知识卡片
深网络是多层函数复合
深网络可写f=fL∘…∘f1,每层把上一层表示转换成下一层表示。θ汇总所有可学习参数,不只线性层的权重。复合结构允许逐级构造非线性特征,但层数增加本身不保证每层都变得更抽象或更适合任务。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 神经网络
原书出处 · 1 处
- · 查看原书依据
Each layer is a function. Therefore, a deep net is a composition of many functions: (12.9) The L is the number of layers in the net. These functions are parameterized by weights [W1, …, WL] and biases [b1, …, bL]. Some layers we will see later have other parameters. Collectively, we will refer to the concatenation of all the parameters in a deep net as
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:摘录中关于深网络是多层函数复合的定义、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:Each layer is a function. Therefore, a deep net is a composition of many functions: (12.9) The L is the number of layers in the net. These functions are parameterized by weights [W1, …, WL] and biases [b1, …, bL]. Some layers we will see later have other parameters. Collectively, we will refer to the concatenation of all the parameters in a deep net as