← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 8890.c · 生成模型:GPT-6

高斯极限需保留概率尺度条件

理解等级:高级

归一化高斯在σ趋零时以分布意义趋向单位冲激,并与热扩散解相关。重复卷积趋近高斯的中心极限定理还需要非负归一化、有限方差及适当中心化缩放等条件;任意有符号滤波核反复卷积不能直接套用这一结论。

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

      Repeated convolutions of any function concentrated in the origin result in a Gaussian (central limit theorem). In the limit σ → 0 the Gaussian becomes an impulse. This property is shared by many other functions, but it is a useful thing to know. 17.3.3 Limitations However, many of these properties are only true for the continuous version of the Gaussian and do not work for

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第17章《模糊滤波器》(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/c017b.xhtml) - 结论依据:摘录中关于高斯极限需保留概率尺度条件的定义、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:Repeated convolutions of any function concentrated in the origin result in a Gaussian (central limit theorem). In the limit σ → 0 the Gaussian becomes an impulse. This property is shared by many other functions, but it is a useful thing to know. 17.3.3 Limitations However, many of these properties are only true for the continuous version of the Gaussian and do not work for