知识卡片
网络数据流与训练职责
用训练数据得到参数,再将输入通过带参数的多层变换得到表示与输出。梯度计算和参数更新另属训练环节。
结构图:
flowchart TD
A[训练数据与目标] --> B[梯度计算及优化更新]
B --> C[各层参数θ]
D[批量输入张量] --> E[仿射变换]
C --> E
E --> F[非线性及归一化]
F --> G[后续层表示]
G --> H[任务输出映射]
H --> I[训练损失]
I --> B
H --> J[预测结果的实际解释]
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
Figure 12.7: Learning is a function that maps a dataset to parameters. Inference, through a neural net, is a function that maps a datapoint to activations. 12.5 Deep Nets Deep nets are neural nets that stack the linear-nonlinear motif many times (figure 12.8): Figure 12.8: Deep nets consist of linear layers interleaved with nonlinearities. Each layer is a function. Therefore, a deep net is a
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:摘录中关于网络数据流与训练职责的定义、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:Figure 12.7: Learning is a function that maps a dataset to parameters. Inference, through a neural net, is a function that maps a datapoint to activations. 12.5 Deep Nets Deep nets are neural nets that stack the linear-nonlinear motif many times (figure 12.8): Figure 12.8: Deep nets consist of linear layers interleaved with nonlinearities. Each layer is a function. Therefore, a deep net is a