introduction_cn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "0b1a3a86",
   "metadata": {},
   "source": [
    "## 1. VSQL 模型简介\n",
    "\n",
    "变分影子量子学习（Variational Shadow Quantum Learning, VSQL）是一个在监督学习框架下的量子–经典混合算法。它使用了参数化量子电路（Parameterized Quantum Circuit, PQC）和经典影子（classical shadow），和通常使用的变分量子算法（Variational Quantum Algorithm, VQA）不同的是，VSQL 只从子空间获取局部特征，而不是从整个希尔伯特空间获取特征。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6aa97879",
   "metadata": {},
   "source": [
    "## 2. 模型原理简介\n",
    "\n",
    "VSQL 的流程图如下：\n",
    "\n",
    "![pipeline](https://ai-studio-static-online.cdn.bcebos.com/36c5194bc48a4af88560172fdd7ec15b8fd3225c27b644acad1981046a48d1ec \"VSQL 流程图\")\n",
    "<div style=\"text-align:center\">VSQL 流程图</div>\n",
    "\n",
    "其训练步骤为：\n",
    "\n",
    "1. 将经典数据 $\\mathbf{x}^i$ 编码到量子态 $\\left|\\mathbf{x}^i\\right>$。\n",
    "2. 准备一个参数化局部量子电路 $U(\\mathbf{\\theta})$ 并且初始化它的参数 $\\mathbf{\\theta}$。\n",
    "3. 在前几个量子比特上作用 $U(\\mathbf{\\theta})$，然后通过测量局部可观测量（比如说泡利 $X\\otimes X\\cdots \\otimes X$ 算符）来获取一个局部影子特征。\n",
    "4. 每次将 $U(\\mathbf{\\theta})$ 向下移动一个量子比特，重复步骤3直到 $U(\\mathbf{\\theta})$ 作用到最后一个量子比特上。\n",
    "5. 将步骤3–4中得到的所有局部影子特征传入经典 FCNN 并通过激活函数得到预测的标签 $\\tilde{\\mathbf{y}}^i$。对于多分类问题来说，我们使用归一化指数函数 (softmax) 作为激活函数。\n",
    "6. 重复步骤3–5直到数据集内所有的数据点都经过了处理。然后计算损失函数 $\\mathcal{L}(\\mathbf{\\theta}, \\mathbf{W}, \\mathbf{b})$。\n",
    "7. 通过梯度下降等优化方法调整参数 $\\mathbf{\\theta}$、$\\mathbf{W}$ 和 $\\mathbf{b}$ 的值，从而最小化损失函数。这样我们就得到了优化后的模型 $\\mathcal{F}$。\n",
    "\n",
    "由于 VSQL 只获取局部影子特征，所以它可以比较容易地在有拓扑连接限制的量子设备上实现。除此之外，因为我们用同一个 $U(\\mathbf{\\theta})$ 来获取整个电路上的局部影子特征，所以需要训练的参数数量相对于通常使用的变分量子分类器来说大大减少。\n",
    "\n",
    "### 2.1 局部影子电路介绍\n",
    "\n",
    "在讲电路的细节之前，我们需要说明几个参数：\n",
    "- $n$：编码后量子态的量子比特数目。\n",
    "- $n_{qsc}$：量子影子电路的宽度。我们每次只在连续 $n_{qsc}$ 个量子比特上作用 $U(\\mathbf{\\theta})$。\n",
    "- $D$：电路的深度，表示 $U(\\mathbf{\\theta})$ 门中某一层电路重复的次数。\n",
    "\n",
    "这里我们给出 $n=4$、$n_{qsc}=2$ 时的一个例子：\n",
    "\n",
    "我们首先在前两个量子比特上作用 $U(\\mathbf{\\theta})$，并且获取第一个影子特征 $O_1$。\n",
    "\n",
    "![qubit0](https://ai-studio-static-online.cdn.bcebos.com/a544360d2b864cd2882c6965bf30e1a0f5fbf089173043cd95a5821b1cdbd799 \"获取第一个影子特征\")\n",
    "<div style=\"text-align:center\">获取第一个影子特征</div>\n",
    "\n",
    "然后我们准备一样的输入态 $\\left|\\mathbf{x}^i\\right>$，在中间两个量子比特上作用 $U(\\mathbf{\\theta})$，得到第二个影子特征 $O_2$。\n",
    "\n",
    "![qubit1](https://ai-studio-static-online.cdn.bcebos.com/9cdb70da3e5f47c2bd089f020975a987e1348d71d724418d890591f298786d4c \"获取第二个影子特征\")\n",
    "<div style=\"text-align:center\">获取第二个影子特征</div>\n",
    "\n",
    "最后，我们再准备一个一样的输入态，在最后两个量子比特上作用 $U(\\mathbf{\\theta})$，得到影子特征 $O_3$。这样我们就处理完了这个数据点！\n",
    "\n",
    "![qubit2](https://ai-studio-static-online.cdn.bcebos.com/6a404995be1141bc9fd6ca36b5e394e00fbe302621c4413bbf86554de6bfc964 \"获取第三个影子特征\")\n",
    "<div style=\"text-align:center\">获取第三个影子特征</div>\n",
    "\n",
