知识卡片

操作参考卡 · 8931.b · 生成模型:GPT-6

示意代码先核对接口语义

理解等级:中级

本章代码图表达conv→ReLU→GAP→fc→Softmax的顺序,但nn.conv、nn.fc等是未指定库的示意接口。落地前需明确批次/通道轴、padding、池化输出形状和Softmax的类别轴,确认损失需要logits还是概率;不能把代码图直接视为可运行的特定框架程序。

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

      Such a network is also very easy to define in code, once we have a library of primitives for basic operations like convolution and softmax: 24.5 A Worked Example In this section, we will analyze the simple network described in section 24.4, trained to discriminate between horizontal and vertical lines. Each subsection will tackle one aspect of the analysis that should be part of training

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第24章《卷积神经网络》(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/c024b.xhtml) - 结论依据:摘录中关于示意代码先核对接口语义的定义、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:Such a network is also very easy to define in code, once we have a library of primitives for basic operations like convolution and softmax: 24.5 A Worked Example In this section, we will analyze the simple network described in section 24.4, trained to discriminate between horizontal and vertical lines. Each subsection will tackle one aspect of the analysis that should be part of training