← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8757.c · 生成模型:GPT-6

变换正确与特征合适分开验收

理解等级:中级

最小值被min-max映射到0是方法定义,不因此自动成为bug。单元测试可确认实现遵从公式,选择它还是log变换则需数据与模型目标评价;不能把“原值正数”作为映射到0必错的理由。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      min-max normalization is the wrong data transformation to apply, as it would baseline the smallest value to 0 (e.g., the “income” of 10 now becomes 0 after normalization). Perhaps we should be using log transformations instead. Through writing this test, we uncovered a bug and found an opportunity to improve our feature engineering logic.

    参考来源

    - 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第05章《软件自动化测试》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)/_epub-src/OEBPS/ch05.html) - 结论依据:摘录中关于变换正确与特征合适分开验收的定义、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:min-max normalization is the wrong data transformation to apply, as it would baseline the smallest value to 0 (e.g., the “income” of 10 now becomes 0 after normalization). Perhaps we should be using log transformations instead. Through writing this test, we uncovered a bug and found an opportunity to improve our feature engineering logic.