apurva

Hi All, I have been working on a recommender engine for a while now and have now stumbled across the cold start problem. The problem here is that whatever data I collect is only an indicator of the likes of the user (for eg., if browsing history is taken as a source, then the basic assumption that people don't browse for what they don't like stands true) So in such a case, any ideas as to how I train the system initially for dislikes?? I do know that the system will gradually tune to the user preferences with continuous feedback, but I would not like the first run to be very erratic either by choosing random dislikes... Any ideas folks?? Any help in the matter is greatly appreciated....

Hey, So for some one who is starting out with Support Vector Machines, there is just loads of material out there which does not make immediate sense. I found the following resources (may suggest in exploring in the order mentioned) really helpful, maybe they can help you out too :

http://www.tristanfletcher.co.uk/SVM%20Explained.pdf

http://pyml.sourceforge.net/doc/howto.pdf ( Personally, I found this brilliant and it took some effort to dig this up)

Finally: http://videolectures.net/mlss06tw_lin_svm/

Hope this helps..

I am currently pursuing my masters in CS with an inclination in machine learning from a reasonably well reputed university. With a reasonable background in AI, I wish to take up a summer internship in a related field. Could people here guide me to apply to places they know for a fact employ machine learning techniques? Any help on this is greatly appreciated. Thanks...