← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6562.f · 生成模型:GPT-6

最小二乘不要求噪声正态

理解等级:中级

最小化平方误差在适当模型下可对应同方差高斯似然的极大化,但平方损失本身不逻辑强制真实噪声为正态。作点预测与作概率推断需要不同假设;不能从使用MSE倒推已验证加性正态DGP。

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

      mean squared error loss function is equivalent to assuming Ny was additive and had a normal distribution. However, it also assumes that the normal distribution had constant variance, while the funnel shape in the scatterplot indicates the variance of Ny might increase with the value of X.

    参考来源

    - 位置:《Causal AI》第06章《结构因果模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-6.xhtml) - 结论依据:摘录中关于最小二乘不要求噪声正态的定义、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:mean squared error loss function is equivalent to assuming Ny was additive and had a normal distribution. However, it also assumes that the normal distribution had constant variance, while the funnel shape in the scatterplot indicates the variance of Ny might increase with the value of X.