{
 "nbformat": 4,
 "nbformat_minor": 5,
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python"
  },
  "lecture_migration": {
   "project_id": "c0b185e7-d7e3-48b6-a605-8f1acdfe243c",
   "notebook_id": "15edefc493064d3382fb418f52d8fabb",
   "destination": "L2",
   "cloud_run": false
  },
  "title": "CLS 03 · 同一台 CNN：局部计算、共享反馈与检验"
 },
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "b7b00745243a4c9b952084ed1f9ac795",
     "type": "text-cell-h1",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-1"
   },
   "source": "# CLS 03 · 同一台 CNN：局部计算、共享反馈与检验",
   "id": "b7b00745243a4c9b952084ed1f9ac795"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "853ea4130617463787ba0ed8903017f7",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-2"
   },
   "source": [
    "[返回第二讲 · 局部计算与共享反馈](https://codingai-lec02.pages.dev/#/convolution-four-products/0) · [第二讲导读](https://codingai-lec02.pages.dev/course/guide.html) · [问卷作业 ↗](https://tally.so/r/VLroVv)（外链可右键在新标签页打开）\n",
    "\n",
    "本册沿第二讲 B 段拆开同一台灰度数字 CNN：两组 Conv／ReLU／Pool，随后 Flatten→32→10。读完要把前向的形状变化和反向的参数共享对应起来。\n",
    "\n",
    "**按眼前的问题读**：核对局部乘加看第 2 节，共享参数怎样收齐反馈看第 3 节，步长与感受野看第 4 节，完整接口看第 5 节，平移检验看第 6 节。第 1 节先提供整机地图；BatchNorm 的补充在册末按需查阅。\n",
    "\n",
    "**动手选择**：[LAB 01 · 学会一个卷积核](https://codingai-lec02.pages.dev/course/lab01.html) 核对窗口与共享梯度；[LAB 02 · 小 CNN 与图像平移](https://codingai-lec02.pages.dev/course/lab02.html) 比较图像平移前后的预测；[LAB 08 · 从真实响应到遮挡对照](https://codingai-lec02.pages.dev/course/lab08.html) 读取真实响应并做遮挡对照。可先读保存结果，再选择一项条件修改；不要求把几项训练全部重跑。"
   ],
   "id": "853ea4130617463787ba0ed8903017f7"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "22957d64a63546b7a9c083534e7a89c2",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-3"
   },
   "source": "## 1. 先认清机器，再区分结构和学习数值",
   "id": "22957d64a63546b7a9c083534e7a89c2"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "cd614fc79856471f893f220a7916ad4b",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-4"
   },
   "source": [
    "**课件与图解**：[先看我们要拆开的这台小CNN](https://codingai-lec02.pages.dev/#/tiny-cnn-preview/0) · [每次读附近，不同位置共用权重](https://codingai-lec02.pages.dev/#/locality-and-sharing/0) · [人规定连接方式，训练调整具体测量](https://codingai-lec02.pages.dev/#/convolution-priors-and-learning/0)\n",
    "\n",
    "第一遍先沿模块向下读：图像怎样变成十个分数？第二遍只看输出形状：哪一步减少空间位置，哪一步只是换一种排列？\n",
    "\n",
    "![同一台26698参数CNN的模块、每张图输出形状与含偏置参数数目；16×7×7展平为784再经32到10](https://codingai-lec02.pages.dev/course/visuals/deepnote-final/foundations-cnn-interfaces.png)\n",
    "\n",
    "图中两组局部处理都含 Conv、ReLU、Pool。池化只缩小宽高，不改变通道数；Flatten 把 16×7×7 个值按固定顺序排成 784 维。类别标签在出口与十个分数形成损失，反馈再经读出和局部模块回到卷积权重。完整逐算子账本在第 5 节，便于把卷积和池化各自的作用拆开核对。\n",
    "\n",