    "通常来说，处理一个数据点需要重复以上步骤 $n - n_{qsc} + 1$ 次。有一点需要指出的是，在上面这个例子中我们只使用了一个影子电路，在获取这三个影子特征时我们使用同样的参数 $\\mathbf{\\theta}$。你可以选择增加影子电路的数量来解决更复杂的问题，这里需要注意的是不同影子电路中的参数 $\\mathbf{\\theta}$ 不同。  \n",
    " \n",
    "在后面的 MNIST 二分类任务中，我们将使用2–局部影子电路，即 $n_{qsc}=2$。下图展示了这个影子电路的结构。\n",
    "\n",
    "![2-local](https://ai-studio-static-online.cdn.bcebos.com/0c1035262cb64f61bd3cc87dbf53253aa6a7ecc170634c4db8dd71d576a9409c \"local数为2时的影子电路结构\")\n",
    "<div style=\"text-align:center\">local数为2时的影子电路结构</div>\n",
    "\n",
    "为了增强量子电路的表达能力，我们将重复 $D$ 次虚线框中的结构。$U(\\mathbf{\\theta})$ 的设计并不是唯一的，这里展示的仅仅是一个例子，也可以设计别的电路结构。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2f0070ae",
   "metadata": {},
   "source": [
    "## 3. 模型效果\n",
    "\n",
    "VSQL 在 MNIST 数据集上的二分类效果如下：\n",
    "\n",
    "![binary-classification](https://ai-studio-static-online.cdn.bcebos.com/03c38b174d0e47ae9dbeea6dfda1333d6d3aa74209594324b78c9159b72e7e8a \"二分类效果图\")\n",
    "<div style=\"text-align:center\">二分类效果图</div>\n",
    "\n",
    "VSQL 在 MNIST 数据集上的十分类效果如下：\n",
    "\n",
    "![10-classification](https://ai-studio-static-online.cdn.bcebos.com/54a0273672ad4ec5ba6b7d73b3b225f8d97380c43ceb495a82d16a258117182b \"十分类效果图\")\n",
    "<div style=\"text-align:center\">十分类效果图</div>\n",
    "\n",
    "由表可见，相比于其它的量子神经网络，VSQL 可以使用很少的参数实现更高的分类准确率。而与经典神经网络相比，在十分类任务上，VSQL 也可以使用较少的参数达到相近的效果。当数据量较小时，VSQL 比经典神经网络的效果会更好。"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "17661d06",
   "metadata": {},
   "source": [
    "## 4. 模型如何使用\n",
    "\n",
    "按照如下代码来安装量桨："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "eb7a2be4",
   "metadata": {},
   "outputs": [],
   "source": [
    "# 安装量桨\n",
    "%pip install --user paddle-quantum\n",
    "# 下载预训练模型\n",
    "!wget https://release-data.cdn.bcebos.com/PaddleQuantum/vsql.pdparams -O vsql.pdparams"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9be7e030",
   "metadata": {},
   "source": [
    "成功安装量桨之后，我们来加载 VSQL 模型和要预测的图片："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "3e883450",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import warnings\n",
    "\n",
    "warnings.filterwarnings('ignore')\n",
    "os.environ['PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION'] = 'python'\n",
    "\n",
    "import toml\n",
    "import matplotlib.pyplot as plt\n",
    "from paddle_quantum.qml.vsql import inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "1ac0fbb3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--2023-01-18 15:25:40--  https://ai-studio-static-online.cdn.bcebos.com/088dc9dbabf349c88d029dfd2e07827aa6e41ba958c5434bbd96bc167fc65347\n",
      "Resolving ai-studio-static-online.cdn.bcebos.com (ai-studio-static-online.cdn.bcebos.com)... 112.132.208.35, 116.177.239.35, 119.188.176.35, ...\n",
      "Connecting to ai-studio-static-online.cdn.bcebos.com (ai-studio-static-online.cdn.bcebos.com)|112.132.208.35|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 290 [image/png]\n",
      "Saving to: ‘data_0.png’\n",
