知识卡片
交叉验证还需独立的选择层次
k折通常将数据分为k个互斥折,轮流留一折评估,并非任意重复随机切分k次。若用折得分选超参数,再用同一最佳得分报告泛化会乐观,可另设外层评价或最终保留集。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 训练机器学习模型
原书出处 · 1 处
- · 查看原书依据
In addition to splitting the data into a training, validation, and testing set once and evaluating performance, it is common (and also good practice) to use cross validation to mitigate biases in the evaluated metrics due to the data set splits. Cross validation divides the data into multiple train/validation/test splits and trains and evaluates the model multiple times. The model is evaluated on each testing
参考来源
- 位置:《Machine Learning Fundamentals: Concepts, Models, and Applications》第09章《训练机器学习模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Machine Learning Fundamentals: Concepts, Models, and Applications/_epub-src/OEBPS/xhtml/23_chapter09.xhtml)
- 结论依据:摘录中关于交叉验证还需独立的选择层次的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:In addition to splitting the data into a training, validation, and testing set once and evaluating performance, it is common (and also good practice) to use cross validation to mitigate biases in the evaluated metrics due to the data set splits. Cross validation divides the data into multiple train/validation/test splits and trains and evaluates the model multiple times. The model is evaluated on each testing