    "局部与共享是两个不同约束：一个输出只读附近窗口，限定读取范围；窗口移到另一处仍使用同一组核权重，限定哪些连接共用参数。人决定核大小、步长、通道数和共享方式，训练决定核中的具体正负数值。给出局部连接，不等于已经手工指定“检测哪种边缘”。更远的关系要由后续层继续组合。\n",
    "\n",
    "**先预测与核对**：在模块图上标出有参数的操作，再口头把其中一层卷积展开为两个窗口；分别说出“由设计固定”和“可由标签反馈调整”的对象。**核对**：窗口位置可以不同，权重对象只有同一份。下一节把大图缩小到四个乘积，核对这个对象究竟算什么。"
   ],
   "id": "cd614fc79856471f893f220a7916ad4b"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "5421d4ab7da348fc9696792e189fd4dc",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-5"
   },
   "source": "## 2. 一个输出、多个通道与多套 filter",
   "id": "5421d4ab7da348fc9696792e189fd4dc"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "3ef076a7435e4f90bc3b1359f8212652",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-6"
   },
   "source": [
    "**课件与图解**：[四个乘积相加，产生一个输出](https://codingai-lec02.pages.dev/#/convolution-four-products/0) · [移动的是窗口，共用的仍是同一个核](https://codingai-lec02.pages.dev/#/convolution-slide-the-kernel/0) · [一个filter读取全部输入通道](https://codingai-lec02.pages.dev/#/one-filter-all-channels/0) · [多套filter，保留多种局部响应](https://codingai-lec02.pages.dev/#/filter-axis-channel-axis/0)\n",
    "\n",
    "取 3×3 输入 [[0,1,2],[3,4,5],[6,7,8]]，核为 [[0,1],[2,3]]，步长 1、不补零。左上窗口与核逐项相乘：0×0+1×1+3×2+4×3=19。四个乘积只产生一个输出；向右、向下移动同一核，得到输出 [[19,25],[37,43]]。一次前向中改变的是读取的输入，核数值不改变。这里沿用深度学习框架常见的不翻转核的计算约定。\n",
    "\n",
    "有多个输入通道时，一套 filter 含每个输入通道的一张核切片。选定输出位置，各通道先做局部乘加，再把贡献相加；Slides 的 RGB 缩例在该位置得到 1+6+1=8。它是另一个形状算例，不是把上一张 3×3 灰度输入直接变成 RGB。\n",
    "\n",
    "张量 [N,C_in,H,W] 中 N 是图片数，C_in 是每张图的通道数；标准卷积权重 [C_out,C_in,K_h,K_w] 的每一套 filter 读取全部输入通道。于是输入 [2,3,5,5]、权重 [4,3,3,3]、stride=1、padding=0，得到 [2,4,3,3]。求和消去的是同一 filter 内的输入通道轴，多套 filter 分别留下多个输出通道，图片轴不会被混合。\n",
    "\n",
    "**先预测与核对**：先补出 25、37、43，再标出上述三个张量中数字 2、3、4 的不同含义。下一问题是：同一核在多处使用，各处错误怎样改变同一份参数？"
   ],
   "id": "3ef076a7435e4f90bc3b1359f8212652"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "23e458bf9c874d6b9181927c02c84d19",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-7"
   },
   "source": "## 3. 路径贡献先相加，再追问反馈来自哪里",
   "id": "23e458bf9c874d6b9181927c02c84d19"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "a7d9778d8fd74ee180110064d4c47a67",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-8"
   },
   "source": [
    "**课件与图解**：[两个窗口的贡献，汇总到同一对核权重](https://codingai-lec02.pages.dev/course/reference/shared-kernel-two-window-gradients.html) · [只有类别标签，局部窗口怎样收到反馈？](https://codingai-lec02.pages.dev/#/shared-feedback-from-class-loss/0) · [重复同一学习循环，核逐步形成目标响应](https://codingai-lec02.pages.dev/course/reference/learn-a-kernel.html)\n",
    "\n",
    "**先找对象**：图中有两个使用位置，但只有一份 K。两个位置各自收到反馈之后，究竟应改哪一份参数？\n",
    "\n",
    "![共享卷积的两窗口梯度相加；分类网络的最终损失经后层传回各位置的delta，再乘局部输入并汇总](data:image/webp;base64,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)\n",
