标签:and tin rate bootstra training ror ppi fit nothing
2.1 经验误差与过拟合
1. error rate/accuracy
2. error: training error/empirical error, generalization error
3. overfitting and underfitting
2.2 评估方法
1. hold out. 直接将数据集$D$划分为两个互斥的集合 $D = S \cap T, S \cap T = \varnothing$ , stratified sampling:分层采样
2. cross validation.
3. bootstrapping
标签:and tin rate bootstra training ror ppi fit nothing
原文地址:http://www.cnblogs.com/vincentcheng/p/6921388.html