Low complexity static and dynamic sparse bayesian learning combining BP, VB and EP message passing

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) provides sophisticated (state) model order selection with unknown support distribution. This allows to handle problems with big state dimensions and relatively limited data by exploiting variations in parameter importance. The
techniques proposed in this paper allow to handle the extension of SBL to time-varying states, modeled as diagonal first-order auto-regressive (DAR(1)) processes with unknown parameters to be estimated also. Adding the parameters to the state leads to an augmented state and a non-linear (at least bilinear) state-space model. The proposed approach, which applies also to more general non-linear models, uses a combination of belief propagation
(BP), Variational Bayes (VB) or mean field (MF) techniques, and Expectation Propagation (EP) to approximate the posterior marginal distributions of the scalar factors. We propose Fisher Information Matrix analysis to determine the variable split between the use of BP and VB allowing to stay optimal in terms of Laplace approximation.

Pacific Grove
Systèmes de Communication
Eurecom Ref:
© 2019 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.

PERMALINK : https://www.eurecom.fr/publication/6055