{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"# 量子神经网络初体验\n",
"\n",
"[](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/r2.5.0/mindquantum/zh_cn/advanced/mindspore_initial_experience_of_quantum_neural_network.ipynb) \n",
"[](https://mindspore-website.obs.cn-north-4.myhuaweicloud.com/notebook/r2.5.0/mindquantum/zh_cn/advanced/mindspore_initial_experience_of_quantum_neural_network.py) \n",
"[](https://gitee.com/mindspore/docs/blob/r2.5.0/docs/mindquantum/docs/source_zh_cn/advanced/initial_experience_of_quantum_neural_network.ipynb)\n",
"\n",
"## 量子神经网络的结构\n",
"\n",
"在MindSpore Quantum中，量子神经网络（Quantum Neural Network, QNN）的结构如下图所示，其通常由三部分构成：\n",
"\n",
"（1）一个（或多个）编码线路，用于将经典数据编码到量子数据（通常称为Encoder）；\n",
"\n",
"（2）一个（或多个）训练线路，用于训练带参量子门中的参数（通常称为Ansatz）；\n",
"\n",
"（3）一个（或多个）测量，用于检测测量值（例如在`Z`方向上测量，就是某个量子比特的量子态在`Z`轴上的投影，该测量得到的是量子态关于泡利`Z`算符（不限定于泡利`Z`算符，换成其他的算符亦可）的期望值）是否接近于目标期望值。\n",
"\n",
"\n",
"\n",
"下面，我们通过一个简单的例子来体验一下如何使用MindSpore Quantum。\n",
"\n",
"## 简单的例子\n",
"\n",
"\n",
"\n",
"我们搭建如上图所示的量子神经网络，其中Encoder由一个 [H](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.HGate.html) 门，1个 [RX](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.RX.html) 门、1个 [RY](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.RY.html) 门和1个 [RZ](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.RZ.html) 门构成，Ansatz由1个 [RX](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.RX.html) 门和1个 [RY](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.RY.html) 门构成，测量则是作用在第0位量子比特上的泡利`Z`算符。\n",
"\n",
"问题描述：我们将Encoder看成是系统对初始量子态的误差影响（参数$\\alpha_0, \\alpha_1$和$\\alpha_2$是将原经典数据经过预处理（可选）后得到的某个固定值，即为已知值，在此分别设为0.2, 0.3和0.4）。我们需要训练一个Ansatz来抵消掉这个误差，使得最后的量子态还是处于$|0\\rangle$态。\n",
"\n",
"思路：对末态执行泡利`Z`算符测量，此时的测量值就是此时的量子态关于泡利`Z`算符的期望值。由于$|0\\rangle$是算符`Z`的本征态，且本征值为1，容易知道\n",
"\n",
"$$\n",
"\\langle 0|Z|0\\rangle=1.\n",
"$$\n",
"\n",
"也就是说，目标期望值为1。可以通过测量得到的期望值来验证此时的状态是否为$|0\\rangle$。\n",
"\n",
"解决方案：通过训练Ansatz中的参数，希望测量值接近于目标期望值，换句话说，我们只需让测量值尽可能接近于$|0\\rangle$态关于泡利`Z`算符对应的期望值，那么此时的状态就是$|0\\rangle$，即Ansatz抵消了Encoder对初始量子态产生的误差。\n",
"\n",
"## 环境准备\n",
"\n",
"导入本教程所依赖的模块"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {},
"outputs": [],
"source": [
"import numpy as np # 导入numpy库并简写为np\n",
"from mindquantum.core.circuit import Circuit # 导入Circuit模块，用于搭建量子线路\n",
"from mindquantum.core.gates import H, RX, RY, RZ # 导入量子门H, RX, RY, RZ\n",
"from mindquantum.core.parameterresolver import PRGenerator # 导入参数生成器模块"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 搭建Encoder\n",
"\n",
"根据图示的量子线路图，我们可以在MindSpore Quantum中搭建Encoder。"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"tags": []
},
"outputs": [
{
"data": {
"text/html": [
"<pre style=\"white-space:pre;overflow-x:auto;line-height:normal;font-family:Menlo,'DejaVu Sans Mono',consolas,'Courier New',monospace\"><span style=\"color: #ff0000; text-decoration-color: #ff0000; font-weight: bold\"> Circuit Summary </span>\n",
"╭──────────────────────┬────────────────────────╮\n",
