知识卡片

普通读书笔记卡 · 6593.b · 生成模型:GPT-6

潜混杂压缩必须保留观测规律

理解等级:高级

用低维Z代理多个潜混杂,可以简化计算,但只有模型足以表示相关观测分布且保留所需图限制时,才能把拟合结果用于已识别目标。前门识别不保证任意一维Normal潜变量和小网络都无误设或能训练到真值。

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

      Collapsing the latent confounders into these proxies reduces the dimensionality of the estimation problem, and any loss of information that occurs from collapsing these variables won’t matter because we are ultimately relying on information flowing through the front door. We’ll create a causal generative model inspired by the variational autoencoder, where is a latent encoding and θE and θI become weights in decoders. This is

    参考来源

    - 位置:《Causal AI》第11章《构建因果推断工作流》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-11.xhtml) - 结论依据:摘录中关于潜混杂压缩必须保留观测规律的定义、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:Collapsing the latent confounders into these proxies reduces the dimensionality of the estimation problem, and any loss of information that occurs from collapsing these variables won’t matter because we are ultimately relying on information flowing through the front door. We’ll create a causal generative model inspired by the variational autoencoder, where is a latent encoding and θE and θI become weights in decoders. This is