Hello Richard, > First, it looks like the function > def _partition_features(forest, n_total_features): > partitions features evenly across cores. Am I reading that correctly?
Yes, this is an internal function used to assign blocks of features to cores. > If so, does this mean that on 2 cores with 100 features, feature 1 and > feature 100 are not going to be in any tree together? ... but this function is actually used internally to sort the data along each feature, in parallel. Which means that each core sorts its block of features and then all blocks are recombined to form the X_argsort table. Decision trees are then build using that table, which is shared across all cores. So the short answer is no. All features will be considered when building a decision tree, as it should. > Second, it looks like someone is planning to sample examples but that > it isn't currently operational: it's setting sample_mask to either > None or all ones. No. This is an internal variable. It is used to track the samples in the current node. At first, all samples are in the root node, hence all ones. If set to None, an array with all ones is built. Hope this helps. Gilles ------------------------------------------------------------------------------ WINDOWS 8 is here. Millions of people. Your app in 30 days. Visit The Windows 8 Center at Sourceforge for all your go to resources. http://windows8center.sourceforge.net/ join-generation-app-and-make-money-coding-fast/ _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
