Difference between revisions of "ISC-5939-03"
Peter Beerli (talk | contribs) |
Peter Beerli (talk | contribs) |
||
| Line 10: | Line 10: | ||
<br> | <br> | ||
[http://peterbeerli.com/classdata/mcmc EXAMPLE CODES IN PYTHON]. Currently there are two files that give an example of the Metropolis method (using a symmetric proposal) and an example that showcases the Metropolis-Hastings method (using an asymmetric proposal that needs the Hastings ratio correction). | [http://peterbeerli.com/classdata/mcmc EXAMPLE CODES IN PYTHON]. Currently there are two files that give an example of the Metropolis method (using a symmetric proposal) and an example that showcases the Metropolis-Hastings method (using an asymmetric proposal that needs the Hastings ratio correction). | ||
| − | You can tinker with the second one by setting a flag to evaluate the Metropolis method with this asymetric proposal | + | You can tinker with the second one by setting a flag to evaluate the Metropolis method with this asymetric proposal. Of course, this gives a horrible result. |
<br> | <br> | ||
---- | ---- | ||
Revision as of 18:09, 24 September 2013
Markov Chain Monte Carlo in Practice Seminar
We will read and discuss papers on MCMC, the goal is to bring the students up to speed on MCMC and to bring the different groups (Material Science, Geology, and Evolution) together. We start the semester by reading a few early seminal papers and then will drift in the directions the lead person gives us.
Class meeting is on Mondays 3-4pm in room 499 (on days with faculty meetings we may need to shift to half an hour earlier, or shorten the meeting).
Wikipedia: Metropolis-Hastings algorithm
EXAMPLE CODES IN PYTHON. Currently there are two files that give an example of the Metropolis method (using a symmetric proposal) and an example that showcases the Metropolis-Hastings method (using an asymmetric proposal that needs the Hastings ratio correction).
You can tinker with the second one by setting a flag to evaluate the Metropolis method with this asymetric proposal. Of course, this gives a horrible result.
September 30 Sachin Shanbhag will lead: Marinari, E. and G. Parisi (1992). Simulated Tempering: A New Monte Carlo Scheme. Europhys. Lett., 19 (6): 451-458 download PDF
September 23 Peter Beerli will lead: Geyer C. J. and E. A. Thompson (1995) Annealing Markov Chain Monte Carlo with Applications to Ancestral Inference. download PDF You definitely want to visit Charlie Geyer's website and specifically his rant about on long chains and burn-in of MCMC.
September 16 Sachin Shanbhag will lead: W. K. Hastings (1970) Monte Carlo Sampling Methods Using Markov Chains and Their Applications. Biometrika 57(1):97-109. download PDF
September 9 Peter Beerli will lead.
We will all read this paper and discuss their approach: interesting topics are assumptions, run time, number of evaluation etc.(download PDF, EXAMPLE CODE IN PYTHON)
Here are some more tidbits from the early MCMC times: first textbook by Hammersley and Handscomb (1962) PDF, news from the Los Alamos Science: The Beginning of the Monte Carlo method, Computing and Computers
