知识卡片

普通读书笔记卡 · 9256.b · 生成模型:GPT-6

纯仿射层串接仍是仿射映射

理解等级:中级

若所有层都仅作仿射变换,复合结果仍可写成一次仿射变换。非线性激活使多层表达超出此范围,但非线性分类边界并非只能由神经网络得到,核方法与树也有相应能力。

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

      Activation functions: these are mathematical functions applied to the outputs of perceptrons. Their purpose is to introduce nonlinear functionality into neural networks to make predictions that may not be possible by linear models. For example, if we consider the weighted sum operation of the perceptron , without the activation function f(.), this model is identical to a linear regression model, where the wi’s represent the

    参考来源

    - 位置:《Machine Learning Fundamentals: Concepts, Models, and Applications》第08章《深度学习模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Machine Learning Fundamentals: Concepts, Models, and Applications/_epub-src/OEBPS/xhtml/21_chapter08.xhtml) - 结论依据:摘录中关于纯仿射层串接仍是仿射映射的定义、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:Activation functions: these are mathematical functions applied to the outputs of perceptrons. Their purpose is to introduce nonlinear functionality into neural networks to make predictions that may not be possible by linear models. For example, if we consider the weighted sum operation of the perceptron , without the activation function f(.), this model is identical to a linear regression model, where the wi’s represent the