      "\n",
      "data_0.png          100%[===================>]     290  --.-KB/s    in 0s      \n",
      "\n",
      "2023-01-18 15:25:40 (5.66 KB/s) - ‘data_0.png’ saved [290/290]\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7fe0b4d39fd0>"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 加载手写数字0\n",
    "!wget https://ai-studio-static-online.cdn.bcebos.com/088dc9dbabf349c88d029dfd2e07827aa6e41ba958c5434bbd96bc167fc65347 -O data_0.png\n",
    "image0 = plt.imread('data_0.png')\n",
    "plt.imshow(image0)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "13aea688",
   "metadata": {},
   "source": [
    "接下来，我们来配置模型参数："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "aedae237",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_toml = r\"\"\"\n",
    "# 模型的整体配置文件。\n",
    "# 输入当前的任务，可以是 'train' 或者 'test'，分别代表训练和预测。这里我们使用 test，表示我们要进行预测。\n",
    "task = 'test'\n",
    "# 要预测的图片的文件路径。\n",
    "image_path = 'data_0.png'\n",
    "# 上面的图片路径是否是文件夹。对于文件夹路径，我们会对文件夹里面的所有图片文件进行预测。这种方式可以一次测试多个图片。\n",
    "is_dir = false\n",
    "# 训练好的模型参数文件的文件路径。\n",
    "model_path = 'vsql.pdparams'\n",
    "# 量子电路所包含的量子比特的数量。\n",
    "num_qubits = 10\n",
    "# 影子电路所包含的量子比特的数量。\n",
    "num_shadow = 2\n",
    "# 电路深度。\n",
    "depth = 1\n",
    "# 我们要预测的类别。这里我们对 0 和 1 进行分类。\n",
    "classes = [0, 1]\n",
    "\"\"\""
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "abf2951f",
   "metadata": {},
   "source": [
    "然后，我们使用 VSQL 模型来进行预测。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "316ff898",
   "metadata": {},
   "outputs": [],
   "source": [
    "config = toml.loads(test_toml)\n",
    "task = config.pop('task')\n",
    "prediction, prob = inference(**config)\n",
    "prob = prob[0]\n",
    "msg = '对于输入的图片，模型有'\n",
    "for idx, item in enumerate(prob):\n",
    "    label = config['classes'][idx]\n",
    "    msg += f'{item:3.2%} 的信心认为它是 {label:d}'\n",
    "    msg += '。' if idx == len(prob) - 1 else '，'\n",
    "print(msg)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cbd83c5b",
   "metadata": {},
   "source": [
    "接下来，我们来测试另外一个图片："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "fb4c9ced",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--2023-01-18 15:25:46--  https://ai-studio-static-online.cdn.bcebos.com/c755f723af3d4a1c8f113f8ac3bd365406decd1be70944b7b7b9d41413e8bc7a\n",
      "Resolving ai-studio-static-online.cdn.bcebos.com (ai-studio-static-online.cdn.bcebos.com)... 116.95.27.35, 116.177.239.35, 119.188.176.35, ...\n",
      "Connecting to ai-studio-static-online.cdn.bcebos.com (ai-studio-static-online.cdn.bcebos.com)|116.95.27.35|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 173 [image/png]\n",
      "Saving to: ‘data_1.png’\n",
      "\n",
      "data_1.png          100%[===================>]     173  --.-KB/s    in 0s      \n",
      "\n",
      "2023-01-18 15:25:46 (82.5 MB/s) - ‘data_1.png’ saved [173/173]\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7fe0b4f24eb0>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 加载手写数字1\n",
