|
I am taking an online class on graphical models, however, it seems to cover a lot of the background theory, and almost no applications. Thus, I have no idea how I would apply the concept to problems arising in industry. Any useful tutorials/papers/books/blogs on this? Perhaps something analogous to Rabiner's tutorial on HMMs. Thanks |
|
Alex Smola has really good slides on Graphical models for the internet if that is your thing. There are also many important applications in vision, speech, etc. |
|
I'm venturing to guess you are taking Koller's course on PGM. It is indeed very focused on theory, as Graphical Models should be. If you are looking more on direct applications, I may suggest you looking for papers where they applied some concepts, if you tell me which kind of specific industries you are interested in, I might help you a bit more. [Applications Update] On Market trading, one of the most basic examples, and a rather simple thing to implement is a Hidden Markov Model to track stock markets, using this model, you can use a HMM and see how a company is doing. You can make it more interesting using Coupled Hidden Markov Models (CHMM) which model the temporal relationship among different variables over time. Let's say you want to model how an oil company performance is affected by the war in the middle east you can do it! If you want to do something like that, I recommend you looking for papers with "Dynamic Bayesian Network" and Trade markets, you will find plenty. Just a quick reminder, HMM and CHMM are nothing but specific instances of Dynamic Bayesian Networks, look in Koller's lecture for the Lecture on Temporal Variables. In particular, I am looking into using them for detecting specific market conditions and using them for trading. Anything that is on some level related to this conceptually(perhaps medical diagnosis paper, given that it is written and explained well) would be very useful. Of course, the most useful would be a paper that addresses my particular field.
(Apr 05 '12 at 11:24)
Viktor Simjanoski
I added a bit to the answer, hope it helps
(Apr 06 '12 at 03:11)
Leon Palafox ♦
Thanks a lot Leon.
(Apr 06 '12 at 10:25)
Viktor Simjanoski
|
|
The Koller and Friedman book actually is chock full of case studies. It is also almost ... too comprehensive. http://pgm.stanford.edu/ HMMs are graphical models. So are Boltzmann machines, conditional random fields, latent Dirichlet allocation. Many things that you might not think of as a graphical model can be viewed as one, such as logistic regression. Just off the top of my head, TrueSkill is a pretty cool application of graphical models. Where exactly are the case studies in Koller & Friedman? I took a look at it, but can't find a special chapter dedicated to them, nor it seems like there is specific section for case studies per chapter.
(Apr 05 '12 at 01:55)
Viktor Simjanoski
1
They are in the grey boxes, those are special boxes dedicated for examples
(Apr 05 '12 at 04:23)
Leon Palafox ♦
|