"│<span style=\"font-weight: bold\"> </span><span style=\"color: #3b3b95; text-decoration-color: #3b3b95; font-weight: bold\">Info</span><span style=\"font-weight: bold\"> </span>│<span style=\"font-weight: bold\"> </span><span style=\"color: #3b3b95; text-decoration-color: #3b3b95; font-weight: bold\">value</span><span style=\"font-weight: bold\"> </span>│\n",
"├──────────────────────┼────────────────────────┤\n",
"│ <span style=\"font-weight: bold\">Number of qubit</span> │ 1 │\n",
"├──────────────────────┼────────────────────────┤\n",
"│ <span style=\"font-weight: bold\">Total number of gate</span> │ 4 │\n",
"│ Barrier │ 0 │\n",
"│ Noise Channel │ 0 │\n",
"│ Measurement │ 0 │\n",
"├──────────────────────┼────────────────────────┤\n",
"│ <span style=\"font-weight: bold\">Parameter gate</span> │ 3 │\n",
"│ 3 encoder parameters │ <span style=\"color: #ffc857; text-decoration-color: #ffc857\">alpha0, alpha1, alpha2</span> │\n",
"╰──────────────────────┴────────────────────────╯\n",
"</pre>\n"
],
"text/plain": [
"\u001b[1;38;2;255;0;0m Circuit Summary \u001b[0m\n",
"╭──────────────────────┬────────────────────────╮\n",
"│\u001b[1m \u001b[0m\u001b[1;38;2;59;59;149mInfo\u001b[0m\u001b[1m \u001b[0m\u001b[1m \u001b[0m│\u001b[1m \u001b[0m\u001b[1;38;2;59;59;149mvalue\u001b[0m\u001b[1m \u001b[0m\u001b[1m \u001b[0m│\n",
"├──────────────────────┼────────────────────────┤\n",
"│ \u001b[1mNumber of qubit\u001b[0m │ 1 │\n",
"├──────────────────────┼────────────────────────┤\n",
"│ \u001b[1mTotal number of gate\u001b[0m │ 4 │\n",
"│ Barrier │ 0 │\n",
"│ Noise Channel │ 0 │\n",
"│ Measurement │ 0 │\n",
"├──────────────────────┼────────────────────────┤\n",
"│ \u001b[1mParameter gate\u001b[0m │ 3 │\n",
"│ 3 encoder parameters │ \u001b[38;2;255;200;87malpha0, alpha1, alpha2\u001b[0m │\n",
"╰──────────────────────┴────────────────────────╯\n"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"image/svg+xml": [
"<svg xmlns=\"http://www.w3.org/2000/svg\" width=\"456.8\" height=\"80.0\" xmlns:xlink=\"http://www.w3.org/1999/xlink\"><rect x=\"0\" y=\"0.0\" width=\"456.8\" height=\"80.0\" fill=\"#ffffff\" /><text x=\"20.0\" y=\"40.0\" font-size=\"16px\" dominant-baseline=\"middle\" text-anchor=\"start\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#252b3a\" >q0: </text><line x1=\"48.8\" x2=\"436.8\" y1=\"40.0\" y2=\"40.0\" stroke=\"#adb0b8\" stroke-width=\"1\" /><rect x=\"72.8\" y=\"20.0\" width=\"40.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#5e7ce0\" fill-opacity=\"1\" /><text x=\"92.8\" y=\"40.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >H </text><rect x=\"132.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"172.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RX </text><text x=\"172.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >alpha0 </text><rect x=\"232.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"272.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RY </text><text x=\"272.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >alpha1 </text><rect x=\"332.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"372.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RZ </text><text x=\"372.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >alpha2 </text></svg>"
],
"text/plain": [
