知识卡片

普通读书笔记卡 · 6595.f · 生成模型:GPT-6

可容许性比较整体风险函数

理解等级:高级

统计决策可容许性指不存在另一规则在所有参数状态下风险不更高、且至少一处严格更低。它不意味着该规则在每个状态、每次数据或每条实现轨迹上都胜过别人;必须说明模型族、损失和比较规则集合。

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

      admissible if there are no other rules that dominate it. A decision rule dominates another rule if the performance of the former is sometimes better, and never worse, than that of the other rule with respect to the utility function. For example, the softmax decision rule is dominated by maximizing expected utility (assuming you know the true shape of P (YX = x )) because

    参考来源

    - 位置:《Causal AI》第12章《因果决策与强化学习》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-12.xhtml) - 结论依据:摘录中关于可容许性比较整体风险函数的定义、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:admissible if there are no other rules that dominate it. A decision rule dominates another rule if the performance of the former is sometimes better, and never worse, than that of the other rule with respect to the utility function. For example, the softmax decision rule is dominated by maximizing expected utility (assuming you know the true shape of P (YX = x )) because