    "\n",
    "**左图：把一次更新算闭合。** X=[1,2,3]，K=[1,−1]，两个窗口输出 y₀=y₁=−1，目标均为 0。使用求和损失 L=½(y₀²+y₁²)=1，不再除以窗口数；上游反馈 δᵢ=∂L/∂yᵢ=−1。\n",
    "\n",
    "| 旧核上的路径 | 对 k₁ 的贡献 | 对 k₂ 的贡献 |\n",
    "|---|---:|---:|\n",
    "| 窗口 0 | −1×1=−1 | −1×2=−2 |\n",
    "| 窗口 1 | −1×2=−2 | −1×3=−3 |\n",
    "| 先求和 | −3 | −5 |\n",
    "\n",
    "η=0.05 时，统一更新为 K′=[1.15,−0.75]；再前向得到 y′=[−0.35,0.05]，新损失 **0.0625**。如果算完第一个窗口就改核，第二个窗口算的就不再是同一次旧参数前向的导数。\n",
    "\n",
    "**右图：保留局部公式，换清反馈来源。** 真实分类只有类别标签，没有给每个中间窗口指定“正确图案”。最终类别损失经后续 ReLU、池化、卷积和分类头传回 δ₀、δ₁，它们可以不同，也不是标签本身。局部公式仍是 ∂L/∂k₁=1δ₀+2δ₁、∂L/∂k₂=2δ₀+3δ₁。1D 核只缩小了计算对象；此处没有给真实 3×3 分类模型虚构梯度数值。\n",
    "\n",
    "**差分核学习页 换了一项实验。** LAB 01 的核学习以差分响应为监督，核逐渐趋近 [1,−1]；目标已从刚才的 [0,0] 换掉。这段训练轨迹不能接在上表后面，也不等于图像分类已经成功。\n",
    "\n",
    "**先预测与核对**：独立算 k₂ 的两份贡献与新损失，然后沿右图指出 δ 从哪里来。下一节先固定这些权重，只改变窗口的合法起点。"
   ],
   "id": "a7d9778d8fd74ee180110064d4c47a67"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "da50b94ef3f44552990db7824fd801ed",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-9"
   },
   "source": "## 4. 窗口起点决定尺寸，层叠与池化改变空间信息",
   "id": "da50b94ef3f44552990db7824fd801ed"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "f8010d2711be449a9e32e7930ec3be0c",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-10"
   },
   "source": [
    "**课件与图解**：[输出尺寸，可以用合法窗口起点数出来](https://codingai-lec02.pages.dev/#/sampling-geometry/0) · [步长2，只计算每隔一格的窗口](https://codingai-lec02.pages.dev/#/stride-changes-sampling/0) · [层叠后，一个局部响应能联系更远的输入](https://codingai-lec02.pages.dev/#/receptive-field-grows/0) · [池化保留一种统计，舍弃部分细节](https://codingai-lec02.pages.dev/#/pooling-trades-detail/0)\n",
    "\n",
    "**先看一条轴，再扩展到二维。** 下图的每一行是一处窗口起点，右侧是它唯一对应的输出地址。\n",
    "\n",
    "![补零后9格使用3格窗口，stride2的合法起点0、2、4、6对应四个输出；起点8越界](https://codingai-lec02.pages.dev/course/visuals/deepnote-final/foundations-stride-origins.png)\n",
    "\n",
    "7×7 输入用 3×3 核，四边补一圈 0（P=1），计算网格为 9×9。步长 1 时每轴有 7 个合法起点，输出 7×7；保持输入、核、补零和核权重不变，只把步长改为 S=2。现在每轴从 0 开始按 2 移动，合法起点是 **0、2、4、6**。起点 6 的三格仍在 9 格计算网格内；再到 8 时窗口会越界，所以**不生成第五个输出**。每轴有 4 个输出，二维合起来就是 **4×4**；变化的是计算哪些窗口，不是换了一套权重。每轴输出数可写为 floor((n+2p−k)/s)+1，这里不使用空洞卷积。补零规定边界外的值，并未增加真实观测。\n",
    "\n",
    "层叠扩大一个输出可能依赖的输入范围。为隔离 stride 的作用，感受野递推页 使用独立三层算例：核宽都为 3，stride 依次为 1、2、1。从范围 r=1、相邻输出间隔 j=1 开始，逐层先算 r′=r+(k−1)j，再算 j′=js，得到 (r,j)=(3,1)、(5,2)、(9,2)。第三层新增的两个邻居间隔已是原图的两格，所以覆盖范围从 5 扩大到 9。这不是本课小 CNN 的逐层结构。\n",
    "\n",
    "本课 CNN 采用 2×2、stride=2 的最大池化，每个通道分别保留局部峰值；宽高减半，通道数不变。多个响应变成一个统计，窗口内精确位置与其他值不能全部恢复。理论覆盖范围说明存在计算路径，不保证范围内每个像素对当前响应都有非零影响。\n",
    "\n",
    "**先预测与核对**：只把 stride 改成 3，先列起点再计算每轴输出数；核对为 0、3、6，输出 3×3。随后做三层范围递推，解释最大池化与仅改变采样起点分别丢掉了什么。接下来把这些空间操作放回真正训练过的同一台小 CNN。"
