Friday, April 29, 2011

Another ML blog

I just noticed that Justin Domke has a blog --

He's one of the strongest researchers in the field of graphical models. I first came across his dissertation when looking for a way to improve loopy-Belief Propagation based training. His thesis gives one such idea -- instead of maximizing the fit of an intractable model, and using BP as intermediate step, maximize the fit of BP marginals directly. This makes sense since approximate (BP-based) marginals are what you ultimately use.

If you run BP for k steps, then likelihood of the BP-approximated model is tractable to minimize -- calculation of gradient is very similar to k steps of loopy BP. I derived formulas for gradient of "BP-approximated likelihood" in equations 17-21 in this note. It looks a bit complicated in my notation, but derivation is basically an application of chain rule to BP-marginal, which is a composition of k-steps of BP update formula.

1 comment:

Anonymous said...

when there is so much hustle among social networking sites, here you get the best, the best working hack for snapchat hack. Now anyone can hack any snapchat account live in seconds. I really loved this snapchat hack, as now its very simple for me to hack any snap chat account. Its working hack for snapchat. I really wondered to get snapchat password hack and I got it lastly on this website. It works with charm and now I am able to hack any snapchat online. I would really be thankful to the mentioned website which helped me in doing so. If you really desire to hack snapchat of your kids, Visit the snaphack generator on the mentioned website, and generate the hack for any snapchat account just by feeding the user name. Visit to explore the snapchat account of your kids and keep track of their daily activity.