Ad b ta oos Rachel Zhang Rachel Zhang http://blog.csdn.net/abcjennifer 大纲 • Adaptive basis function models • CART • Boosting • Adaboost Rachel Zhang http://blog.csdn.net/abcjennifer Adaptive basis function models • Kernel methods: 所有数据或部分数据 是啥• Kernel ? – 距离度量 怎样定义好k l？• erne • 怎样学kernel? M i i i lik lih d– ax m z ng e oo – MKL (multiple kernel learning) – Adaptive basis function model (ABM) Rachel Zhang http://blog.csdn.net/abcjennifer Basis function, Learnt from data CART Rm: region m, 由basis function定义 Wm: mean response Vm: encodes the variable to split on Rachel Zhang http://blog.csdn.net/abcjennifer CART • CART model: • Find best split: • Algorithm: Cost减少太小？ 树高超过指定值？ 是否所有label分布已经pure了? Rachel Zhang http://blog.csdn.net/abcjennifer CART • 实战： – 加载数据集 – 计算gini index 根据最佳分割feature进行数据分割– – 根据最大信息增益选择最佳分割feature – 递归构建决策树 – 样本分类 Rachel Zhang http://blog.csdn.net/abcjennifer CART • Misclassification rate • Entropy • G
