知识卡片
奇数近邻数不能解决全部平票
二类等权投票中奇数k可避免两类票数各半,但多类仍可能同票,距离加权或边界同距离也需规则。k≈√n等经验式仅作候选,最终用验证选择,不动用最终测试调k。
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
- · 监督学习模型
原书出处 · 1 处
- · 查看原书依据
The value of K should be an odd number to avoid ties. A conventional way of finding k value is by using the expression: where n is the number of data points. The data should be divided into training and validation sets. Usually, the test set is not used. The validation set is used to tune the value of K. Then select K that works
参考来源
- 位置:《Machine Learning Fundamentals: Concepts, Models, and Applications》第06章《监督学习模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Machine Learning Fundamentals: Concepts, Models, and Applications/_epub-src/OEBPS/xhtml/19_chapter06.xhtml)
- 结论依据:摘录中关于奇数近邻数不能解决全部平票的定义、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:The value of K should be an odd number to avoid ties. A conventional way of finding k value is by using the expression: where n is the number of data points. The data should be divided into training and validation sets. Usually, the test set is not used. The validation set is used to tune the value of K. Then select K that works