Christian Robert addressing BRAG about improving MCMC

Christian took some time to finish off the abridged meeting from last week so he could give us a few more leads about how to improve MCMC mixing and convergence.

  1. samclifford
    An extra special meeting of my research group to have #xian tell us a bit more about improving MCMC. Tempering to flatten target density.
    Wed, Aug 22 2012 18:06:44
  2. samclifford
    Run parallel chains with diff starting values and look at posterior density as you go to check convergence and importance of region #xian
    Wed, Aug 22 2012 18:27:25
  3. samclifford
    Change of variable to map posterior space to a hypercube may lead to concentration in small dots but you have a more compact space #xian
    Wed, Aug 22 2012 18:31:34
  4. samclifford
    Introducing an auxiliary variable z so that a prior on x becomes a prior on (x,z) can allow the prior on x to be discovered quicker #xian
    Wed, Aug 22 2012 18:36:12
  5. samclifford
    Idea of energy in MCMC and looking at level sets (isocontours of posterior density) means “large” jumps possible #xian
    Wed, Aug 22 2012 18:42:33
  6. samclifford
    Question: a lot of knowledge about low density rejected MH steps. How do we use this info to avoid these regions? #xian
    Wed, Aug 22 2012 18:47:12
  7. samclifford
    Answer: we can use this to develop jumping rules. #xian We want to know about this region but we can’t use it without thinking abt acc. rate
    Wed, Aug 22 2012 18:50:00

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s