    "!wget https://ai-studio-static-online.cdn.bcebos.com/c755f723af3d4a1c8f113f8ac3bd365406decd1be70944b7b7b9d41413e8bc7a -O data_1.png\n",
    "image1 = plt.imread('data_1.png')\n",
    "plt.imshow(image1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "ecb2ba3d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "对于输入的图片，模型有18.29% 的信心认为它是 0，81.71% 的信心认为它是 1。\n"
     ]
    }
   ],
   "source": [
    "test_toml = r\"\"\"\n",
    "# 模型的整体配置文件。\n",
    "# 输入当前的任务，可以是 'train' 或者 'test'，分别代表训练和预测。这里我们使用 test，表示我们要进行预测。\n",
    "task = 'test'\n",
    "# 要预测的图片的文件路径。\n",
    "image_path = 'data_1.png'\n",
    "# 上面的图片路径是否是文件夹。对于文件夹路径，我们会对文件夹里面的所有图片文件进行预测。这种方式可以一次测试多个图片。\n",
    "is_dir = false\n",
    "# 训练好的模型参数文件的文件路径。\n",
    "model_path = 'vsql.pdparams'\n",
    "# 量子电路所包含的量子比特的数量。\n",
    "num_qubits = 10\n",
    "# 影子电路所包含的量子比特的数量。\n",
    "num_shadow = 2\n",
    "# 电路深度。\n",
    "depth = 1\n",
    "# 我们要预测的类别。这里我们对 0 和 1 进行分类。\n",
    "classes = [0, 1]\n",
    "\"\"\"\n",
    "\n",
    "config = toml.loads(test_toml)\n",
    "task = config.pop('task')\n",
    "# 代码还需要修改\n",
    "prediction, prob = inference(**config)\n",
    "if config['is_dir']:\n",
    "    print(f\"对输入图片的预测结果分别是 {str(prediction)[1:-1]}。\")\n",
    "else:\n",
    "    prob = prob[0]\n",
    "    msg = '对于输入的图片，模型有'\n",
    "    for idx, item in enumerate(prob):\n",
    "        label = config['classes'][idx]\n",
    "        msg += f'{item:3.2%} 的信心认为它是 {label:d}'\n",
    "        msg += '。' if idx == len(prob) - 1 else '，'\n",
    "    print(msg)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8f6f3b91",
   "metadata": {},
   "source": [
    "## 5. 注意事项\n",
    "\n",
    "我们提供的模型为二分类模型，仅可以用来分辨手写数字0和1。对于其它分类任务，需要重新进行训练。\n",
    "\n",
    "更详细的使用介绍可以参考：https://github.com/PaddlePaddle/Quantum/blob/master/applications/handwritten_digits_classification/introduction_cn.ipynb\n",
    "\n",
    "VSQL 模型的具体介绍可以参考：https://github.com/PaddlePaddle/Quantum/blob/master/tutorials/machine_learning/VSQL_CN.ipynb"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4857182b",
   "metadata": {},
   "source": [
    "## 6. 相关论文以及引用信息\n",
    "\n",
    "```\n",
    "@inproceedings{li2021vsql,\n",
    "  title={VSQL: Variational shadow quantum learning for classification},\n",
    "  author={Li, Guangxi and Song, Zhixin and Wang, Xin},\n",
    "  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},\n",
    "  volume={35},\n",
    "  number={9},\n",
    "  pages={8357--8365},\n",
    "  year={2021}\n",
    "}\n",
    "```"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pq-dev",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.15"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "vscode": {
   "interpreter": {
    "hash": "5fea01cac43c34394d065c23bb8c1e536fdb97a765a18633fd0c4eb359001810"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}