知识卡片

普通读书笔记卡 · 6575.e · 生成模型:GPT-6

反事实采样可计算目标函数

理解等级:中级

一旦能从所需反事实分布取样,可逐样本计算遗憾、阈值事件或其他函数,再平均估计其期望。若事实购买额固定为50,替代购买减50就是对应差值样本;估计仍受模型、近似推断及蒙特卡洛误差影响。

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

      9.1.4 Counterfactual Monte Carlo The output of the general probabilistic counterfactual inference algorithm produces samples from a distribution. Recall from chapter 2 that once you can sample from a distribution, you can apply the Monte Carlo techniques to make inferences based on that distribution. That same is true with counterfactual distributions. For example, in chapter 8, I introduced the idea of regret, where we compare

    参考来源

    - 位置:《Causal AI》第09章《通用反事实推断算法》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-9.xhtml) - 结论依据:摘录中关于反事实采样可计算目标函数的定义、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:9.1.4 Counterfactual Monte Carlo The output of the general probabilistic counterfactual inference algorithm produces samples from a distribution. Recall from chapter 2 that once you can sample from a distribution, you can apply the Monte Carlo techniques to make inferences based on that distribution. That same is true with counterfactual distributions. For example, in chapter 8, I introduced the idea of regret, where we compare