Are there any efficient ways to do structure learning in an online fashion? All the papers I find on structured output prediction just assume you can quickly find the optimal configuration.

Am I suppose to just run a greedy algorithm at every time step as I update my parameters?

asked Jun 25 '12 at 21:29

zaxtax's gravatar image

zaxtax ♦
1051122545

Do you know "(Online) Subgradient Methods for Structured Prediction" http://repository.cmu.edu/cgi/viewcontent.cgi?article=1054&context=robotics? Not sure if it will help you, though.

I am not sure I understood you question: what does the online setting have to do with finding the argmax for a single example? And where do you need a greedy algorithm?

(Jun 26 '12 at 03:17) Andreas Mueller

Andreas: there are two senses in which online can be understood in structured prediction. One is the one you mention: you get one input and one structured output at a time, make a prediction, suffer a loss, move on. What zaxtax seems to be talking about, however, is a setting where you can make updates before inference has converged, which can be useful if inference is intractable or just expensive. This sense is the one in which samplerank can help.

(Jun 26 '12 at 03:55) Alexandre Passos ♦

That is a weird meaning of "online". zaxtax: What kind of objective are you interested in? Maximum margin or maximum likelihood? What does your inference problem look like and what kind of inference algorithm are you using?

(Jun 26 '12 at 04:27) Andreas Mueller

Usually you use the BIC score, Maximum likelihood is not a very good estimator for optimal graph structure

(Jun 26 '12 at 04:37) Leon Palafox ♦

What's the correct terminology to clarify that this is a setting where inference is ridiculously intractable?

(Jun 26 '12 at 05:10) zaxtax ♦

2 Answers:

You should read the SampleRank paper. It's a simple algorithm that does very fast structured learning. It's not online, but it can train (say) a trigram POS tagger on wall-street-journal data in a few minutes, because it updates the parameters as you gibbs sample around a solution.

answered Jun 25 '12 at 23:43

Alexandre%20Passos's gravatar image

Alexandre Passos ♦
2554154278421

Did you check this paper by Murphy, he is pretty good with algorithms to do fast inference on Bayesian Nets using prior structures over the possible graph spaces so you do not do just random changes and hope for the best.

answered Jun 26 '12 at 03:37

Leon%20Palafox's gravatar image

Leon Palafox ♦
40857194128

Your answer
toggle preview

powered by OSQA

User submitted content is under Creative Commons: Attribution - Share Alike; Other things copyright (C) 2010, MetaOptimize LLC.