Metropolis-Hastings algorithm
Metropolis-Hastings algorithm
Metropolis-Hastings algorithm
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
MCMC methods<br />
• Markov chain Monte Carlo methods:<br />
– Class of <strong>algorithm</strong>s for sampling from a probability distribution.<br />
– Based on constructing a Markov chain.<br />
– Desired distribution is the equilibrium distribution.<br />
• State of Markov chain after many steps then used as a sample of<br />
the desired distribution.<br />
– Quality of the sample improves as the number of steps increases.<br />
• Difficult problem is to determine how many steps are needed to<br />
converge to the stationary distribution within an acceptable<br />
error.<br />
– A good <strong>algorithm</strong> will have rapid mixing: the stationary<br />
distribution is reached quickly starting from an arbitrary position.<br />
– <strong>Metropolis</strong>-<strong>Hastings</strong> <strong>algorithm</strong> has rapid-mixing gproperties.<br />
p