Statistical-physics-based reconstruction in compressed sensing


Krzakala, F. ; Mézard, M. ; Sausset, F. ; Sun, Y. F. ; Zdeborová, L.


Compressed sensing has triggered a major evolution in signal acquisition. It consists of sampling a sparse signal at low rate and later using computational power for the exact reconstruction of the signal, so that only the necessary information is measured. Current reconstruction techniques are limited, however, to acquisition rates larger than the true density of the signal. We design a new procedure that is able to reconstruct the signal exactly with a number of measurements that approaches the theoretical limit, i.e., the number of nonzero components of the signal, in the limit of large systems. The design is based on the joint use of three essential ingredients : a probabilistic approach to signal reconstruction, a message-passing algorithm adapted from belief propagation, and a careful design of the measurement matrix inspired by the theory of crystal nucleation. The performance of this new algorithm is analyzed by statistical-physics methods. The obtained improvement is confirmed by numerical studies of several cases.

Version originale

PDF - 1.9 Mo

Texte complet : PDF

Voir en ligne


Citez cet article comme :

  author = {{Krzakala}, F. and {M{\'e}zard}, M. and {Sausset}, F. and {Sun}, Y.~F. and
        {Zdeborov{\'a}}, L.},
   title = "{Statistical-Physics-Based Reconstruction in Compressed Sensing}",
 journal = {Physical Review X},
    year = 2012,
   month = apr,
  volume = 2,
  number = 2,
   pages = {021005},
     doi = {10.1103/PhysRevX.2.021005},

Haut de page

À lire aussi...

Un laboratoire sur puce pour produire des plaquettes sanguines

Les plaquettes sanguines sont des cellules indispensables à la coagulation du sang. L’augmentation croissante des besoins en plaquettes encourage (...) 

> Lire la suite...

Coupling spin to velocity : collective motion of Hamiltonian polar particles

Coupling spin to velocity: collective motion of Hamiltonian polar particles Sigbjørn Løland Bore, Michael Schindler, Khanh-Dang Nguyen Thu Lam, (...) 

> Lire la suite...