Department of Information Technology


This page contains a few code packages for illustrating some of the algorithms that I have been working on. See also my github page:

Particle Markov chain Monte Carlo

This small code package implements two Particle Markov chain Monte Carlo (PMCMC) methods for Bayesian parameter inference. The two algorithms are particle Gibbs with ancestor sampling (PGAS) and particle marginal Metropolis-Hastings (PMMH). The intention is to illustrate the algorithm on a simple example. The MATLAB code can be accessed here, [code].

Stochastic approximation EM using conditional particle filters

This code package implements the PSAEM algorithm (particle stochastic approximation EM) for maximum likelihood identification of a nonlinear system (the same system as used in the Bayesian setting in the PMCMC illustration; see above). The algorihtm is described in this paper. As comparison, a particle smoothing EM (PSEM) algorithm is also implemented. The main advantage of PSAEM over PSEM is that it makes more efficient use of the simulated particles, reducing the total computational cost (by orders of magnitude). The MATLAB code can be accessed here, [code].

Data driven Wiener system identification

We have developed a method for identification of Wiener systems based on a semiparametric Bayesian model. The related paper is available here. MATLAB code can be accessed here, [code].

Updated  2017-10-20 13:14:09 by Fredrik Lindsten.