线性回归和随机梯度下降.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import datasets\n",
    "from sklearn.model_selection import train_test_split\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import random\n",
    "\n",
    "data_x, data_y = datasets.load_svmlight_file('housing_scale.txt')\n",
    "x_train, x_test, y_train, y_test = train_test_split(data_x, data_y, test_size = 0.2)\n",
    "x_train = x_train.toarray()\n",
    "x_test = x_test.toarray()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# loss函数：$J(\\theta)=\\frac{1}{2m}\\sum_{i=0}^m (h(x^i)-y^i)^2$\n",
    "# 梯度：$G(\\theta_j) = \\frac{1}{m}\\sum_{i=0}^m ((h(x^i)-y^i)*x_j^i)$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#计算训练集的行数\n",
    "train_array_row_size, = y_train.shape\n",
    "test_array_row_size, = y_test.shape\n",
    "#生成一个行数=train_array_row_size的全1向量\n",
    "temp_vector = np.ones(train_array_row_size)\n",
    "#算出X\n",
    "X_train = np.column_stack((x_train,temp_vector))\n",
    "temp_vector = np.ones(test_array_row_size)\n",
    "X_test = np.column_stack((x_test,temp_vector))\n",
    "#初始化学习率\n",
    "learnRate = 0.0001"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeLoss_train():\n",
    "    loss_train = (1 / 2 / train_array_row_size)*np.dot(np.transpose((np.transpose(y_train) - np.dot(X_train, w))),(y_train - np.dot(X_train, w)))\n",
    "    return loss_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeLoss_test():\n",
    "    loss_test = (1 / 2 / test_array_row_size)*np.dot(np.transpose((np.transpose(y_test) - np.dot(X_test, w))),(y_test - np.dot(X_test, w)))\n",
    "    return loss_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeGradient_train():\n",
    "    temp = random.randint(0,train_array_row_size-1)\n",
    "    pred_y = np.dot(X_train,w)[temp]\n",
    "    for i in range (14):\n",
