知识卡片

普通读书笔记卡 · 6570.b · 生成模型:GPT-6

最小反事实依赖距离定义

理解等级:中级

寻找最近的翻转输入,需要先选择特征尺度、距离、可变字段及约束。不同度量会得到不同“最小改动”,相关特征也未必能独立改变。最短向量距离本身不是最低行动成本或最可行方案。

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

      defining a distance metric in the feature space and then finding the feature value on the other side of the class boundary. In this example, the hypothetical condition “if the applicant had $20,000 more a year in income . . .” corresponds to the smallest change to the feature vector (in terms of distance on the decision surface) that would have led to approval. This

    参考来源

    - 位置:《Causal AI》第08章《反事实与平行世界》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-8.xhtml) - 结论依据:摘录中关于最小反事实依赖距离定义的定义、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:defining a distance metric in the feature space and then finding the feature value on the other side of the class boundary. In this example, the hypothetical condition “if the applicant had $20,000 more a year in income . . .” corresponds to the smallest change to the feature vector (in terms of distance on the decision surface) that would have led to approval. This