知识卡片
反向传播求梯度而非选更新
网络结构、梯度计算与参数更新是不同职责。反向传播高效计算损失对参数的导数,梯度下降等优化器再决定如何更新。原书局部把反向传播称优化器较宽泛;硬阈值处不可微且其余位置梯度为零,通常还需替代目标、专门规则或无梯度搜索。
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
general-purpose optimizers like gradient descent (for differentiable objectives) or zeroth-order methods (for nondifferentiable objectives). The next chapter will cover the backpropagation algorithm, which is a general-purpose gradient-based optimizer that applies to essentially all neural networks we will see in this book (but, note that for the perceptron objective, because it has a nondifferentiable threshold function, we would instead opt for a zeroth order optimizer).
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:摘录中关于反向传播求梯度而非选更新的定义、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:general-purpose optimizers like gradient descent (for differentiable objectives) or zeroth-order methods (for nondifferentiable objectives). The next chapter will cover the backpropagation algorithm, which is a general-purpose gradient-based optimizer that applies to essentially all neural networks we will see in this book (but, note that for the perceptron objective, because it has a nondifferentiable threshold function, we would instead opt for a zeroth order optimizer).