quantum_model_MERA_1_layers.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Quantum model\n",
    "\n",
    "The fair model is designed from the data set that was made in the data analysis section and his can compare with the classical fair model, for this will be with the use of neural networks to predict when there is a backorder and thus pass the data to the company responsible for the distribution and storage."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "fKJm0SmcDpUD",
    "outputId": "93105aae-e162-4f7b-eb9e-100cce222d6d"
   },
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "In case you don't have installed tensorflow uncomment the following cell"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#pip install pennyane"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Callign the dependencies fro mpennylane the optimizers, and methods to generate a quantum circuit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "cK5Ry9dADtir",
    "outputId": "b438ff3d-7cc8-45f8-e3db-1d436e81c1c6"
   },
   "outputs": [],
   "source": [
    "#improt pennylane dependnecies\n",
    "import pennylane as qml\n",
    "from pennylane import numpy as np\n",
    "from pennylane.optimize import NesterovMomentumOptimizer\n",
    "\n",
    "\n",
    "# load the csv files\n",
    "import pandas as pd\n",
    "\n",
    "# plot the historical acc and cost\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "\n",
    "\n",
    "from IPython.display import clear_output\n",
    "clear_output(wait=False)\n",
    "\n",
    "import os\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Load the train and test sets from the Data analysis module"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {
    "id": "ZTdT7iH2DpUG"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(((2000, 16), (2000,)), ((6159, 16), (6159,)))"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_train = pd.read_csv(\"dataset/fair_train.csv\")\n",
    "X_train,y_train = data_train[data_train.columns[:16]].values, data_train[data_train.columns[16]].values\n",
    "\n",
    "data_test = pd.read_csv(\"dataset/classic_test.csv\")\n",
    "X_test,y_test = data_test[data_test.columns[:16]].values, data_test[data_test.columns[16]].values\n",
    "\n",
    "(X_train.shape, y_train.shape),(X_test.shape, y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Init the device that called the backend default.qubit and using 4 qubits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "s804AwFCEFV8"
   },
   "outputs": [],
   "source": [
    "n_wires = 4\n",
    "dev = qml.device(\"default.qubit\", wires=n_wires)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define the structure of the entenglament for MERA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def block(weights, wires):\n",
    "    qml.CNOT(wires=[wires[0],wires[1]])\n",
    "    qml.RY(weights[0], wires=wires[0])\n",
    "    qml.RY(weights[1], wires=wires[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define the parameters for MERA quantum circuit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_block_wires = 2\n",
    "n_params_block = 2\n",
    "n_blocks = qml.MERA.get_n_blocks(range(n_wires),n_block_wires)\n",
    "n_blocks"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Called the quantum circuit for all the data train using amplitude embedding, where the  data is encoded into the amplitudes of a quantum state. A normalized classical $N$ -dimensional datapoint $x$ is represented by the amplitudes of a $n$-qubit quantum state $\\left|\\psi_{x}\\right\\rangle$ as\n",
    "$$\n",
    "\\left|\\psi_{x}\\right\\rangle=\\sum_{i=1}^{N} x_{i}|i\\rangle\n",
    "$$\n",
    "where $N=2^{n}, x_{i}$ is the $i$-th element of $x$, and $|i\\rangle$ is the $i$-th computational basis state. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "P2gS_r0dEN5s"
   },
   "outputs": [],
   "source": [
    "@qml.qnode(dev)\n",
    "def circuit(weights, x):\n",
    "\n",
    "    qml.AmplitudeEmbedding(x, wires=[0,1,2,3],normalize=True,pad_with=True)\n",
    "\n",
    "    for w in weights:\n",
    "\n",
    "        qml.MERA(range(n_wires),n_block_wires,block, n_params_block, w)\n",
    "        #print(w)\n",
    "    #print(x)\n",
    "    \n",
    "    return qml.expval(qml.PauliZ(1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "id": "xLSeVH8CEPUi"
   },
   "outputs": [],
   "source": [
    "def variational_classifier(weights, bias, x):\n",