"<mindquantum.io.display.circuit_svg_drawer.SVGCircuit at 0x7f1f78ae3eb0>"
]
},
"execution_count": 2,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# pylint: disable=W0104\n",
"alpha = PRGenerator('alpha') # 初始化参数生成器\n",
"encoder = Circuit() # 初始化量子线路\n",
"encoder += H.on(0) # H门作用在第0位量子比特\n",
"encoder += RX(alpha.new()).on(0) # RX(alpha_0)门作用在第0位量子比特\n",
"encoder += RY(alpha.new()).on(0) # RY(alpha_1)门作用在第0位量子比特\n",
"encoder += RZ(alpha.new()).on(0) # RZ(alpha_2)门作用在第0位量子比特\n",
"encoder = encoder.no_grad() # Encoder作为整个量子神经网络的第一层，不用对编码线路中的梯度求导数，因此加入no_grad()\n",
"encoder.as_encoder() # 将量子线路设置为encoder线路\n",
"encoder.summary() # 总结Encoder\n",
"encoder.svg()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从对Encoder的Summary中可以看到，该量子线路由4个量子门组成，其中有3个含参量子门且参数为$\\alpha_0,\\alpha_1,\\alpha_2$，该量子线路调控的量子比特数为1。\n",
"\n",
"然后，我们需要对Encoder中的参数进行赋值。由于Encoder中的参数$\\alpha_0, \\alpha_1$和$\\alpha_2$分别为已知值0.2、0.3和0.4，因此可以直接对参数进行赋值，并打印此时的状态。"
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"(0.5669903122552596-0.1753906567580312j)¦0⟩\n",
"(0.800814626197614+0.08034947292077024j)¦1⟩\n"
]
}
],
"source": [
"alpha0, alpha1, alpha2 = 0.2, 0.3, 0.4 # alpha0、alpha1、alpha2为已知的固定值，分别赋值0.2、0.3 和0.4\n",
"state = encoder.get_qs(pr={'alpha0': alpha0, 'alpha1': alpha1, 'alpha2': alpha2}, ket=True)\n",
"print(state)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"上述步骤为了展示MindSpore Quantum可以演化量子线路（若量子线路中的量子门带参数，则需要对参数赋值）并得到演化后的末态。从上述打印可以看到，演化后得到的末态为$|0\\rangle$和$|1\\rangle$组成的叠加态，各项对应的振幅为上述打印的状态左边对应的数值。\n",
"\n",
"说明：\n",
"\n",
"（1）通过调用量子线路的 [get_qs](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/circuit/mindquantum.core.circuit.Circuit.html#mindquantum.core.circuit.Circuit.get_qs) 函数，我们能够得到该量子线路在全零态基础上演化出来的量子态。\n",
"\n",
"（2）[get_qs](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/circuit/mindquantum.core.circuit.Circuit.html#mindquantum.core.circuit.Circuit.get_qs) 的 `pr` 参数代表变分量子线路中的参数值，`ket` 表示是否将量子态输出为右矢形式。\n",
"\n",
"## 搭建Ansatz\n",
"\n",
"同样地，我们也可以在MindSpore Quantum中搭建Ansatz。"
]
},
{
"cell_type": "code",
"execution_count": 4,
"metadata": {
"tags": []
},
"outputs": [
{
"data": {
"image/svg+xml": [
"<svg xmlns=\"http://www.w3.org/2000/svg\" width=\"296.8\" height=\"80.0\" xmlns:xlink=\"http://www.w3.org/1999/xlink\"><rect x=\"0\" y=\"0.0\" width=\"296.8\" height=\"80.0\" fill=\"#ffffff\" /><text x=\"20.0\" y=\"40.0\" font-size=\"16px\" dominant-baseline=\"middle\" text-anchor=\"start\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#252b3a\" >q0: </text><line x1=\"48.8\" x2=\"276.8\" y1=\"40.0\" y2=\"40.0\" stroke=\"#adb0b8\" stroke-width=\"1\" /><rect x=\"72.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"112.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RX </text><text x=\"112.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >theta0 </text><rect x=\"172.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"212.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RY </text><text x=\"212.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >theta1 </text></svg>"
],
"text/plain": [
"<mindquantum.io.display.circuit_svg_drawer.SVGCircuit at 0x7f1f78a87ca0>"
]
},
"execution_count": 4,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# pylint: disable=W0104\n",
"theta = PRGenerator('theta') # 初始化参数生成器\n",
"ansatz = Circuit() # 初始化量子线路\n",
"ansatz += RX(theta.new()).on(0) # RX(theta_0)门作用在第0位量子比特\n",
"ansatz += RY(theta.new()).on(0) # RY(theta_1)门作用在第0位量子比特\n",
"ansatz.as_ansatz() # 将量子线路设置成待训练线路\n",
"ansatz.svg() # 打印量子线路"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从对Ansatz的Summary中可以看到，该量子线路由2个量子门组成，其中有2个含参量子门且参数为$\\theta_0, \\theta_1$，该量子线路调控的量子比特数为1。\n",
"\n",
"然后，对Ansatz中的参数进行赋值。由于Ansatz为需要训练的量子线路，因此Ansatz中的参数$\\theta_0$和$\\theta_1$可以随机设定，通常默认设为初始值0。我们同样可以打印此时的量子态，不过这并不是必要的步骤，只是为了再次熟悉一下 [get_qs](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/circuit/mindquantum.core.circuit.Circuit.html#mindquantum.core.circuit.Circuit.get_qs) 函数。"
]
},
{
"cell_type": "code",
"execution_count": 5,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"1¦0⟩\n"
]
}
],
"source": [
"theta0, theta1 = 0, 0 # 对theta0、theta1进行赋值，设为初始值0、0\n",
"state = ansatz.get_qs(pr=dict(zip(ansatz.params_name, [theta0, theta1])), ket=True)\n",
"print(state)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从上述打印可以看到，此时的状态为$|0\\rangle$且振幅为1。这是因为对于Ansatz来说，默认的输入量子态为$|0\\rangle$，而且其中的参数$\\theta_0$和$\\theta_1$都为0，此时的 [RX(0)](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.RX.html) 门和 [RY(0)](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.RY.html) 门都相当于 [I](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/gates/mindquantum.core.gates.IGate.html) 门，因此整个线路演化的过程就是$|0\\rangle$经过$I\\cdot I$，那么最后输出的态当然就是$|0\\rangle$了。\n",
"\n",
"那么完整的量子线路就是Encoder加上Ansatz。"
]
},
{
"cell_type": "code",
"execution_count": 6,
"metadata": {
"tags": []
},
"outputs": [
{
"data": {
"image/svg+xml": [
"<svg xmlns=\"http://www.w3.org/2000/svg\" width=\"656.8\" height=\"80.0\" xmlns:xlink=\"http://www.w3.org/1999/xlink\"><rect x=\"0\" y=\"0.0\" width=\"656.8\" height=\"80.0\" fill=\"#ffffff\" /><text x=\"20.0\" y=\"40.0\" font-size=\"16px\" dominant-baseline=\"middle\" text-anchor=\"start\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#252b3a\" >q0: </text><line x1=\"48.8\" x2=\"636.8\" y1=\"40.0\" y2=\"40.0\" stroke=\"#adb0b8\" stroke-width=\"1\" /><rect x=\"72.8\" y=\"20.0\" width=\"40.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#5e7ce0\" fill-opacity=\"1\" /><text x=\"92.8\" y=\"40.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >H </text><rect x=\"132.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"172.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RX </text><text x=\"172.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >alpha0 </text><rect x=\"232.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"272.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RY </text><text x=\"272.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >alpha1 </text><rect x=\"332.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"372.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RZ </text><text x=\"372.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >alpha2 </text><rect x=\"432.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"472.