← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8763 · 生成模型:GPT-6

不变性测试先证明扰动保标签

理解等级:高级

改变输入某属性而预期标签不变,可检查模型对无关变化的敏感性。前提是扰动在目标任务确实应保留含义或结果,例如照明变化未抹掉对象;不能任意改职业就宣称真实贷款风险必须完全不变。

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

      Invariance tests In an invariance test, we apply label-preserving perturbations to the input data and expect the model’s prediction to remain the same. In our loan default prediction example, if there’s a dimension that we expect our model to be invariant to (e.g., occupation), we can write a test where all examples have the same attributes, except for the applicant’s occupation, and assert that the

    参考来源

    - 位置:《Effective Machine Learning Teams (David Tan, Ada Leung, David Colls)》第06章《模型自动化测试》(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/ch06.html) - 结论依据:摘录中关于不变性测试先证明扰动保标签的定义、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:Invariance tests In an invariance test, we apply label-preserving perturbations to the input data and expect the model’s prediction to remain the same. In our loan default prediction example, if there’s a dimension that we expect our model to be invariant to (e.g., occupation), we can write a test where all examples have the same attributes, except for the applicant’s occupation, and assert that the