0)1 40 79 47 31
Ecole Normale Superieure
24 Rue Lhomond
75231 PARIS Cedex 05
To come and see me
ENS is in the center of Paris, close to the Pantheon. My office is L284 in the Physics Departement (second floor).
University Pierre et Marie Curie