    "    return circuit(weights, x) + bias\n",
    "\n",
    "def square_loss(labels, predictions):\n",
    "    loss = 0\n",
    "    for l, p in zip(labels, predictions):\n",
    "        loss = loss + (l - p) ** 2\n",
    "\n",
    "    loss = loss / len(labels)\n",
    "    return loss\n",
    "def accuracy(labels, predictions):\n",
    "\n",
    "    loss = 0\n",
    "    for l, p in zip(labels, predictions):\n",
    "        if abs(l - p) < 1e-5:\n",
    "            loss = loss + 1\n",
    "    loss = loss / len(labels)\n",
    "\n",
    "    return loss\n",
    "\n",
    "def cost(weights, bias, X, Y):\n",
    "    #print(1)\n",
    "    predictions = [variational_classifier(weights, bias, x) for x in X]\n",
    "    return square_loss(Y, predictions)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "yFN174tZEU22",
    "outputId": "7de4359e-7953-487e-8f08-30a7623d639e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[[11.08386778  2.51426189]\n",
      "  [ 6.14959212 14.07994722]\n",
      "  [11.73421292 -6.14041802]\n",
      "  [ 5.96958159 -0.95100539]\n",
      "  [-0.64854317  2.57986647]]] 0.0\n"
     ]
    }
   ],
   "source": [
    "np.random.seed(0)\n",
    "num_layers = 1\n",
    "weights_init = 2*np.pi * np.random.randn(num_layers,n_blocks, n_params_block, requires_grad=True)\n",
    "bias_init = np.array(0.0, requires_grad=True)\n",
    "\n",
    "print(weights_init, bias_init)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0: ─╭QubitStateVector(M0)─╭X──RY(2.51)────────────────╭●──RY(5.97)──╭●──RY(-0.65)─┤     \n",
      "1: ─├QubitStateVector(M0)─╰●──RY(11.08)─╭X──RY(-6.14)─│─────────────╰X──RY(2.58)──┤  <Z>\n",
      "2: ─├QubitStateVector(M0)─╭●──RY(6.15)──│─────────────╰X──RY(-0.95)───────────────┤     \n",
      "3: ─╰QubitStateVector(M0)─╰X──RY(14.08)─╰●──RY(11.73)─────────────────────────────┤     \n"
     ]
    }
   ],
   "source": [
    " print(qml.draw(circuit,expansion_strategy='device',wire_order=[0,1,2,3])(weights_init,np.asarray(X_train[0])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "g2IhcVFlEV_V",
    "outputId": "3b78e6fc-f121-4095-a4b3-47e1b07ce621"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[11.08386778  2.51426189]\n"
     ]
    }
   ],
   "source": [
    "for i in weights_init:\n",
    "    print(i[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "id": "1WEWkuIOEXnm"
   },
   "outputs": [],
   "source": [
    "y_train = np.where(y_train < 1, -1, y_train)\n",
    "y_test = np.where(y_test < 1, -1, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "id": "Vrs1X5UiMnWJ"
   },
   "outputs": [],
   "source": [
    "from sklearn.utils import shuffle\n",
    "X,y = shuffle(X_train, y_train, random_state=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "s8Z8HLOXEY1n",
    "outputId": "249d3445-bac4-4885-8c86-49053fd8b78a"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iter:    50 | Cost: 0.6213926 | Acc train: 0.8122222 | Acc validation: 0.8300000 \n",
      "[[[11.37998162  4.3055446 ]\n",
      "  [ 4.52880898 15.13469819]\n",
      "  [11.73421292 -6.84553768]\n",
      "  [ 8.31808883 -0.95100539]\n",
      "  [-0.64854317  3.70480243]]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "opt = NesterovMomentumOptimizer(0.4)\n",
    "batch_size = 32\n",
    "\n",
    "num_data = len(y_train)\n",
    "num_train = 0.9\n",
    "# train the variational classifier\n",
    "weights = weights_init\n",
    "bias = bias_init\n",
    "print()\n",
    "\n",
    "cost_g = []\n",
    "acc_train = []\n",
    "acc_test = []\n",
    "plt.show()\n",
    "for it in range(50):\n",
    "    X_train_70, X_test_30, y_train_70, y_test_30 =train_test_split(np.asarray(X), np.asarray(y), train_size=num_train, test_size=1.0-num_train, shuffle=True)\n",
    "    # Update the weights by one optimizer step\n",
    "    batch_index = np.random.randint(0, len(X_train_70), (batch_size,))\n",
    "    feats_train_batch = X_train_70[batch_index]\n",
    "    Y_train_batch = y_train_70[batch_index]\n",
    "    weights, bias, _, _ = opt.step(cost, weights, bias, feats_train_batch, Y_train_batch)\n",
    "\n",
    "    # Compute predictions on train and validation set\n",
    "    predictions_train = [np.sign(variational_classifier(weights, bias, f)) for f in X_train_70]\n",
    "    predictions_val = [np.sign(variational_classifier(weights, bias, f)) for f in X_test_30]\n",
    "    # Compute accuracy on train and validation set\n",
    "    acc_tra = accuracy(y_train_70, predictions_train)\n",
    "    acc_val = accuracy(y_test_30, predictions_val)\n",
    "    cost_train = cost(weights, bias,X_train, y_train)\n",
    "    cost_g.append(cost_train)\n",
    "    \n",
    "    acc_train.append(acc_tra)\n",
    "    acc_test.append(acc_val)\n",
    "    \n",
    "    clear_output(wait=True)\n",
    "    plt.plot(cost_g,label='cost')\n",
    "    plt.plot(acc_train,label='acc_train')\n",
    "    plt.plot(acc_test,label='acc_test')\n",
    "    plt.legend(['cost','acc_train','acc_test'])\n",
    "    plt.show()\n",
    "     \n",
    "    \n",
    "    print(\n",
    "         \"Iter: {:5d} | Cost: {:0.7f} | Acc train: {:0.7f} | Acc validation: {:0.7f} \"\n",
    "         \"\".format(it + 1, cost_train, acc_tra, acc_val)\n",
    "     )\n",
    "    print(weights)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.4423974 , 0.5898632 , 0.        , 0.        , 0.        ,\n",
       "       0.        , 0.        , 0.1474658 , 0.2949316 , 0.4423974 ,\n",
       "       0.3686645 , 0.        , 0.07225824, 0.07299557, 0.        ,\n",
       "       0.0737329 ])"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_test = []\n",
    "for x in X_test.tolist():\n",
    "    if sum(x) == 0:\n",
    "        x[0]=1\n",
    "    x_test.append( x/ np.linalg.norm(x))\n",
    "x_test[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([-1, -1, -1, ...,  1,  1,  1], requires_grad=True)"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test_pred = [np.sign(variational_classifier(weights, bias, f)) for f in x_test]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_test_pred = []\n",
    "for i in y_pred:\n",
    "    if i < 0:\n",
    "        y_test_pred.append(-1)\n",
    "    else:\n",
    "        y_test_pred.append(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.796720\n",
      "Precision: 0.469509\n",
      "Recall: 0.838795\n",
      "F1 score: 0.602034\n",
      "Cohens kappa: 0.479749\n",
      "ROC AUC: 0.813036\n",
      "[[3960 1070]\n",
      " [ 182  947]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "from sklearn.metrics import confusion_matrix, roc_curve, auc\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "# demonstration of calculating metrics for a neural network model using sklearn\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import precision_score\n",
    "from sklearn.metrics import recall_score\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.metrics import cohen_kappa_score\n",
    "from sklearn.metrics import roc_auc_score\n",
    "\n",
    "accuracy = accuracy_score(y_test, y_test_pred)\n",
    "print('Accuracy: %f' % accuracy)\n",
    "# precision tp / (tp + fp)\n",
    "precision = precision_score(y_test, y_test_pred)\n",
    "print('Precision: %f' % precision)\n",
    "# recall: tp / (tp + fn)\n",
    "recall = recall_score(y_test, y_test_pred)\n",
    "print('Recall: %f' % recall)\n",
    "# f1: 2 tp / (2 tp + fp + fn)\n",
    "f1 = f1_score(y_test, y_test_pred)\n",
    "print('F1 score: %f' % f1)\n",
    " \n",
    "# kappa\n",
    "kappa = cohen_kappa_score(y_test, y_test_pred)\n",
    "print('Cohens kappa: %f' % kappa)\n",
    "# ROC AUC\n",
    "auc = roc_auc_score(y_test, y_test_pred)\n",
    "print('ROC AUC: %f' % auc)\n",
    "# confusion matrix\n",
    "test_matrix = confusion_matrix(y_test, y_test_pred)\n",
    "print(test_matrix)\n",
    "ax = sns.heatmap(test_matrix, annot=True, cmap='Blues', fmt='g')\n",
    "\n",
    "ax.set_title('Seaborn Confusion Matrix with labels\\n\\n');\n",
    "ax.set_xlabel('\\nPredicted Values')\n",
    "ax.set_ylabel('Actual Values ');\n",
    "\n",
    "ax.xaxis.set_ticklabels(['0','1'])\n",
    "ax.yaxis.set_ticklabels(['0','1'])\n",
    "\n",
    "## Display the visualization of the Confusion Matrix.\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred_1 = [int(i) for i in y_test_pred ]\n",
    "y_pred_1 = [\"{}\\n\".format(i) for i in y_pred_1]\n",
    "with open(r'mera_1_layers.csv', 'w') as fp:\n",
    "    fp.writelines(y_pred_1)"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "name": "classic_BackOrders_accuracy_1_v4.ipynb",
   "provenance": []
  },
  "gpuClass": "standard",
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.8.13"
  },
  "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
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}