1: \begin{abstract}

2: 	\noindent

3: 	The computation of Bayesian estimates of

4: 	system parameters and functions of them on the basis of observed

5: 	system performance data is a common problem within system identification.

6: 	This is a previously studied issue where

7: 	stochastic simulation approaches have been examined using the

8: 	popular Metropolis--Hastings (MH) algorithm.  This prior study has

9: 	identified a recognised difficulty of tuning the \emph{proposal

10: 	distribution} so that the MH method provides realisations with

11: 	sufficient mixing to deliver efficient convergence.  This paper

12: 	proposes and empirically examines a method of tuning the proposal

13: 	using ideas borrowed from the numerical optimisation literature

14: 	around efficient computation of Hessians so that gradient and

15: 	curvature information of the target posterior can be incorporated in

16: 	the proposal. \\

17: 

18: 	\noindent \textbf{Keywords}:

19: 	Bayesian parameter inference, state-space models, quasi-Newton, BFGS.

20: \end{abstract}

21: