4 ms·
Crudely: combine the speed of classification trees with the generally low error rates (flexibility) of the SVM.
by mturmon 14y ago
Crudely: combine the speed of classification trees with the generally low error rates (flexibility) of the SVM.
- sgt101 14y agoThere is a good implementation in Mahout and it scales well on Hadoop, so its a pretty good out of the box whole data set learning algorithm. No advantage over taking a decent statistically valid subset and running C4.5 (or a variant with floating point splits) apart from it's less work.