    "        w[i] = w[i] - learnRate * (pred_y - y_train[temp]) * X_train[temp][i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def computeGradient_test():\n",
    "    temp = random.randint(0,test_array_row_size-1)\n",
    "    pred_y = np.dot(X_test,w)[temp]\n",
    "    for i in range (14):\n",
    "        w[i] = w[i] - learnRate * (pred_y - y_test[temp]) * X_test[temp][i]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "#初始化w\n",
    "w = np.zeros(14)\n",
    "train_loss_tuple = []\n",
    "loss_train1 = 0\n",
    "for i in range (10000):\n",
    "    computeGradient_train()\n",
    "    loss_train1 = computeLoss_train()\n",
    "    train_loss_tuple.append(computeLoss_train())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#初始化w\n",
    "w = np.zeros(14)\n",
    "test_loss_tuple = []\n",
    "loss_val = 0\n",
    "for i in range (10000):\n",
    "    computeGradient_test()\n",
    "    loss_val = computeLoss_test()\n",
    "    test_loss_tuple.append(computeLoss_test())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAElCAYAAAD+wXUWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3Xd4VGX2wPHvyUwqhIRm6CDFEkFK6KiAIoq6i+taV0UFZfenrrqWVdfdtbEu6+5asBdwWRuyKhZEsRERC0iTFpEuVWqAQHrO7497g0OcJJOQyc1kzud57jP3vvPeO+edC3Ny2/uKqmKMMcaUFeN1AMYYY+omSxDGGGOCsgRhjDEmKEsQxhhjgrIEYYwxJihLEMYYY4KyBGEinojcIyIvHcH6y0VkSA2GVLrdTBG5uqa3a0xtsQRhooqI/EdExgWWqeoJqprpUUhBWXIxdYElCFOniYjf6xiMiVaWIEydIyLrReR2EVkCHBARv4i0EpE3RGSHiKwTkRsqWP9/IrJNRPaKyGwROcEtHwtcCvxRRHJE5N2AzxvmzseLyCMissWdHhGRePe9ISKySURuEZHtIrJVRK6qpDmdRGSeG8vbItIkIM7+IvKliGSLyLelp7lE5G/AycDjbpyPi8i9IvKY+36siBwQkQfd5UQRyRORxhVt130vRUQmurFvFpFxIuJz37tSROaIyL9EZI/7PY8Ifc+Z+sYShKmrLgHOBlKBEuBd4FugNXAacJOInFHOuu8DXYCjgIXAywCq+qw7/6CqNlTVXwRZ9y6gP9AD6A70Bf4c8H4LIMWNYwzwROkPczlGAaOBVkARMAFARFoD7wHjgCbArcAbItJcVe8CPgeud+O8HvgMGOJusw+wDRjsLg8AVqrqnoq269ad7MbRGegJDAcCT2X1A1YCzYAHgYkiIhW0z9RjliBMXTVBVTeqai7OD2JzVb1PVQtUdS3wHHBxsBVVdZKq7lfVfOAeoLuIpIT4uZcC96nqdlXdAdwLXB7wfqH7