知识卡片
训练目标通过反向传播求导
反向传播沿计算图使用链式法则计算参数对损失的梯度,优化器再按梯度更新。求梯度与选择更新规则是两件事,梯度下降也不保证找到复杂网络的全局最优;特定训练判定问题的复杂度不能直接概括为所有训练步骤都NP完全。
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
- · 人工神经网络
原书出处 · 1 处
- · 查看原书依据
This process is nontrivial, as calculating the partial derivatives across the computation graph of the neural network requires the chain rule. An algorithm known as backpropagation is described in section C.5.4.
参考来源
- 位置:《Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)》第10章《人工神经网络》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Introduction to Autonomous Robots Mechanisms, Sensors, Actuators, and Algorithms (Nikolaus Correll, Bradley Hayes etc.)/_epub-src/OEBPS/xhtml/chapter_10.xhtml)
- 结论依据:摘录中关于训练目标通过反向传播求导的定义、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:This process is nontrivial, as calculating the partial derivatives across the computation graph of the neural network requires the chain rule. An algorithm known as backpropagation is described in section C.5.4.