Andy, I just recently discussed this with Gilles; There are a number of things involved here: Gilles told me that his experience shows that randomized trees are usually deeper than regular trees thus the increased training time.
After looking at the code I also found that ``_find_random_split`` requires one more pass over the training data compared to ``_find_best_split``. The reason is the following: in order to draw a random number between [min_x, max_x] one has to determine min_x and max_x first. This requires one pass over the training data, after the random split point has been chosen we have to determine the cost of the split - for this we perform the same linear scan over the input data as for ``_find_best_split`` but in this case we know exactly where the split point is. Apart from that, there are also some variables in ``find_random_split`` which are not ctyped - I've fixed this and pushed the modifications to master. best, Peter 2012/6/25 <[email protected]>: > I just read the Post and i was wodering: shouldn't extra trees be faster > than random forests? In the Blog Post they are slower. > Andy > -- > Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet. > > > > Olivier Grisel <[email protected]> schrieb: >> >> Here is the link: >> >> >> http://blog.explainmydata.com/2012/06/ntrain-24853-ntest-25147-ncorrupt.html >> >> -- >> Olivier >> http://twitter.com/ogrisel - http://github.com/ogrisel >> >> ________________________________ >> >> Live Security Virtual Conference >> Exclusive live event will cover all the ways today's security and >> threat landscape has changed and how IT managers can respond. Discussions >> will include endpoint security, mobile security and the latest in malware >> threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ >> ________________________________ >> >> Scikit-learn-general mailing list >> [email protected] >> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general > > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Scikit-learn-general mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/scikit-learn-general > -- Peter Prettenhofer ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
