← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8953.c · 生成模型:GPT-6

广义拉普拉斯调节尾形

理解等级:中级

分布p(x)正比于exp(−|x/s|ʳ),r>0、s>0,其中Γ是归一化中的Gamma函数。r=2对应高斯形状,r=1对应拉普拉斯形状,r更尖且相对厚尾。s是尺度参数而非直接等于标准差;固定s令r趋零时不能据图形尖峰直接断言收敛为零点质量。

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

      Where Γ is the Gamma distribution. This function, equation (27.13), known as the generalized Laplacian distribution, has two parameters: the exponent r, which alters the distribution's shape; and the scale parameter s, which influences its variance. This distribution fits the outputs of band-pass filters of natural images remarkably well. In natural images, typical values of r lie within the range [0.4, 0.8].

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第27章《统计图像模型》(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/c027b.xhtml) - 结论依据:摘录中关于广义拉普拉斯调节尾形的定义、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:Where Γ is the Gamma distribution. This function, equation (27.13), known as the generalized Laplacian distribution, has two parameters: the exponent r, which alters the distribution's shape; and the scale parameter s, which influences its variance. This distribution fits the outputs of band-pass filters of natural images remarkably well. In natural images, typical values of r lie within the range [0.4, 0.8].