PGM-class and MRF parameter learning

I’m taking Stanford CS 228 (a.k.a. pgm-class) on Coursera. The class is great, I guess it provides close to the maximum one can do under the constraints of remoteness and bulkness. The thing I miss is theoretical problems, which were taken aside from the on-line version because they could not be graded automatically.

There is an important thing about graphical models I fully realized only recently (partly due to the class). This thing should be articulated clearly in every introductory course, but is often just mentioned, probably because lecturers consider it obvious. The thing is there is no probabilistic meaning of MRF potentials whatsoever. The partition function is there not only for amenity: in contrast to Bayesian networks, there is no general way to assign potentials of an undirected graphical model to avoid normalization. The loops make it impossible. The implication is one should not assign potentials by estimating frequencies of assignments to factors (possibly conditioned on features) like I did earlier. This is quite a bad heuristic because it is susceptible to overcounting. Let me give an example.

For the third week programming assignment we needed to implement a Markov network for handwriting OCR. The unary and pairwise potentials are somewhat obvious, but there was also a task to add ternary factors. The accuracy of the pairwise model is 26%. Mihaly Barasz tried to add ternary factors with values proportional to trigram frequencies in English, which decreased performance to 21% (link for those who have access). After removing pairwise factors, the performance rose to 38%. Why has the joint model failed? The reason is overcounting evidence: different factor types enforce the same co-occurrences, thus creating bias towards more frequent assignments, and this shows it can be significant. Therefore, we should train models with cycles discriminatively. 

One more thought I’d like to share: graphical model design is similar to software engineering in the way that the crucial thing for the both is eliminating insignificant dependencies on the architecture design stage. 

Read Users' Comments (7)

7 Response to "PGM-class and MRF parameter learning"

  1. Yaroslav Bulatov says:
    20 April 2012 at 10:41

    Here's an interesting side point -- if you could assign potentials to avoid global normalization then computing the partition function could be done efficiently with dynamic programming. And vica versa, if you can compute the partition function efficiently, then you could rewrite your model as a product of factors which can be estimated using simple counting.

  2. reviews says:
    19 April 2017 at 17:40

    Good sharing. It is a very informative blog. Anyone who wants detailed summary of computer classes that theoretical concept can benefit from this blog to have a detailed overview and information.

  3. says:
    30 July 2017 at 18:07

    Online Computer Training Classes - Get the free computer courses, classes, training and certification online at your home free of cost. Nitro India one of the famous organization provide these services to you free of cost. Get now free classes, courses and certification from Nitro all over the world.

  4. Gokul Ravi says:
    26 May 2018 at 11:48

    nice blog
    android training in bangalore
    ios training in bangalore
    machine learning online training

  5. Gokul Ravi says:
    26 May 2018 at 12:32

    useful blog
    python interview questions
    cognos interview questions
    perl interview questions
    vlsi interview questions
    web api interview questions
    msbi interview questions

  6. Gokul Ravi says:
    26 May 2018 at 12:33

    laravel interview questions aem interview questions
    salesforce interview questions
    oops abab interview questions itil interview questions
    informatica interview questions
    extjs interview questions

  7. Gokul Ravi says:
    26 May 2018 at 12:33

    sap bi interview questions
    hive interview questions
    seo interview questions
    as400 interview questions
    wordpress interview questions accounting interview questions
    basic accounting and financial interview questions

Post a Comment