Improvement using Mondrian Forests Sample Clauses

Improvement using Mondrian Forests. A novel algorithm that addresses the above problems is the Mondrian Forest by Xxxxxxxxxxxxxxxx et al. [3]. The major difference between a Mondrian tree and a standard decision tree is that Mondrian trees also store the extent of the data that corresponds to each node. While the decision tree uses splits that range over the entire potential range of the data, the splits of a Mondrian tree only cover the actual data range. This is achieved by keeping bounding box information for each sub tree. Note that the data itself is not stored in the tree, only the bounding boxes. As in random decision trees, splits are generated randomly, but samples are drawn from an exponential distribution whose rate parameter is proportional to the data extent of the sub tree.