   ],
   "id": "f8010d2711be449a9e32e7930ec3be0c"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "4b43af8676604ab58a716ffce93630ce",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-11"
   },
   "source": "## 5. 从真实响应到 784 维，形状和参数逐项对上",
   "id": "4b43af8676604ab58a716ffce93630ce"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "ab7af89027eb4e45955862717b2b06dd",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-12"
   },
   "source": [
    "**课件与图解**：[真实特征图里，仍能看见响应在哪里](https://codingai-lec02.pages.dev/#/features-are-learned-responses/0) · [沿同一台小CNN，核对三个关键接口](https://codingai-lec02.pages.dev/#/tiny-cnn-shape-ledger/0)\n",
    "\n",
    "**这次看真实数值。** 固定官方 MNIST test 0（标签 7），让它经过已经训练好的同一台 CNN。先找：哪些通道的亮处相近？哪些通道在同一位置几乎没有响应？\n",
    "\n",
    "![真实CNN对test0数字7的中间响应：第一层全部8通道、第二层前8通道完整显示](data:image/webp;base64,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)\n",
    "\n",
    "第一行是 ReLU 后、pool1 前的全部 8 个 28×28 通道；第二行是 ReLU 后、pool2 前 16 个 14×14 通道中的前 8 个。**通道保留不同响应，空间格保留响应发生在哪里。** 这些图还不足以给通道命名为某类物体检测器，同通道的三个固定输入比较紧接这一观察问题；它仍不足以建立语义命名。\n",
    "\n",
    "读图条件：输入 uint8/255；每层跨 test 0/1/2 的所有通道共用 Blues 色标，L1 为 [0,2.3630995750427246]，L2 为 [0,8.801921844482422]，nearest 显示。两层独立定标，不能凭“哪层更蓝”比较绝对幅值。主讲金框表示理论输入支持范围，不是因果归因。\n",
    "\n",
    "**再把图放回形状账本。**\n",
    "\n",
    "| 操作后，每张图 | 形状 | 本操作可训练参数（含偏置） |\n",
    "|---|---|---:|\n",
    "| Conv 1→8，3×3，p=1，stride=1 / ReLU | 8×28×28 | 80 |\n",
    "| MaxPool 2×2，stride=2 | 8×14×14 | 0 |\n",
    "| Conv 8→16，3×3，p=1，stride=1 / ReLU | 16×14×14 | 1168 |\n",
    "| MaxPool 2×2，stride=2 | 16×7×7 | 0 |\n",
    "| Flatten | 784 | 0 |\n",
    "| Linear 784→32 / ReLU | 32 | 25120 |\n",
    "| Linear 32→10 | 10 | 330 |\n",
    "\n",
    "总参数 **26,698**。Flatten 按固定顺序重排 16×7×7 个值，没有删值；前面的池化已舍弃过部分细节。十个输出是类别分数，训练时再与标签形成损失。\n",
    "\n",
    "**先预测与核对**：遮住参数列，用“输出数×每个输出的权重数＋偏置”核对参数最多的 Linear 层，再解释为什么 Flatten 参数数为零。保持这个模型，下一节检验图像移动后类别分数是否稳定。\n",
    "\n",
    "*来源：既有 `cnn-plain-state.pt`，最小验证交叉熵选择 epoch 6。本图重新排版已保存的前向数组，未重新训练。完整身份、色标及通道值见[响应清单](https://codingai-lec02.pages.dev/responses-v8/manifest.json)；LAB 08 可重新前向核对。*"
   ],
   "id": "ab7af89027eb4e45955862717b2b06dd"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "2cf5d53796cd4338b1c7c3584cc0b348",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-13"
   },
   "source": "## 6. 响应能随位置移动，分类分数就一定不变吗？",
   "id": "2cf5d53796cd4338b1c7c3584cc0b348"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "9b610aecad6e47d088a090cae43aa7ef",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-14"
   },
   "source": [
    "**课件与图解**：[响应跟着移动，不等于类别分数保持不变](https://codingai-lec02.pages.dev/#/equivariance-is-not-invariance/0) · [只移一格，同一条纹会被采成另一组数](https://codingai-lec02.pages.dev/course/reference/stride-sampling-phase.html) · [同一模型加上增强，哪些结果改变了？](https://codingai-lec02.pages.dev/#/cnn-shift-evidence/0) · [回到同一台CNN，解释预测与学习](https://codingai-lec02.pages.dev/course/reference/tiny-cnn-shapes.html)\n",
    "\n",
    "共享卷积在不同位置使用相同规则。固定同一份参数，在 stride=1 的内部区域，输入平移可带来对应的响应平移，这叫等变。先保留原输入与原响应，移动输入后预测响应的新位置，再揭示比较。有限边界、稀疏采样、池化和后续读出仍可能改变整网结果；“响应随输入移动”和“最终类别分数不变”是不同的主张，后者要另做检验。\n",
    "\n",
    "采样相位页 固定采样位置 0、2、4、6。交替序列 [0,1,0,1,0,1,0,1] 被采成 [0,0,0,0]，周期平移一格后被采成 [1,1,1,1]。若先两格平均，两种相位都变成 0.5，同时丢掉条纹细节。这个构造例只说明采样相位的风险，分类表现要在真实模型与指定协议上另外检查。\n",
    "\n",
    "平移对照页 读取既有四条件实验，固定数据划分、样本量、验证选择规则和指定的 right_2px 测试。每行应先在同一模型内部比较有无训练增强：\n",
    "\n",
    "| 模型与训练设置 | 原图测试准确率 | 右移 2px 测试准确率 |\n",
    "|---|---:|---:|\n",
    "| MLP，无增强 | 92.11% | 66.39% |\n",
    "| MLP，有增强 | 89.94% | 84.21% |\n",
    "| CNN，无增强 | 95.33% | 84.64% |\n",
    "| CNN，有增强 | 95.81% | 93.84% |\n",
    "\n",
    "本次 CNN 的指定平移测试从 84.64% 升到 93.84%，原图从 95.33% 升到 95.81%；MLP 的原图准确率却下降。证据支持这套协议中的具体比较，不能推广为增强改善所有模型、所有指标，更不能把有无增强的不同检查点混称同一参数快照。未增强 CNN 的原图 9533/10000、右移 8464/10000 已与现有检查点前向复核一致。\n",
    "\n",
    "**先预测与核对**：选出直接支持“CNN 加增强后右移测试改善”的两格，计算提高 9.20 个百分点，再给出不能推出的结论。最后沿 CNN 学习闭环页 复述：局部表示→Flatten 读出→类别损失→各位置反馈→共享权重求和更新。这个失败说明必须在完整网络和固定协议下检查结构主张。第二讲随后比较深度、分支、连接和成本；分布变化与适应的按需补充见第二讲 CLS 08；可迁移表示与学习信号进入第三讲。"
   ],
   "id": "9b610aecad6e47d088a090cae43aa7ef"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "89580af05e804ae9875f420b06075d72",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-15"
   },
   "source": "## 把本段主线连起来",
   "id": "89580af05e804ae9875f420b06075d72"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "a5c8885454bf4e6eb038bc5d35ffe6f5",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-16"
   },
   "source": [
    "**课件与图解**：[本段起点](https://codingai-lec02.pages.dev/#/tiny-cnn-preview/0)；逐页入口见上方各节。\n",
    "\n",
    "请沿同一台 26,698 参数的 Flatten CNN 讲清两个方向。前向：局部乘加、通道汇总、空间采样、展平与类别读出。反向：最终标签形成损失，后层传来 δ，每个共享位置贡献“δ×局部输入”，汇总后才更新。\n",
    "\n",
    "四种证据要分开：窗口小算例核对计算；零目标例核对一次共享更新；差分目标记录检验核学习；既有 MNIST 记录检验分类与指定平移。不要互相替代。完整共享梯度动手见 [LAB 01 · 学会一个卷积核](https://codingai-lec02.pages.dev/course/lab01.html)，真实检查点的响应与遮挡操作见 [LAB 08 · 从真实响应到遮挡对照](https://codingai-lec02.pages.dev/course/lab08.html)。\n",
    "\n",
    "来源沿用主讲：[D2L · 卷积](https://d2l.ai/chapter_convolutional-neural-networks/conv-layer.html)、[D2L · 多通道](https://d2l.ai/chapter_convolutional-neural-networks/channels.html)、[Araujo 等 · 理论感受野](https://distill.pub/2019/computing-receptive-fields/)、[Zhang · 平移与采样](https://proceedings.mlr.press/v97/zhang19a.html)。"
   ],
   "id": "a5c8885454bf4e6eb038bc5d35ffe6f5"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "db76d5ee73514223bd9b75d1e2c0a606",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-17"
   },
   "source": "## 课堂核对：先预测，再运行",
   "id": "db76d5ee73514223bd9b75d1e2c0a606"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "2f06ac8813ce4d60941c75a8fc0d791d",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-18"
   },
   "source": [
    "**课件与图解**：[本段起点](https://codingai-lec02.pages.dev/#/convolution-four-products/0)；逐页入口见上方各节。\n",
    "\n",
    "下面只用标准库核对 2D 乘加、零目标共享梯度与参数账单。先预测每个窗口输出，再写出两条梯度路径。代码对同一旧核收齐梯度后只更新一次；它没有运行完整 CNN 训练，也不会产生新的准确率。"
   ],
   "id": "2f06ac8813ce4d60941c75a8fc0d791d"
  },
  {
   "cell_type": "code",
   "metadata": {
    "deepnote": {
     "block_id": "3d60bf49bebc44ce9e254762f671e664",
     "type": "code",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-19"
   },
   "source": "from math import isclose\n\nimage = [[0, 1, 2], [3, 4, 5], [6, 7, 8]]\nkernel = [[0, 1], [2, 3]]\nout = [[sum(image[r+a][c+b]*kernel[a][b] for a in range(2) for b in range(2))\n        for c in range(2)] for r in range(2)]\nassert out == [[19, 25], [37, 43]]\nprint('二维输出:', out)\n\nX, K, target = [1.0, 2.0, 3.0], [1.0, -1.0], [0.0, 0.0]\ndef forward(k):\n    return [sum(X[i+j]*k[j] for j in range(2)) for i in range(2)]\ndef loss(k):\n    return 0.5*sum((y-t)**2 for y, t in zip(forward(k), target))\ny = forward(K)\ndelta = [a-b for a, b in zip(y, target)]\npaths = [[delta[i]*X[i+j] for j in range(2)] for i in range(2)]\ngrad = [sum(row[j] for row in paths) for j in range(2)]\nassert K == [1, -1] and y == [-1, -1] and grad == [-3, -5]\nassert isclose(loss(K), 1.0)\neta = 0.05\nnew_K = [k-eta*g for k, g in zip(K, grad)]\nassert all(isclose(a, b) for a, b in zip(new_K, [1.15, -.75]))\nassert all(isclose(a, b, abs_tol=1e-12) for a, b in zip(forward(new_K), [-.35, .05]))\nassert isclose(loss(new_K), .0625)\nprint('各窗口梯度贡献:', paths, '总梯度:', grad)\nprint('新核:', new_K, '新输出:', forward(new_K), '新损失:', loss(new_K))\n\ncounts = {'conv1': 8*(1*3*3+1), 'conv2': 16*(8*3*3+1),\n          'linear1': 32*(16*7*7+1), 'linear2': 10*(32+1)}\nassert 16*7*7 == 784 and sum(counts.values()) == 26698\nprint('参数账单:', counts, '总数:', sum(counts.values()))",
   "id": "3d60bf49bebc44ce9e254762f671e664",
   "execution_count": null,
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "48d30d53689a4023be2bd013062f09da",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-20"
   },
   "source": [
    "**课件与图解**：[本段起点](https://codingai-lec02.pages.dev/course/reference/shared-kernel-two-window-gradients.html)；逐页入口见上方各节。\n",
    "\n",
    "**再改一个条件**：仍从旧核 [1,−1] 出发，只改学习率，先预测新损失是否还会下降。请把新输出标为零目标构造例，不覆写 [LAB 01 · 学会一个卷积核](https://codingai-lec02.pages.dev/course/lab01.html) 的差分目标记录。然后不借用 δ=−1 的特例，用符号解释只有类别标签时 δ₀、δ₁ 怎样来到共享核。"
   ],
   "id": "48d30d53689a4023be2bd013062f09da"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "14d15e0337d1415389c774c830f2c5ba",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": "a936def539a64390912a377e38196407"
    },
    "html_anchor": "cell-21"
   },
   "source": "## 架构支撑：BatchNorm 的统计与 affine 参数",
   "id": "14d15e0337d1415389c774c830f2c5ba"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "b898d98b969748588d10d91583525f81",
     "type": "markdown",
     "source_block_id": "c953031239cd43efa604a68a726e0859",
     "logical_key": "bde18a59ffe8413e972dfc78052f8102"
    },
    "html_anchor": "cell-22"
   },
   "source": [
    "对输入 $X\\in\\mathbb R^{N\\times C\\times H\\times W}$，BatchNorm2d 在每个通道内先规范化，再作可学习变换：\n",
    "\n",
    "$$\\hat x=\\frac{x-\\mu_c}{\\sqrt{\\sigma_c^2+\\epsilon}},\\qquad y=\\gamma_c\\hat x+\\beta_c.$$\n",
    "\n",
    "$\\mu_c,\\sigma_c^2$ 是所选统计；$\\gamma_c,\\beta_c$ 是每通道各一个的 affine 参数。使用当前 batch 时，统计汇总同通道的 $N,H,W$ 位置。因此同一张图的响应也会依赖 batch 同伴。一个与主讲一致的精确构造：单通道、每图一个位置，固定当前图的激活 $x=3$；同伴为1时均值2、方差1，同伴为5时均值4、方差仍为1。固定 $\\gamma=1,\\beta=0$，当前图输出便从 $+1/\\sqrt{1+\\epsilon}$ 变为 $-1/\\sqrt{1+\\epsilon}$。[BatchNorm2d 公式、统计维度与 buffers](https://docs.pytorch.org/docs/stable/generated/torch.nn.BatchNorm2d.html)。\n",
    "\n",
    "\n",
    "**运行统计不是可学习参数。** 训练模式下可以估计当前 batch 统计，并把估计积累为 running buffers；推理模式通常读取保存统计。$\\gamma,\\beta$ 由优化器的梯度步骤改变，buffers 的滑动积累不需要参数梯度。因此 requires_grad=False、no_grad、eval 控制不同对象。具体框架配置见 LAB03/LAB05。\n",
    "\n",
    "先在单通道构造中固定当前值3，比较同伴1和5；然后问模型是否采用当前 batch 统计。这个小构造解释依赖关系，不声称真实图像分类已经发生相同变化。"
   ],
   "id": "b898d98b969748588d10d91583525f81"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "8501746b77054f97962af65ccbae55ac",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-23"
   },
   "source": "## 课程导航",
   "id": "8501746b77054f97962af65ccbae55ac"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "81594a2db77e437399a76b55d8349bf3",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-24"
   },
   "source": [
    "[第二讲导读与完整索引](https://codingai-lec02.pages.dev/course/guide.html) · [继续 CLS 04 · 加深与残差](https://codingai-lec02.pages.dev/course/c04.html)\n",
    "\n",
    "[在 Deepnote 返回 CLS 00 · 第二讲导读](https://deepnote.com/workspace/ise3309jingao-5ba66b0a-4283-4f5a-ae06-3f68d3201e99/project/c0b185e7-d7e3-48b6-a605-8f1acdfe243c/notebook/008c71041d2d43569af77c9f25509be8?utm_source=openai&utm_medium=mcp&utm_campaign=openaimcp&utm_content=008c71041d2d43569af77c9f25509be8&utm_term=get_notebook)。静态课件链接返回对应主题；需保留原 Slides 停点时，请在新标签页打开材料。"
   ],
   "id": "81594a2db77e437399a76b55d8349bf3"
  }
 ]
}