1: \begin{abstract}

2: Probabilistic modeling provides the capability to represent and

3: manipulate \emph{uncertainty} in data, models, predictions and

4: decisions. We are concerned with the problem of learning

5: probabilistic models of dynamical systems from measured

6: data. Specifically, we consider learning of probabilistic

7: nonlinear state-space models.  There is no closed-form solution

8: available for this problem, implying that we are forced to use

9: approximations.  In this tutorial we will provide a self-contained

10: introduction to one of the state-of-the-art methods---the particle

11: Metropolis--Hastings algorithm---which has proven to offer a practical approximation.

12: This is a Monte Carlo based method, where the particle

13: filter is used to guide a Markov chain Monte Carlo method through

14: the parameter space.  One of the key merits of the particle

15: Metropolis--Hastings algorithm is that it is guaranteed to converge to

16: the ``true solution'' under mild assumptions, despite being based

17: on a particle filter with only a finite number of particles. 

18: We will also provide a motivating numerical example illustrating the method using a modeling language tailored for sequential Monte Carlo methods.

19: The intention of modeling languages of this kind is to open up the power of sophisticated Monte Carlo methods---including particle Metropolis--Hastings---to a large group of users

20: without requiring them to know all the underlying mathematical

21: details.

22: \end{abstract}

23: