{
 "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": "52d5359cf8504f069e29cebf107c5674",
   "destination": "L2",
   "cloud_run": false
  },
  "title": "CLS 04 · 为什么还要加深：训练起点、规模与残差"
 },
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "bd91f4d57622486fbb6e6f1532efa19d",
     "type": "text-cell-h1",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-1"
   },
   "source": "# CLS 04 · 为什么还要加深：训练起点、规模与残差",
   "id": "bd91f4d57622486fbb6e6f1532efa19d"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "6a2c9dea7d234508842a1f2ca7e11130",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-2"
   },
   "source": [
    "[返回第二讲 · 加深、退化与残差](https://codingai-lec02.pages.dev/#/deep-network-training-difficulty/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",
    "本册承接第二讲 C 段：能写出更深的网络，为什么还不等于训练得好？先区分表达、优化和泛化，再看恒等旁路如何改变参数化，并用同深度比较判断证据。\n",
    "\n",
    "**课堂后先读**：第 4 节找训练退化现象，第 5 节沿残差的两条路径手算，第 6 节读 ImageNet 配对验证表。第 1–3 节补齐 LeNet、系统规模和 AlexNet 的历史条件；末尾短代码核对残差数值与百分点差。\n",
    "\n",
    "**结构练习**：进入 [LAB 03 · 架构实验室](https://codingai-lec02.pages.dev/course/lab03.html)，找分支／残差相关小节。先检查相加接口，再提出一个只改变连接方式的比较；写清实际运行了什么，不能用机制算例替代模型训练结果。"
   ],
   "id": "6a2c9dea7d234508842a1f2ca7e11130"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "43470c5bfc4f421192bb21153f8ba0f0",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-3"
   },
   "source": "## 1. 卷积已经能学，新的困难为什么出现？",
   "id": "43470c5bfc4f421192bb21153f8ba0f0"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "041ef403e63040cb8ead2b72ec66393c",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-4"
   },
   "source": [
    "**课件与图解**：[LeNet说明：端到端卷积学习早已存在](https://codingai-lec02.pages.dev/#/convolution-before-imagenet/0) · [可以写出更深网络，不代表它容易训练](https://codingai-lec02.pages.dev/#/deep-network-training-difficulty/0)\n",
    "\n",
    "LeNet 已把局部共享、下采样与监督训练用于字符识别，因此 2012 年的转折不是第一次发明卷积或端到端学习。自然图像有更多类别、尺度、视角、背景和更大数据规模，需要把这条路线扩大。\n",
    "\n",
    "更多层可能组合出更复杂的表示，但能写出函数，不等于从随机起点就容易训练出有用参数。反馈沿多层局部导数传播，初始化、激活尺度与结构都会影响有效调整。观察到深层网络表现不好时，要先分开“函数族能否表达”“训练是否找到好参数”“新样本是否泛化”，不能仅用层数替代诊断。\n",
    "\n",
    "**先预测与核对**：分别给“可以表示某个解”和“这次训练找到这个解”写一句需要的证据。下一节先固定端到端监督目标，分清表达能力、优化与泛化；随后看扩大数据和计算怎样支持这条训练链。"
   ],
   "id": "041ef403e63040cb8ead2b72ec66393c"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "42e63a5b3e7b4026ae4cf89f96490bec",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-5"
   },
   "source": "## 2. 能写出计算图，还要检查训练能否找到解",
   "id": "42e63a5b3e7b4026ae4cf89f96490bec"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "230b9665660e4e81b0c7f70e4bdf909a",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-6"
   },
   "source": "增加深度先扩大了可表达的函数关系，但不保证同一个优化流程能找到已有的好解。如果浅网络已能实现某个映射，给更深网络加恒等层可以保留该映射；现实训练是否找到这样的参数，要用训练曲线检查，不能只凭包含关系证明。\n\n面对性能下降，分三步判断：模型类是否含有需要的函数；当前训练是否拟合好训练集；拟合的函数是否在留出样本泛化。后面的 ResNet 证据将分别使用训练退化观察和同深度验证比较。\n\n2006 年 DBN 的逐层生成学习完整解释已归 [历史参考](https://codingai-lec02.pages.dev/course/reference.html#history)；它与这里的监督分类目标不同，不作为继续学习残差的前置。",
   "id": "230b9665660e4e81b0c7f70e4bdf909a"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "251ac650702944f79e8f963b75b1ae98",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-7"
   },
   "source": "## 3. AlexNet：系统条件与结果必须一起读",
   "id": "251ac650702944f79e8f963b75b1ae98"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "ef9b448188bc4b8da6fba3568b670665",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-8"
   },
   "source": [
    "**课件与图解**：[AlexNet把数据、算力与训练方法组织成一个可扩展系统](https://codingai-lec02.pages.dev/#/alexnet-system/0) · [2012年的结果显示了路线转折，也要保留比较口径](https://codingai-lec02.pages.dev/#/alexnet-2012-evidence/0)\n",
    "\n",
    "AlexNet 把大量类别标注、GPU 张量计算、非线性、增强、正则与优化组织在一起，使更大规模的监督表示学习可行。把结果只归因于“层数增加”会遗漏这些同时存在的条件；系统成功也不是对单一因素的独立消融。\n",
    "\n",
    "主讲沿原论文 Table 2 使用同年 2012 竞赛分类任务的 **test top-5 错误率**，保留一位小数：\n",
    "\n",
    "| 比较对象 | test top-5 error | 必须保留的口径 |\n",
    "|---|---:|---|\n",
    "| 最佳非 CNN 系统 | 约 26.2% | 同年任务参照 |\n",
    "| 仅比赛训练数据的 CNN 集成 | 约 16.4% | 集成，不是单个网络 |\n",
    "| 含额外 ImageNet 监督预训练的集成 | 约 15.3% | 训练数据条件不同 |\n",
    "\n",
    "top-5 错误判断真实类别是否未进入前五个预测，不能写成 top-1 准确率。16.4% 与 15.3% 不是“同一个单网络又多训练一点”的直接对照，后者还涉及额外数据；这里按表中显示精度保留一位小数。\n",
    "\n",
    "**先预测与核对**：为每个百分比补上数据、模型数量与指标三个标签，再说一句这张表不能隔离的因素。规模化路线取得进展以后，新的问题出现：更深的普通网络是否至少能把训练集拟合得更好？"
   ],
   "id": "ef9b448188bc4b8da6fba3568b670665"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "3f61dbcd905347fb942fc8f1bbf0a11e",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-9"
   },
   "source": "## 4. 训练本身退化，为什么不能只说过拟合？",
   "id": "3f61dbcd905347fb942fc8f1bbf0a11e"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "80a9a42924224276b232b894048c1af3",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-10"
   },
   "source": [
    "**课件与图解**：[更深的普通网络，为什么连训练误差也更高？](https://codingai-lec02.pages.dev/#/resnet-degradation/0) · [存在一个好解，不等于训练过程容易找到它](https://codingai-lec02.pages.dev/#/depth-can-preserve-a-solution/0)\n",
    "\n",
    "**先只看左图。** 同一 CIFAR-10 任务中，56 层普通网络的训练误差比 20 层更高。这还能只解释成“训练集拟合太好”吗？\n",
    "\n",
    "![ResNet原论文Figure1：CIFAR10普通20层与56层网络的训练误差和测试误差](data:image/webp;base64,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)\n",
    "\n",
    "*复用 Slides 已保存的 ResNet 原论文 Figure 1。左侧为训练误差，右侧为测试误差；横轴迭代单位为 10⁴。这里展示原图，未生成或重跑曲线。*\n",
    "\n",
    "**先排除一个不完整解释。** 56 层连训练集本身也拟合得更差，因此仅说“过拟合”不够。这指向优化困难，但也不能凭这一张图断言原因就是梯度消失；原研究已使用归一化等措施，并讨论了这个简单解释的不足。\n",
    "\n",
    "**再提出一个可以构造的问题。** 若新增层能保持输入表示，较深模型原则上就能包含浅层模型的行为：浅层做原来的事，新增部分先什么也不改。可是存在这个候选解，不代表随机起点和训练容易找到它。\n",
    "\n",
    "于是设计问题变得具体：能否让“先保留已有表示”对应一组更直接的参数选择，让新增部分只学习需要改变的内容？\n",
    "\n",
    "**先预测与核对**：各用一句话区分“更深模型能表达好解”和“这次训练找到好解”。图上哪条曲线说明后者遇到了困难？下一节直接写出保持路径。"
   ],
   "id": "80a9a42924224276b232b894048c1af3"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "51e58c2a5a254e83bb9bc9d551850a89",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-11"
   },
   "source": "## 5. 先保留、再修正；反馈也逐路径相加",
   "id": "51e58c2a5a254e83bb9bc9d551850a89"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "d8148a2cd3fc499996ceff3ab4f25ae5",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-12"
   },
   "source": [
    "**课件与图解**：[先保留已有表示，再学习需要改变的部分](https://codingai-lec02.pages.dev/#/residual-preserve-and-correct/0) · [残差相加以后，反馈也有一条直接路径](https://codingai-lec02.pages.dev/course/reference/residual-two-gradient-paths.html)\n",
    "\n",
    "**先读结构，再代入数值。** 保持路径把 h 送到相加处，分支给出修正 F(h)。不要把 F(h) 当成另一个完整输出 y。\n",
    "\n",
    "![残差核心y=h+F(h)的保持路径和分支；同一h=(2,1,-1)下零修正和非零修正的完整输出](data:image/webp;base64,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)\n",
    "\n",
    "这里两路形状相同，且**停在相加节点，没有相加后激活**。因此 F(h)=0 会严格保留 h；修正为 (0.2,−0.3,0.1) 时，逐项得到 (2.2,0.7,−0.9)。修正可以正也可以负，也不要求永远很小。形状不同需先匹配。\n",
    "\n",
    "**沿同一结构反向走一次。** 为方便手算，换成标量 x=2、F(x)=0.2x，输出 y=2.4。若上游 ∂L/∂y=1：\n",
    "\n",
    "| 路径 | 局部导数 | 给 x 的反馈 |\n",
    "|---|---:|---:|\n",
    "| 直达 x | 1 | 1 |\n",
    "| 分支 F(x)=0.2x | 0.2 | 0.2 |\n",
    "| 汇总 | 1+0.2 | 1.2 |\n",
    "\n",
    "这仍是上一册的原则：每条路径先算贡献，再相加。它不是梯度永不消失的保证——若 F(x)=−x，总反馈为 1−1=0；若相加后接 ReLU，还需乘该点的 ReLU 导数。向量例中的 −1 在相加后 ReLU 下也会被截断。\n",
    "\n",
    "**先预测与核对**：遮住图右侧输出，先算零修正与非零修正，再算两种标量分支的总反馈。局部机制给出了预期；要判断连接方案是否帮助实际训练，还得读同深度、同协议的结果。"
   ],
   "id": "d8148a2cd3fc499996ceff3ab4f25ae5"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "a456ce1ecf81468fa4cb3f45c5093ca4",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-13"
   },
   "source": "## 6. 同深度配对，验证表不能冒充训练曲线",
   "id": "a456ce1ecf81468fa4cb3f45c5093ca4"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "f14de7a240b448b7938b75a2081dcb45",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-14"
   },
   "source": [
    "**课件与图解**：[连接改了，怎样检验它是否有用？](https://codingai-lec02.pages.dev/#/residual-easier-to-optimize/0)\n",
    "\n",
    "先固定数据与评价协议，再横向比较同深度 plain 与 residual。以下是 ResNet 原论文 Table 2：**ImageNet validation top-1 error，10-crop，越低越好**。\n",
    "\n",
    "| 深度 | plain | residual | 错误率减少（百分点） |\n",
    "|---|---:|---:|---:|\n",
    "| 18 层 | 27.94% | 27.88% | 0.06 |\n",
    "| 34 层 | 28.54% | 25.03% | 3.51 |\n",
    "\n",
    "先遮住最后一列，用同一行的两个数相减。34 层的差是 3.51 **个百分点**，不是相对减少 3.51%；它支持这套设置中残差连接相对 plain 的验证表现改善。表中 18 层差别小，提醒我们结构效果要结合具体设置读取。\n",
    "\n",
    "**这张表回答不了什么？** 它没有给出训练拟合降低了多少。前节 CIFAR-10 Figure 1 的 20／56 层训练退化，与这里 ImageNet 18／34 层验证比较不是同一组实验轨迹。若要继续追问优化过程，应另读原论文 ImageNet Figure 4 或 CIFAR-10 Figure 6 的训练／测试曲线；跨深度也不能宣称计算量相同。\n",
    "\n",
    "**把证据写成一句话**：在 ImageNet、10-crop 评价下，同为 34 层，改变连接方式后验证错误率从 28.54% 降到 25.03%；解释训练机制还需要训练行为的证据。然后改写一次：只看 18 层行时，你的结论力度应怎样变化？\n",
    "\n",
    "下一步在 [DenseNet 对比页](https://codingai-lec02.pages.dev/#/frontier-dense-feature-reuse/0) 比较相加与拼接：前者合并同形数值，后者保留来源的通道组，也带来通道增长与激活存储。"
   ],
   "id": "f14de7a240b448b7938b75a2081dcb45"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "264ca6124c984ae19a6497e7f94aea9f",
     "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": "264ca6124c984ae19a6497e7f94aea9f"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "0d3c15d6496b46c19cdad9d1843095f1",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-16"
   },
   "source": "本册的连续问题是：端到端监督卷积学习怎样扩大？增加深度为何会连训练拟合都变差？恒等路径怎样改变参数化？同深度配对是否支持连接方案？旧表示相加与通道拼接怎样改变访问和存储？\n\nAlexNet 的 test top-5、ResNet 的训练观察与 validation top-1 各自标明数据和协议。构造残差向量与两路导数仅验证局部计算。DBN 的生成目标在历史参考独立保存，不混作监督 CNN 的训练步骤。",
   "id": "0d3c15d6496b46c19cdad9d1843095f1"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "05ef58e7676b44659ee68c6fd1c26d77",
     "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": "05ef58e7676b44659ee68c6fd1c26d77"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "5a970917928345a5b8bb98c4f936c3d3",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-18"
   },
   "source": [
    "**课件与图解**：[先保留已有表示，再学习需要改变的部分](https://codingai-lec02.pages.dev/#/residual-preserve-and-correct/0) · [残差相加以后，反馈也有一条直接路径](https://codingai-lec02.pages.dev/course/reference/residual-two-gradient-paths.html) · [连接改了，怎样检验它是否有用？](https://codingai-lec02.pages.dev/#/residual-easier-to-optimize/0)\n",
    "\n",
    "以下核对只用标准库。先算向量修正与两条反馈，再读取明确标注的文献验证数值，计算百分点差。代码不会训练 ImageNet 模型，表中数字是原论文报告值，不是此代码测出的性能。"
   ],
   "id": "5a970917928345a5b8bb98c4f936c3d3"
  },
  {
   "cell_type": "code",
   "metadata": {
    "deepnote": {
     "block_id": "1f90f9405ffa4b51994883e03998d3b7",
     "type": "code",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-19"
   },
   "source": "from math import isclose\n\nh = (2.0, 1.0, -1.0)\nzero = (0.0, 0.0, 0.0)\ncorrection = (0.2, -0.3, 0.1)\nadd = lambda a, b: tuple(x+y for x, y in zip(a, b))\nassert add(h, zero) == h\nassert all(isclose(a, b) for a, b in zip(add(h, correction), (2.2, .7, -.9)))\nprint('零修正:', add(h, zero), '非零修正:', add(h, correction))\nprint('若相加后加ReLU:', tuple(max(0, v) for v in add(h, zero)))\n\nx, upstream = 2.0, 1.0\nfor slope in [0.2, -1.0]:\n    y = x+slope*x\n    direct, branch = upstream, upstream*slope\n    total = direct+branch\n    print('F斜率=', slope, 'y=', y, '两路贡献=', (direct, branch), '总反馈=', total)\nassert isclose(upstream*(1+.2), 1.2)\nassert upstream*(1-1) == 0\n\n# 文献报告：ImageNet validation top-1 error, 10-crop；不是本代码训练输出。\nreported_validation_error = {18: {'plain': 27.94, 'residual': 27.88},\n                             34: {'plain': 28.54, 'residual': 25.03}}\nfor depth, row in reported_validation_error.items():\n    print(depth, '层验证错误率减少（百分点）:', round(row['plain']-row['residual'], 2))\nassert isclose(reported_validation_error[34]['plain']-reported_validation_error[34]['residual'], 3.51)\nprint('此表不能代替训练曲线。')",
   "id": "1f90f9405ffa4b51994883e03998d3b7",
   "execution_count": null,
   "outputs": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "aa0544912dcb473fb3afd513f7080e0c",
     "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/#/resnet-degradation/0)；逐页入口见上方各节。\n",
    "\n",
    "**运行后完成一句证据句**：“在 ____ 数据、____ 指标、____ 评价设置下，同为 ____ 层时，改变 ____ 后观察到 ____；这还不能直接说明 ____。”然后用训练误差退化例检验自己的解释是否把优化和泛化混在一起。"
   ],
   "id": "aa0544912dcb473fb3afd513f7080e0c"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "452d88c593614c068f3b1af0cb08d11b",
     "type": "text-cell-h2",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-21"
   },
   "source": "## 课程导航",
   "id": "452d88c593614c068f3b1af0cb08d11b"
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "deepnote": {
     "block_id": "fce362c6a5cd4cee968fe1843e534f7a",
     "type": "markdown",
     "source_block_id": null,
     "logical_key": null
    },
    "html_anchor": "cell-22"
   },
   "source": [
    "[第二讲导读与完整索引](https://codingai-lec02.pages.dev/course/guide.html) · [继续 CLS 06 · 现代骨干与公平比较](https://codingai-lec02.pages.dev/course/c07.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": "fce362c6a5cd4cee968fe1843e534f7a"
  }
 ]
}