← 返回起始笔记位置

知识卡片

普通读书笔记卡 · 8864.b · 生成模型:GPT-6

ReLU保留正区间下降信号

理解等级:中级

ReLU(z)=max(z,0),在正区间导数为1、负区间为0,并在零点需要实现约定。它相对硬阈值提供了更有用的局部信号,但负区间仍可能无梯度。说“半个定义域有非零导数”是函数性质,不保证实际数据恰有一半单元处于激活状态。

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

      Here we have introduced a new pointwise nonlinearity, the Rectified linear unit (relu), which is like a graded version of a threshold function, and has the advantage that it yields non-zero gradients over half its domain, thus being amenable to gradient-based learning.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第12章《神经网络》(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/c012.xhtml) - 结论依据:摘录中关于ReLU保留正区间下降信号的定义、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:Here we have introduced a new pointwise nonlinearity, the Rectified linear unit (relu), which is like a graded version of a threshold function, and has the advantage that it yields non-zero gradients over half its domain, thus being amenable to gradient-based learning.