← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 9163.a · 生成模型:GPT-6

梯度为零不够证明最优

理解等级:中级

可微目标的内部极值常满足导数为零,但驻点也可能是极大值、鞍点或退化解。求出直线角度候选后,还需核对真正最小残差与参数约定,不能把任何零导数结果直接标成最佳拟合。

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

      This approach is known as the least-squares method and can be used to fit data to any parametric model (i.e., a model that has numbers sought to make it fit our data best). The general approach is to describe the fit between the data and the model in terms of a difference, known as an “error.” The best fit will minimize this error—that is, the

    参考来源

    - 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第09章《特征提取》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)/_epub-src/OEBPS/xhtml/chapter_9.xhtml) - 结论依据:摘录中关于梯度为零不够证明最优的定义、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 approach is known as the least-squares method and can be used to fit data to any parametric model (i.e., a model that has numbers sought to make it fit our data best). The general approach is to describe the fit between the data and the model in terms of a difference, known as an “error.” The best fit will minimize this error—that is, the