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RX </text><text x=\"472.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >theta0 </text><rect x=\"532.8\" y=\"20.0\" width=\"80.0\" height=\"40\" rx=\"4\" ry=\"4\" stroke=\"#ffffff\" stroke-width=\"0\" fill=\"#fac209\" fill-opacity=\"1\" /><text x=\"572.8\" y=\"36.0\" font-size=\"20px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >RY </text><text x=\"572.8\" y=\"52.0\" font-size=\"14.0px\" dominant-baseline=\"middle\" text-anchor=\"middle\" font-family=\"Arial\" font-weight=\"normal\" fill=\"#ffffff\" >theta1 </text></svg>"
],
"text/plain": [
"<mindquantum.io.display.circuit_svg_drawer.SVGCircuit at 0x7f1f78a87520>"
]
},
"execution_count": 6,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# pylint: disable=W0104\n",
"circuit = encoder.as_encoder() + ansatz.as_ansatz() # 完整的量子线路由Encoder和Ansatz组成\n",
"circuit.svg()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从对完整的量子线路的Summary中可以看到，该量子线路由6个量子门组成，其中有5个含参量子门且参数为$\\alpha_0,\\alpha_1,\\alpha_2,\\theta_0,\\theta_1$，该量子线路调控的量子比特数为1。\n",
"\n",
"## 构建哈密顿量\n",
"\n",
"我们对第0位量子比特执行泡利`Z`算符测量，构建对应的哈密顿量。"
]
},
{
"cell_type": "code",
"execution_count": 7,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"-1 [Z0]\n"
]
}
],
"source": [
"from mindquantum.core.operators import QubitOperator # 导入QubitOperator模块，用于构造泡利算符\n",
"from mindquantum.core.operators import Hamiltonian # 导入Hamiltonian模块，用于构建哈密顿量\n",
"\n",
"ham = Hamiltonian(QubitOperator('Z0', -1)) # 对第0位量子比特执行泡利Z算符测量，且将系数设置为-1，构建对应的哈密顿量\n",
"print(ham)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从上述打印可以看到，此时构建的哈密顿量为对第0位量子比特执行泡利`Z`算符测量，且系数为-1。之所以将系数设为-1，是因为在量子神经网络的训练中，Ansatz中的参数的梯度会一直下降，同时测量值也会一直减少。如果最后收敛于-1，那么此时对应的量子态是$|1\\rangle$而不是$|0\\rangle$，如下所示\n",
"\n",
"$$\n",
"\\langle 1|Z|1\\rangle=-1.\n",
"$$\n",
"\n",
"而我们所希望得到的是$|0\\rangle$态。所以，将系数设为-1，那么当测量值为-1时，此时对应的量子态就是$|0\\rangle$态，如下所示\n",
"\n",
"$$\n",
"\\langle 0|(-Z)|0\\rangle=-1.\n",
"$$\n",
"\n",
"说明：\n",
"\n",
"（1）[QubitOperator](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/operators/mindquantum.core.operators.QubitOperator.html) 是作用于量子比特的算子的总和，主要用于构造泡利算符；一般格式如下：`QubitOperator(term=None, coefficient=1.0)`；\n",
"\n",
"（2）[Hamiltonian](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/operators/mindquantum.core.operators.Hamiltonian.html) 是哈密顿量包装器，主要用于构建哈密顿量，一般格式如下：`Hamiltonian(QubitOperator('X0 Y2', 0.5))`，`X0` 和 `Y2` 表示泡利 `X` 算符作用在第0位量子比特，泡利 `Y` 算符作用在第2位量子比特，系数为0.5。\n",
"\n",
"## 生成变分量子线路模拟算子\n",
"\n",
"对于上述搭建的量子线路，我们可以在MindSpore Quantum生成一个变分量子线路模拟算子对其进行模拟。\n",
"\n",
"首先，为了方便，我们对Encoder和Ansatz中的参数数组分别命名为encoder_names和ansatz_names。"
]
},
{
"cell_type": "code",
"execution_count": 8,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"encoder_names = ['alpha0', 'alpha1', 'alpha2'] \n",
"ansatz_names = ['theta0', 'theta1']\n"
]
}
],
"source": [
"encoder_names = encoder.params_name # Encoder中所有参数组成的数组，encoder.para_name系统会自动生成\n",
"ansatz_names = ansatz.params_name # Ansatz中所有参数组成的数组，ansatz.para_name系统会自动生成\n",
"\n",
"print('encoder_names = ', encoder.params_name, '\\nansatz_names =', ansatz.params_name)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从上述打印可以看到，encoder_names为Encoder中所有参数$\\alpha_0, \\alpha_1, \\alpha_2$组成的数组，ansatz_names为Ansatz中所有参数$\\theta_0,\\theta_1$组成的数组，这两个数组会在生成变分量子线路模拟算子时用到。\n",
"\n",
"然后，我们通过 [Simulator](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/simulator/mindquantum.simulator.Simulator.html) 模块得到变分量子线路演化和梯度求解的算子。"
]
},
{
"cell_type": "code",
"execution_count": 9,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Measurement result: [[0.29552022+0.j]]\n",
"Gradient of encoder parameters: [[[0.+0.j 0.+0.j 0.+0.j]]]\n",
"Gradient of ansatz parameters: [[[-0.37202556+0.j 0.87992317+0.j]]]\n"
]
}
],
"source": [
"# 导入Simulator模块\n",
"from mindquantum.simulator import Simulator\n",
"\n",
"# 生成一个基于mqvector后端的模拟器，并设置模拟器的比特数为量子线路的比特数。\n",
"sim = Simulator('mqvector', circuit.n_qubits)\n",
"\n",
"# 获取模拟器基于当前量子态的量子线路演化以及期望、梯度求解算子\n",
"grad_ops = sim.get_expectation_with_grad(ham,\n",
" circuit)\n",
"\n",
"# Encoder中的alpha0, alpha1, alpha2这三个参数组成的数组，\n",
"# 将其数据类型转换为float32，并储存在encoder_data中。\n",
"# MindSpore Quantum支持多样本的batch训练，Encoder数组是两个维度，\n",
"# 第一个维度为样本，第二个维度为特征（即参数）\n",
"encoder_data = np.array([[alpha0, alpha1, alpha2]]).astype(np.float32)\n",
"\n",
"# Ansatz中的theta0, theta1这两个参数组成的数组，将其数据类型转换为float32，\n",
"# 并储存在ansatzr_data中，Ansatz数据只有一个维度，特征（即参数）\n",
"ansatz_data = np.array([theta0, theta1]).astype(np.float32)\n",
"\n",
"# 根据Encoder和Ansatz的数据，输出变分量子线路的测量值，Encoder中的参数的导数和Ansatz中的参数的导数\n",
"measure_result, encoder_grad, ansatz_grad = grad_ops(encoder_data, ansatz_data)\n",
"\n",
"print('Measurement result: ', measure_result)\n",
"print('Gradient of encoder parameters: ', encoder_grad)\n",
"print('Gradient of ansatz parameters: ', ansatz_grad)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从上述打印可以看到，测量结果（期望值）为0.29552022，Encoder中的3个参数的导数为0，0，0（因为我们对Encoder设置了no_grad()），Ansatz中的2个参数的导数为-0.37202555，-0.87992316。\n",
"\n",
"这里通过 [get_expectation_with_grad](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/simulator/mindquantum.simulator.Simulator.html#mindquantum.simulator.Simulator.get_expectation_with_grad) 产生的只是一个算子，还不能进行训练，要把它放到量子神经网络里面才能进行训练。通过训练Ansatz中的参数，可以使得Ansatz中的参数的导数一直下降并接近于0，那么测量值也就会接近于-1。\n",
"\n",
"说明：\n",
"\n",
"（1）[Simulator](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/simulator/mindquantum.simulator.Simulator.html) 的 [get_expectation_with_grad](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/simulator/mindquantum.simulator.Simulator.html#mindquantum.simulator.Simulator.get_expectation_with_grad) 用于生成变分量子线路来模拟算子，一般格式如下：\n",
"\n",
"```python\n",
"\n",
"Simulator.get_expectation_with_grad(ham,\n",
" circ_right,\n",
" circ_left,\n",
" simulator_left,\n",
" parallel_worker=1)\n",
"\n",
"```\n",
"\n",
"此函数适用于计算如下模型：\n",
"\n",
"$$E=\\left<\\varphi\\right|U^\\dagger_l(\\theta) H U_r(\\theta)\\left|\\psi\\right>$$\n",
"\n",
"其中`circ_right`是$U_r$，`circ_left`是$U_l$，当不提供时，默认跟`circ_right`是相同的线路，$\\left|\\psi\\right>$是当前模拟器的量子态，$\\left|\\varphi\\right>$是 `simulator_left` 中的量子态。 `parallel_worker`指定并行数，当需要编码的经典数据是一个batch时，合理设置此参数可以提高计算效率。\n",
"\n",
"（2）MindSpore是一个全场景深度学习框架，旨在实现易开发、高效执行、全场景统一部署三大目标，提供支持异构加速的张量可微编程能力，支持云、服务器、边和端多种硬件平台。\n",
"\n",
"## 搭建量子神经网络"
]
},
{
"cell_type": "code",
"execution_count": 10,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"MQLayer<\n",
" (evolution): MQOps<1 qubit mqvector VQA Operator>\n",
" >"
]
},
"execution_count": 10,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# pylint: disable=W0104\n",
"from mindquantum.framework import MQLayer # 导入MQLayer\n",
"import mindspore as ms # 导入mindspore\n",
"\n",
"ms.set_seed(1) # 设置生成随机数的种子\n",
"ms.set_context(mode=ms.PYNATIVE_MODE, device_target=\"CPU\")\n",
"\n",
"QuantumNet = MQLayer(grad_ops)\n",
"QuantumNet"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"上述打印可以看到，我们已经成功搭建了量子机器学习层，其可以无缝地跟MindSpore中其他的算子构成一张更大的机器学习网络。\n",
"\n",
"说明：\n",
"\n",
"（1）MindSpore Quantum中的量子线路梯度计算算子都是在`PYNATIVE_MODE`下的，因此需要设置MindSpore的运行模式。\n",
"\n",
"（2）我们也可以通过如下代码方式搭建量子机器学习层，只是在MindSpore Quantum中，已经将下述过程封装打包，这样我们就可以直接利用 [MQLayer](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/framework/layer/mindquantum.framework.MQLayer.html) 模块搭建量子机器学习层。对于更复杂的量子-经典混合神经网络，如下搭建方式会展示它的优势。"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"```python\n",
"from mindspore import nn\n",
"\n",
"class MQLayer(nn.Cell):\n",
" def __init__(self, expectation_with_grad, weight='normal'):\n",
" super(MQLayer, self).__init__()\n",
" self.evolution = MQOps(expectation_with_grad)\n",
" weight_size = len(\n",
" self.evolution.expectation_with_grad.ansatz_params_name)\n",
" self.weight = Parameter(initializer(weight,\n",
" weight_size,\n",
" dtype=ms.float32),\n",
" name='ansatz_weight')\n",
"\n",
" def construct(self, x):\n",
" return self.evolution(x, self.weight)\n",
"\n",
"```"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## 训练\n",
"\n",
"我们采用 [Adam](https://www.mindspore.cn/docs/zh-CN/r2.5.0/api_python/nn/mindspore.nn.Adam.html#mindspore.nn.Adam) 优化器优化Ansatz中的参数。"
]
},
{
"cell_type": "code",
"execution_count": 11,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"0 : [[0.30633032]]\n",
"10 : [[-0.87929225]]\n",
"20 : [[-0.9726834]]\n",
"30 : [[-0.9935059]]\n",
"40 : [[-0.9934296]]\n",
"50 : [[-0.996459]]\n",
"60 : [[-0.99859095]]\n",
"70 : [[-0.9995031]]\n",
"80 : [[-0.9998626]]\n",
"90 : [[-0.9999718]]\n",
"100 : [[-0.9999983]]\n",
"110 : [[-0.9999993]]\n",
"120 : [[-0.99999785]]\n",
"130 : [[-0.9999986]]\n",
"140 : [[-0.9999997]]\n",
"150 : [[-1.]]\n",
"160 : [[-0.99999994]]\n",
"170 : [[-1.]]\n",
"180 : [[-1.]]\n",
"190 : [[-1.]]\n"
]
}
],
"source": [
"from mindspore.nn import Adam, TrainOneStepCell # 导入Adam模块和TrainOneStepCell模块\n",
"\n",
"opti = Adam(QuantumNet.trainable_params(), learning_rate=0.5) # 需要优化的是Quantumnet中可训练的参数，学习率设为0.5\n",
"net = TrainOneStepCell(QuantumNet, opti)\n",
"\n",
"for i in range(200):\n",
" res = net(ms.Tensor(encoder_data))\n",
" if i % 10 == 0:\n",
" print(i, ': ', res)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从上述打印可以看到，最后测量值收敛于-1。\n",
"\n",
"说明：\n",
"\n",
"（1）[Adam](https://www.mindspore.cn/docs/zh-CN/r2.5.0/api_python/nn/mindspore.nn.Adam.html#mindspore.nn.Adam) 模块通过自适应矩估计算法更新梯度，可以优化Ansazt中的参数，输入的是神经网络中可训练的参数；一般格式如下：nn.Adam(net.trainable_params(), learning_rate=0.5)；\n",
"\n",
"（2）[TrainOneStepCell](https://www.mindspore.cn/docs/zh-CN/r2.5.0/api_python/nn/mindspore.nn.TrainOneStepCell.html#mindspore.nn.TrainOneStepCell) 模块为网络训练包类，用优化器包装网络。生成的单元格使用输入“inputs”进行训练，将在构造函数中创建反向图，以更新参数，有不同的并行模式可用于训练。一般格式如下：`nn.TrainOneStepCell(network, optimizer, sens=1.0)`。\n",
"\n",
"## 结果呈现\n",
"\n",
"由于测量值已经收敛于-1，所以我们可以打印此时Ansatz中的参数。"
]
},
{
"cell_type": "code",
"execution_count": 12,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"[ 2.241999 -1.0756955]\n"
]
}
],
"source": [
"theta0, theta1 = QuantumNet.weight.asnumpy()\n",
"\n",
"print(QuantumNet.weight.asnumpy())"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从上述打印可以看到，此时Ansatz中的参数$\\theta_1, \\theta_2$分别为2.2420275和-1.0756909。\n",
"\n",
"通过 [get_qs](https://www.mindspore.cn/mindquantum/docs/zh-CN/r0.10/core/circuit/mindquantum.core.circuit.Circuit.html#mindquantum.core.circuit.Circuit.get_qs)，可以输出量子线路在最优参数时的量子态。"
]
},
{
"cell_type": "code",
"execution_count": 13,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"(0.371309288022958-0.9285092418490042j)¦0⟩\n",
"(2.0012262153618066e-05+5.327365456997413e-06j)¦1⟩\n"
]
}
],
"source": [
"pr = {'alpha0': alpha0, 'alpha1': alpha1, 'alpha2': alpha2, 'theta0': theta0, 'theta1': theta1}\n",
"state = circuit.get_qs(pr=pr, ket=True)\n",
"\n",
"print(state)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"从上述打印可以看到，这就是量子线路在最优参数时的量子态。从其数值表示可以看到，这是一个接近于目标态$|0\\rangle$的态。最后，我们计算一下此量子态与目标态$|0\\rangle$的保真度（用于验证两个量子态的相似程度），并将保真度打印。"
]
},
{
"cell_type": "code",
"execution_count": 14,
"metadata": {
"tags": []
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"0.9999999995711284\n"
]
}
],
"source": [
"state = circuit.get_qs(pr=pr)\n",
"fid = np.abs(np.vdot(state, [1, 0]))**2 # 保真度fidelity为向量内积的绝对值的模平方，即计算此时量子态对应的向量与|0>态对应的向量[1,0]的内积的模平方\n",
"\n",
"print(fid)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"可以看到，此时的保真度为100.00%，也就是说，该状态与目标态$|0\\rangle$的相似程度为100.00%。\n",
"\n",
"综上所述，我们搭建了一个简单的量子神经网络，通过训练Ansatz中的参数，抵消了Encoder对初始量子态产生的误差，使得最后的量子态仍为$|0\\rangle$，且保真度达到100.00%。\n",
"\n",
"至此，我们通过MindSpore Quantum完成了对量子神经网络的初体验！赶紧动手体验一下量子编程的乐趣吧！"
]
},
{
"cell_type": "code",
"execution_count": 15,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
"<table border=\"1\">\n",
" <tr>\n",
" <th>Software</th>\n",
" <th>Version</th>\n",
" </tr>\n",
"<tr><td>mindquantum</td><td>0.9.11</td></tr>\n",
"<tr><td>scipy</td><td>1.10.1</td></tr>\n",
"<tr><td>numpy</td><td>1.23.5</td></tr>\n",
"<tr>\n",
" <th>System</th>\n",
" <th>Info</th>\n",
"</tr>\n",
"<tr><td>Python</td><td>3.9.16</td></tr><tr><td>OS</td><td>Linux x86_64</td></tr><tr><td>Memory</td><td>8.3 GB</td></tr><tr><td>CPU Max Thread</td><td>8</td></tr><tr><td>Date</td><td>Sun Dec 31 21:28:06 2023</td></tr>\n",
"</table>\n"
],
"text/plain": [
"<mindquantum.utils.show_info.InfoTable at 0x7f1ff025f490>"
]
},
"execution_count": 15,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from mindquantum.utils.show_info import InfoTable\n",
"\n",
"InfoTable('mindquantum', 'scipy', 'numpy')"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "base",
"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": "3.9.16"
}
},
"nbformat": 4,
"nbformat_minor": 2
}