Ecole d'ingénieur et centre de recherche en Sciences du numérique

Convergence analysis of sparse Bayesian learning under approximate inference techniques

Kurisummoottil Thomas, Christo; Slock, Dirk TM

ASILOMAR 2019, Asilomar Conference on Signals, Systems, and Computers, 3-6 November 2019, Pacific Grove, CA, USA

Sparse Bayesian Learning (SBL) is an efficient and well-studied framework for sparse signal recovery. SBL relies on a parameterized prior on the sparse signal to be estimated. The prior is chosen (with estimated hyperparameters) such that it encourages sparsity in the representation of the signal. However, SBL doesn’t scale with problem dimensions due to the computational complexity associated with matrix inversion. To address this issue, there exists low complexity methods based on approximate Bayesian inference. Various state of the art approximate inference methods are based on variational Bayesian (VB) inference or message passing algorithms such as belief propagation (BP) or expectation propagation. Moreover, these approximate inference methods can be unified under the optimization of Bethe free energy with appropriate constraints. SBL allows to treat more general signal models by the use of hierarchical prior formulation which eventually becomes more sparsity inducing than e.g., Laplacian prior. In this paper, we study the convergence behaviour of the mean and variance of the unknown parameters in SBL under approximate Bayesian inference.

Bibtex

Titre:Convergence analysis of sparse Bayesian learning under approximate inference techniques
Type:Conférence
Langue:English
Ville:Pacific Grove
Pays:ÉTATS-UNIS
Date:
Département:Systèmes de Communication
Eurecom ref:6056
Copyright: Asilomar
Bibtex: @inproceedings{EURECOM+6056, year = {2019}, title = {{C}onvergence analysis of sparse {B}ayesian learning under approximate inference techniques}, author = {{K}urisummoottil {T}homas, {C}hristo and {S}lock, {D}irk {TM}}, booktitle = {{ASILOMAR} 2019, {A}silomar {C}onference on {S}ignals, {S}ystems, and {C}omputers, 3-6 {N}ovember 2019, {P}acific {G}rove, {CA}, {USA}}, address = {{P}acific {G}rove, {\'{E}}{TATS}-{UNIS}}, month = {11}, url = {http://www.eurecom.fr/publication/6056} }
Voir aussi: