| 摘要 |
The present invention relates to a method for incremental learning of a classification model, where pre-defined weak incremental learners are distributed over the distinct regions in a set of partitionings of the input domain. The partitionings and regions are organized via a binary tree and they are allowed to vary in a data-driven way, i.e., in a way to minimize the classification error rate. Moreover, to test a given data point, a mixture of decisions is obtained through the models learned in the regions that this point falls in. Hence, naturally, in the cold start phase of the data stream, the simpler models belonging to the larger regions are favored and as more data get available, the invention automatically puts more weights on the more complex models. |
| 主权项 |
1. A method for incremental learning for a given data point at any time, wherein the method comprises the steps of:
S101, propagating forward on the decision tree to locate a leaf node; S102, computing a node decision by,
computing first a local decision at a node, then computing the node decision, if this node is non-leaf, as the weighted sum of the local decision and the node decision of the corresponding child; if this node is the leaf, as the local decision; S103, updating a local classification model; S104, updating a node classification model. i.e., the weighting coefficients; S105, updating a node partitioning model; if this is not a root node, S108, back propagating to the upper parent and go to b, otherwise, S106, returning the sign of this node decision as the class label predicts. |
