知识卡片

普通读书笔记卡 · 8861 · 生成模型:GPT-6

正则化的先验解释要对齐符号

理解等级:中级

后验与似然乘先验成正比,取负对数后成为负对数似然与负对数先验之和,形成正则目标。样本平均和总和会改变系数尺度,因此λ需与噪声和样本量约定对齐。不是任意拟合损失都可在未指定概率模型时直接称作似然。

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

      this interpretation, the data fit loss (e.g., L2 loss) is a likelihood function and the regularizer is a prior p(θ). Bayes’ rule then states that the posterior is proportional to the product of the prior and the likelihood. The log posterior is then the sum of the log likelihood and the log prior, plus a constant. Hence we arrive at the form of equation (11.12).

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第11章《泛化问题》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c011.xhtml) - 结论依据:摘录中关于正则化的先验解释要对齐符号的定义、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:this interpretation, the data fit loss (e.g., L2 loss) is a likelihood function and the regularizer is a prior p(θ). Bayes’ rule then states that the posterior is proportional to the product of the prior and the likelihood. The log posterior is then the sum of the log likelihood and the log prior, plus a constant. Hence we arrive at the form of equation (11.12).