Sequentially Adaptive Bayesian Learning research project (SABL)

The Sequentially Adaptive Bayesian Learning (SABL) project develops new algorithms for inference and optimisation that exploit massively parallel desktop computing. It produces publications, working papers and technical reports on theoretical and applied problems, as well as software providing a platform for applied statistics and econometrics. The project is part of the Australian Research Council (ARC) Centre of Excellence for Mathematical and Statistical Frontiers (opens an external site). The project is also supported by Australian Research Council Grant DP130103356, Massively parallel algorithms for Bayesian inference and decision making.

SABL algorithm and toolbox

The SABL algorithm is a generalisation of adaptive posterior simulators described in Durham and Geweke (2014) (PDF, 841kB). That work is motivated by the pleasingly parallel structure of sequential Monte Carlo algorithms in conjunction with the power of graphics processing unit (GPU) hardware and software that together provide inexpensive, massively parallel desktop scientific computing. The SABL algorithm builds on a substantial literature in particle filtering, as discussed in Durham and Geweke (2014).

The generalisations incorporated in the SABL toolbox include quite a few variants of the algorithm, and the toolbox readily accommodates the incorporation of more. The variants include the extension of sequential Monte Carlo to optimization problems producing algorithms that can also be viewed as extensions of simulated annealing algorithms; see Geweke and Frishknecht (2014) (PDF, 934kB) and references there.

The SABL toolbox augments core Matlab functions as do all Matlab toolboxes, for example the Matlab Statistics and Matlab Parallel Computing Toolboxes. More important, the SABL toolbox exploits the modular structure of the algorithm. Incorporating new variants of the algorithm amounts to providing Matlab (or C) code that respects a simple interface. The same is true of new models, which amount to code for prior densities and likelihood functions in the case of Bayesian inference and code for objective functions in the case of optimization. SABL is specifically designed to facilitate incorporation of new models by third parties. SABL is also designed as a vehicle for applied scientific work drawing on models already incorporated in SABL.

SABL source code is open. It is freely available and may be used subject to the terms of the BSD license of the Open Software Initiative that protects it. The terms of this license are provided with the software.

Project members

Dr Huaxin Xu
Postdoctoral Research Fellow, School of Mathematical and Physical Sciences

Bin Peng
Postdoctoral Research Fellow, School of Mathematical and Physical Sciences

Simon Yin
Scientific Programmer, Economics Discipline Group

The project also involves collaborators in other countries, including Garland Durham (US), Bart Frischknecht (US), Mike Keane (UK), and Bill McCausland (Canada).

Acknowledgement of Country

UTS acknowledges the Gadigal people of the Eora Nation, the Boorooberongal people of the Dharug Nation, the Bidiagal people and the Gamaygal people, upon whose ancestral lands our university stands. We would also like to pay respect to the Elders both past and present, acknowledging them as the traditional custodians of knowledge for these lands.