fqGqzgBygGMr2N6LqrpMVQ8AfwEudP9ivwyYoaozVLVEVT8C5gNnlbOdr4AuItIUOAWYCLQWkYY4ieIzt1652xWRNGAEcJOqHlDV7cDDHP49blDV51S1GCeZtATSKv3WTL1k53dNXbUxYL490EpEsgPKfDh/ZR/G/fH9G3AB0Bzn6AOcv4j3hvC5rYANAcsb3LJSu1S1KGD5INCwgu0FtmMDEOvG0h64QEQCj2JigVnBNqKquSIyHycZnILTxh7AILfsMbdqRdtt785vDTgoiCkT47aAzzzo1quofaYeswRh6qrAboY3AutUtUsI6/0GGAkMA9bjnA7aA5T+IlbWffEWnB/S5e5yO7esutoGzLfDOQLZidOmF1X1mnLWCxbnZ8CpOKeGvnGXz8A5DTbbrVPudkWkJZAPNCuT5IwJyk4xmUgwD9jnXrhOFBGfiHQVkT5B6ibj/AjuApKAB8q8/yPQsYLPehX4s4g0F5FmwF+Baj9jAVwmIukikgTcB7zunr55CfiFiJzhtifBvQjepoI4P8O5prFCVQuATJzrB+vc02FUtF1V3Qp8CPxbRBqJSIyIdBKRwRgThCUIU+e5P6i/wDmlsg7nL/DncY4OyvovzqmczcAK4Osy708E0t07fN4Ksv44nHP2S4ClOBe5xwWpF6oXgf/gnLpJAG5w27QR50jnT8AOnL/8b+On/5OPAue7dxNNcMu+BBL56WhhBZAXsBzKdkcBce66e4DXca4zGPMzYgMGGWOMCcaOIIwxxgRlCcIYY0xQliCMMcYEZQnCGGNMUJYgTI0I7M8okpT2rxTmz3hfRK6o6bpHSkT+JCLP18ZnmchkD8oZUwERUaCLqq6u7jZUNeQO76pStyrcDvteUtXS5yxQ1bLPiBhzGDuCMOYIiHVHbuoxSxCmxlXSZXYzEZnuPqi2W0Q+F5EY973b3S6o94vIShE5rZztny0ii0Rkn4hsFJF7At7rICIqIleIyA8islNE7gp4P1GcQYP2iMgKnI4Ay2tH6QNo37rdbl8kP3X5fbuIbANeEJHGbpt2uNudHvBE9GGD/1TWpXYV6x4tTnfm+0XkYxF5QoKMrCciDXB6uG3ltiNHnO7TD43EF/C9XeV+p3tE5Hci0kdElrj76/Ey2x0tIllu3Zki0t4tFxF5WJwu0fe663ct73s2dZclCBMOFXWZfQuwCacjvTScJ35VRI4Frgf6qGoyTh9D68vZ/gGcJ4JTcboE/z8RObdMnZNwelk9DfiriBzvlt8NdHKnM4Byz/er6inubHe32+3X3OUWOF1ptwfG4vw/esFdbgfkAo9Tvqp0qV1R3VdwuiFpitNr7eXBNuD2JDsC2OK2o6Gqlte/VD+crtIvAh7B2ZfDgBNweqIdDOB+338CzsPZl5/jdFMCThfipwDH4Oyji3C6PjERxhKECYeKuswuxOnaob3bZfbn6jzOXwzE43SDEauq61V1TbCNq2qmqi51u7NegvPDVLY/oXtVNVdVv8UZR6K7W34h8DdV3e12SzGBqisB7lbVfPczdqnqG6p6UFX34/S0WlH/RlXpUjtoXRFph3P081e3C/Q5wDvVaEtZ96tqnqp+iJOIX3X342acJNDTrfdb4O+qmuV2/PcA0MM9iijE6RPrOJzeGrLcfqBMhLEEYcKhoi6z/wmsBj4UkbUicgeAexH4Jpy/hLeLyBQRCexm+xAR6Scis9xTOnuB3+H8hR1oW8B8YJfcrfh5F9xVtUNV8wLiSRKRZ0Rkg4jsw+kbKVXckdqCOKxLbXe2vC61y6vbCtgdUAaHt6u6fgyYzw2yXBpne+BR99RTNrAbp8fc1qr6Kc4R1BPAjyLyrIg0qoHYTC2zBGHCobTL7FKHusx2B/K5RVU74nTAd3PptQZVfUVVT3LXVeAf5Wz/FZy/ltuqagrwND91512Zrfy8C+6qKtuB2S04p7P6qWojnNMrVCGm6tgKNBGnl9hSbcurTOXdnFfVRuC3qpoaMCWq6pcAqjpBVTNwTk0dg9NhoIkwliBMOJTbZbaInCMind3z6PtwTi0Vi8ixInKqOBez83D+Wi0uZ/vJOH8954lIX5wxIEI1FbjTvbDcBvh9JfUr6x68NJ5cIFucMafvrkI81aKqG3B6nb1HROJEZABOwi3Pj0BTCX1kvco8jfM9lo73nSIiF7jzfdyjvFic01R5lL8vTR1mCcKEQ0VdZncBPsYZqvMr4ElVzcS5/jAepyvvbTjjSf+pnO1fC9wnIvtxks/UKsR2L85ppXU4YyO8WEn9e4DJ7qmUC8up8whON9w7cboX/6AK8RyJS3HGo96F8/2+hjMWxs+o6nc4iXut25agp+9CparTcI7wprin1ZbhXAgHaIQzJOwenO96F/CvI/k84w3r7tuYekJEXgO+U9WwH8GY6GBHEMZEKPdUTidxRoY7E2egoGCDIBlTLfYUqDGRqwXwJs5zEJuA/1PVRd6GZOoTO8VkjDEmKDvFZIwxJqiIPsXUrFkz7dChQ7XWPXDgAA0aNKjZgOo4a3N0sDZHhyNp84IFC3aqavPK6kV0gujQoQPz58+v1rqZmZkMGTKkZgOq46zN0cHaHB2OpM0iElIPAnaKyRhjTFCWIIwxxgRlCcIYY0xQEX0NwhhT/xQWFrJp0yby8vIqr+xKSUkhKysrjFHVPaG0OSEhgTZt2hAbG1utz7AEYYypUzZt2kRycjIdOnSg/HGUDrd//36Sk5PDHFndUlmbVZVdu3axadMmjj766Gp9hp1iMsbUKXl5eTRt2jTk5GCCExGaNm1apSOxssKWIEQkQUTmici3IrJcRO51y48WkbkiskpEXhOROLc83l1e7b7fIVyxGWPqNksONeNIv8dwHkHkA6eqanecsYnPFJH+OF0EP6yqXXC6Ax7j1h8D7FHVzsDDlD9YzBFbvhyeeqoTubnh+gRjjIl8YUsQ6shxF2PdSYFTgdfd8slA6WDzI91l3PdPq2Ag9yOydi1MndqWhQvDsXVjTCTbtWsXPXr0oEePHrRo0YLWrVsfWi4oKAhpG1dddRUrV64M+TOff/55brrppuqGHDZhvUjtjsm7AOiMMz7tGiDbHeQcnB4oW7vzrXHH1FXVInes4aY4g7AEbnMsMBYgLS2NzMzMKse1d28i0I933/2OwsJtldavL3Jycqr1fUUya3PkSUlJYf/+/VVap7i4uMrrlCcuLo7PP/8cgAceeICGDRtyww03AJCfn09+fj6qiqoSExP8b+wJEyYAhBxTXl4eBQUFVWpDqG3Oy8ur9r+HsCYIVS0GeohIKjANOD5YNfc12NHCz7qaVdVngWcBevfurdV51LygAK68soSYmOMYMuS4Kq8fqaw7gugQ6W3Oysqq8h1J4bqLKT4+nvj4eJKTk1m9ejXnnnsuJ510EnPnzmX69Once++9LFy4kNzcXC666CL++te/AnDSSSfx+OOP07VrV5o1a8bvfvc73n//fZKSknj77bc56qijDvuchIQE4uLiSE5OZt26dYwePZpdu3aRlpbGCy+8QJs2bZgyZQrjxo3D5/PRpEkT3nnnHdavX8/o0aMpLCykpKSEt956i44dO/5s2z179qxW+2vlNldVzRaRTKA/kCoifvcoog3uYPY4RxNtgU0i4gdSgN3hiCcuDtq1O8jSpQ3DsXljTE256SZYvLjSaonFxeDzhbbNHj3gkUeqFc6KFSt44YUXePrppwEYP348TZo0oaioiKFDh3L++eeTnp5+2Dp79+5l8ODBjB8/nptvvplJkyZxxx13lPsZ1157LVdffTWXXnopzz77LDfddBOvv/469957L5mZmaSlpZGdnQ3Ak08+ya233spFF1106MimJoXzLqbm7pEDIpIIDAOygFnA+W61K4C33fl33GXc9z/VcA1W8cMPnJCwiiXfloRl88aY+qlTp0706dPn0PKrr75Kr1696NWrF1lZWaxYseJn6yQmJjJihDNcd0ZGBuvXr6/wM+bOncvFF18MwKhRow6d7ho0aBCjRo3i+eefp6TE+e0aOHAg48aN48EHH2Tjxo0kJCTURDMPCecRREucwd59OIloqqpOF5EVOAOdjwMWARPd+hOBF0VkNc6Rw8Vhi+ybb+iXtYjp9CQ3FxITw/ZJxpgjEeJf+rm19KBcYPfaq1at4tFHH2XevHmkpqZy2WWXBX3mIC4u7tC8z+ejqKjoZ3VC8dxzzx06tdW9e3e++OILLr/8cgYMGMB7773H6aefzuTJkznllFOqtf1gwnkX0xJV7amqJ6pqV1W9zy1fq6p9VbWzql6gqvlueZ673Nl9f224YuOYY+jEGsC5o8kYY6pq3759JCcn06hRI7Zu3crMmTNrZLv9+/dn6tSpALz00kuHfvDXrl1L//79uf/++2ncuDFbt25l7dq1dO7cmRtvvJGzzz6bJUuW1EgMpaKzq43OnekQswFKYM0aOOEErwMyxkSaXr16kZ6eTteuXenYsSODBg2qke0+/vjjjBkzhr///e+HLlID/OEPf2DdunWoKsOHDyc9PZ1HH32UV199ldjYWFq1asW4ceNqJIZSET0mde/evbW6AwZt7NCTdhsW8eCDcNttNRxYHRXpd7dUh7U58mRlZXH88cFueCyf9cVUvmDfp4gsUNXela0btX0xxXVqwlExO6nCsyzGGBNVojZBHGzfnuNKlpO1vNjrUIwxpk6K2gRxoH17juM7VmZF7ik2Y4wJp6hNEAc7dKAry9i1188PP3gdjTHG1D1RmyByW7dmkG8uAF984XEwxhhTB0VtglC/nxOPySMxJp+5c72Oxhhj6p6oTRAA/vRj6Bq7kmXLvI7EGFNX1ER33wCTJk1i27bgvUVfdtllvPXWWzUVcthE54NypY4/nvQ3FvHhim4E70zWGBNtmjZtymK3g8B77rmHhg0bcuutt1Z5O5MmTaJXr160aNGipkOsNVF9BEG3bqSznK1bhb17vQ7GGFPXTZ48mb59+9KjRw+uvfZaSkpKKCoq4vLLL6dbt2507dqVCRMm8Nprr7F48WIuuuiiSo88PvroI3r06EG3bt245pprDtW97bbbSE9P58QTT+T2228HYMqUKXTt2pXu3btz9tlnh7290X0E0b07x/MiAFlZ0L+/x/EYYw4TYm/fFBcnhr2372XLljFt2jS+/PJL/H4/Y8eOZcqUKXTq1ImdO3eydOlSALKzs0lNTeWxxx7j8ccfp0ePHuVu8+DBg4wePZrMzEw6dep0qIvvCy64gBkzZrB8+XJE5FD33oFdfm/cuLHqjaii6D6C6NKFrg03AFDDfVwZY+qZjz/+mG+++YbevXvTo0cPPvvsM9asWUPnzp1ZuXIlN954IzNnziQlJSXkbWZlZdGlSxc6deoEON17z549myZNmhATE8M111zDtGnTDvUiG6zL73CK7iOImBg69G9Bk1nZzJ+fytixXgdkjAkU6l/6+/fnhr0vJlVl9OjR3H///T97b8mSJbz//vtMmDCBN954g2effTbkbQYTGxvL/Pnz+eijj5gyZQpPPfUUH3744WFdfg8aNIilS5fSuHHjI2pXRaL7CAKQQQPpXTyPb+ZalxvGmPINGzaMqVOnsnPnTsC52+mHH35gx44dqCoXXHDBoSFIAZKTkysdMzo9PZ1Vq1ax1h134KWXXmLw4MHs37+fffv2cc455/Dwww+zaNEi4PAuv1NTU9m8eXMYWxztRxAAAwfSh3mMXzbMBg8yxpSrW7du3H333QwbNoySkhJiY2N5+umn8fl8jBkzBlVFRPjHP/4BwFVXXcXVV19NYmIi8+bNO2zgoFJJSUlMnDiR8847j+LiYvr168c111zD9u3bOe+888jPz6ekpISHHnoIOLzL7yFDhtC1a9fwNlpVI3bKyMjQ6po1a5Yzk52tb3Ceguq8edXeXEQ41OYoYm2OPCtWrKjyOvv27QtDJHVbqG0O9n0C8zWE39ioP8VESgq92u8CwD0yNMYYg12DAKB9vxY0lmwWzLeeXY0xppQlCEBOGkQvnc+CuYVeh2KMofy7e0zVHOn3aAkCoGdP+jGXxctjOXjQ62CMiW4JCQns2rXLksQRUlV27dpFQkJCtbdhdzEBdOtGLx6mpERYtgz69vU6IGOiV5s2bdi0aRM7duwIeZ28vLwj+iGMRKG0OSEhgTZt2lT7MyxBAKSk0Lf9dtgAc+dagjDGS7GxsRx99NFVWiczM5OePXuGKaK6qTbabKeYXG0GtqNlzI82NoQxxrgsQbikT2/6l3zBV3OKvA7FGGPqBEsQpfr0oS/zWLvBz+7dXgdjjDHeswRRqmdP+sgCABYs8DgWY4ypA8KWIESkrYjMEpEsEVkuIje65feIyGYRWexOZwWsc6eIrBaRlSJyRrhiC6pBA3oe69zj6vaLZYwxUS2cdzEVAbeo6kIRSQYWiMhH7nsPq+q/AiuLSDpwMXAC0Ar4WESOUdVa62a1yaDjaff9RhYvaoMNQWqMiXZhO4JQ1a2qutCd3w9kAa0rWGUkMEVV81V1HbAaqN0bTk8+mR4lC1k0N/SByY0xpr6qlWsQItIB6AmU3kR6vYgsEZFJIlI62kVrIHAMvU1UnFBqXkYGPVnEyvVxHDhQq59sjDF1TtgflBORhsAbwE2quk9EngLuB9R9/TcwmuDndH72rL2IjAXGAqSlpZGZmVmtuHJycn62rhQX0923FC0WJk9eSHr6vmptu64K1ub6ztocHazNYRJKn+DVnYBYYCZwcznvdwCWufN3AncGvDcTGFDR9mtkPIgy1nc7R0H1yServek6K9LHCagOa3N0sDZXDV6PByEiAkwEslT1oYDylgHVfgUsc+ffAS4WkXgRORroAswLV3zladevJY1lD4sWWkdhxpjoFs5TTIOAy4GlIrLYLfsTcImI9MA5fbQe+C2Aqi4XkanACpw7oK7TWryDqZT07EFPXcjib04C4mv7440xps4IW4JQ1TkEv64wo4J1/gb8LVwxhSQjgx7M4ckVQygqAr91Z2iMiVL2JHVZvXrRM34FeYU+vvvO62CMMcY7liDKio2lZ2/nsMGeqDbGRDNLEEEcO6QlCeSy+BsbgtQYE70sQQTh75dBN5ay6At7Ws4YE70sQQTTpw89WcSirARsWFxjTLSyBBFMixb0OGor2bkJ/PCD18EYY4w3LEGUo+fARAAWza/1RzGMMaZOsARRjhPP7UgMxSz6YJvXoRhjjCfsMbByJJ1xMseykgVf2NPUxpjoZEcQ5WnRgoFJ3/LlmjRKSrwOxhhjap8liAqccsw29hQ0ZPlyryMxxpjaZwmiAicPdy5Uz35vv8eRGGNM7bMEUYEO53SlDRuZPb1+DRxkjDGhsARRAenTm1Ni5vD54mR7YM4YE3UsQVQkIYFTOmxk64FGrFnjdTDGGFO7LEFU4uTBzlc0+xPruM8YE10sQVTi+F92oRk7+Pzt3V6HYowxtcoSRCVk0EBOYTazvrKO+4wx0cUSRGWaN+e0o5axITuFdeu8DsYYY2qPJYgQDDwxB4B58zwOxBhjapEliBCccEYbktnHx2/u9ToUY4ypNZYgQhB73i84ixnMmOmz6xDGmKhhCSIUHTsyuMVKtu5raNchjDFRwxJEiE4ekQzAnI9yPY7EGGNqhyWIEKVf2pNU9jBn2navQzHGmFphCSJEMYMGMCBmLl/Pj/U6FGOMqRWWIEKVkEDP9nvI2tWcnByvgzHGmPCzBFEFQ0/3U0Qsn72x0+tQjDEm7MKWIESkrYjMEpEsEVkuIje65U1E5CMRWeW+NnbLRUQmiMhqEVkiIr3CFVt1DRx9HH4KyZyyzetQjDEm7MJ5BFEE3KKqxwP9getEJB24A/hEVbsAn7jLACOALu40FngqjLFVS1LfrvTyL2XeAp/XoRhjTNiFLUGo6lZVXejO7weygNbASGCyW20ycK47PxL4rzq+BlJFpGW44qsWEQZ32cJXOzqxf2+J19EYY0xY+WvjQ0SkA9ATmAukqepWcJKIiBzlVmsNbAxYbZNbtrXMtsbiHGGQlpZGZmZmtWLKycmp1rrdem6mMCuOx/74KQMviaxLONVtcySzNkcHa3OYqGpYJ6AhsAA4z13OLvP+Hvf1PeCkgPJPgIyKtp2RkaHVNWvWrGqtV/TDZk1hj17dd3G1P9sr1W1zJLM2Rwdrc9UA8zWE3++w/gksIrHAG8DLqvqmW/xj6akj97X0ybNNQNuA1dsAW8IZX3X42rZiSPICPl3a3OtQjDEmrMJ5F5MAE4EsVX0o4K13gCvc+SuAtwPKR7l3M/UH9qp7KqquGdprH2tzW7FplXW7YYypv8J5BDEIuBw4VUQWu9NZwHjgdBFZBZzuLgPMANYCq4HngGvDGNsR6fuLNAAWvLLS40iMMSZ8wnaRWlXnAFLO26cFqa/AdeGKpyZ1H9Wd2FsL+HLGHkbe7XU0xhgTHpF1G04dkdS8Af1SvmPW0mZeh2KMMWFjCaKahvY9yILcdPYuXON1KMYYExaWIKrp1DFHU4KPWY9+63UoxhgTFpYgqmngr9KIl3zmfGh3Mhlj6idLENUUFwf9OvzIR9u6wurVXodjjDE1zhLEERh5WSOW0J11T8/0OhRjjKlxliCOwDmXpQLw3nT1OBJjjKl5liCOQJcu0CllB++v6gQHD3odjjHG1ChLEEdABM4afJBZJYPJ/eRLr8MxxpgaZQniCJ11VRq5JPHZ5PVeh2KMMTXKEsQRGnxGAokxeczITPQ6FGOMqVGWII5QYiKcesxmZuzqBxs3Vr6CMcZECEsQNeCsCxqwhs58P2mO16EYY0yNCSlBiMiNItLIHathoogsFJHh4Q4uUpx5hdP998ypez2OxBhjak6oRxCjVXUfMBxoDlzFT+M4RL2OncS53fW7Dna7qzGm3gg1QZSO63AW8IKqfkv5Yz1EpXOGure7vp/pdSjGGFMjQk0QC0TkQ5wEMVNEkoGS8IUVec64qhV5JPL5C9YvkzGmfgh1RLkxQA9graoeFJEmOKeZjGvwsFiS/Qd57tOODFd1nqIzxpgIFuoRxABgpapmi8hlwJ8BuyIbICkJrjt9FW/mjuCHmVleh2OMMUcs1ATxFHBQRLoDfwQ2AP8NW1QR6v/ubYGgjL/bxogwxkS+UBNEkaoqMBJ4VFUfBZLDF1ZkatcnjTNTvuLNRR0oLvY6GmOMOTKhJoj9InIncDnwnoj4gNjwhRW5Rp2+jR8LmzJnerbXoRhjzBEJNUFcBOTjPA+xDWgN/DNsUUWws37fiUQO8r8JW7wOxRhjjkhICcJNCi8DKSJyDpCnqnYNIoiGJ/dkRMM5vPH5UZTYjcDGmAgWalcbFwLzgAuAC4G5InJ+OAOLWCJccL6yrbAZc55c4nU0xhhTbaGeYroL6KOqV6jqKKAv8JfwhRXZzhl/Eg3I4cVHd3kdijHGVFuoCSJGVbcHLO+qwrpRp2FaA87vsoTXVmdwcKf1zWSMiUyh/sh/ICIzReRKEbkSeA+YUdEKIjJJRLaLyLKAsntEZLOILHanswLeu1NEVovIShE5ozqNqUuuvErYTyPeemiN16EYY0y1hHqR+jbgWeBEoDvwrKreXslq/wHODFL+sKr2cKcZACKSDlwMnOCu86R7K23EOuW6bnSI2cALE9XrUIwxplpCPk2kqm+o6s2q+gdVnRZC/dnA7hA3PxKYoqr5qroOWI1znSNixTRqyKW9svh0+wlsW7TV63CMMabKKuysT0T2A8H+BBZAVbVRNT7zehEZBcwHblHVPTjPVXwdUGeTWxYsprHAWIC0tDQyMzOrEQLk5ORUe91Qdb2wiJL5Pp4Z9R6DH+sc1s8KRW20ua6xNkcHa3OYqGrYJqADsCxgOQ3w4Ry5/A2Y5JY/AVwWUG8i8OvKtp+RkaHVNWvWrGqvWxXdm2zQ3jJfS7b9WCufV5HaanNdYm2ODtbmqgHmawi/4bV6J5Kq/qiqxapaAjzHT6eRNgFtA6q2AerFo8hjr4tjvmYw69b3vA7FGGOqpFYThIi0DFj8FVB6h9M7wMUiEi8iRwNdcB7Mi3hj7mpB49j9TJyaDPn5XodjjDEhC1uCEJFXga+AY0U8xEfTAAAZHUlEQVRkk4iMAR4UkaUisgQYCvwBQFWXA1OBFcAHwHWqWi/6Q42PhwtP38O0grPY/+p0r8MxxpiQhTqiXJWp6iVBiidWUP9vONcl6p3L72jDMzNiePOfa7jiSq+jMcaY0NjT0LVg4EkxdGyyhxdX9IJ59eLMmTEmCliCqAUicNk1SXzKqWy66ymvwzHGmJBYgqgll18djxLDfz5uDYsXex2OMcZUyhJELencGU4fWsgTcj159z3odTjGGFMpSxC16M6/xLJNW/DctGawdq3X4RhjTIUsQdSioUNhUJ98/s3NFI3/l9fhGGNMhSxB1LJb/xTPBjrw1gu7YfNmr8MxxphyWYKoZb/4BXRqX8hDRTfCP/7hdTjGGFMuSxC1zOeD314Xy1cMYOUzmfDDD16HZIwxQVmC8MBFFzmvLxZdAg884G0wxhhTDksQHmjXDn79a/iX3MqOie/AsmWVr2SMMbXMEoRH/vIXyC+O5Unf7+HPf/Y6HGOM+RlLEB7p3h1GjIAn/DdQ+PZ7MHeu1yEZY8xhLEF46JprYMeBBnzY6AK46y6vwzHGmMNYgvDQ2WdD+/ZwT8pDFH8yCz75xOuQjDHmEEsQHoqLc25imr+xBU+k3OVci3DG5DbGGM9ZgvDYJZfAySfDv2JuY9/Xy2H2bK9DMsYYwBKE50Rg3DjYlN2Quxv8G26/3Y4ijDF1giWIOuCUU2DsWOGx3DEsm5sDU6Z4HZIxxliCqCv+9jdolCLc0ehJ5yGJoiKvQzLGRDlLEHVE06Zw443Ce/tO4f01XeDll70OyRgT5SxB1CG33gpduyrXxj1PwZ/vg5wcr0MyxkQxSxB1SIMG8OCDwvqC1vxz08Vw331eh2SMiWKWIOqYESPgggvgft89rP732/D9916HZIyJUpYg6qBHHoH4JB/X8xj89a9eh2OMiVKWIOqgVq3gnntjmFkynI9e2wWvv+51SMaYKGQJoo669lpo3075U4NH0ctHwcqVXodkjIkyYUsQIjJJRLaLyLKAsiYi8pGIrHJfG7vlIiITRGS1iCwRkV7hiitSxMfD3fcI8w+kM813Ppx/Phw44HVYxpgoEs4jiP8AZ5YpuwP4RFW7AJ+4ywAjgC7uNBZ4KoxxRYzLL4fjjoM7Up6iYNn3cMstXodkjIkiYUsQqjob2F2meCQw2Z2fDJwbUP5fdXwNpIpIy3DFFin8fnjoIVi1pQGjj/8KnnkG3n3X67CMMVFCNIwdw4lIB2C6qnZ1l7NVNTXg/T2q2lhEpgPjVXWOW/4JcLuqzg+yzbE4RxmkpaVlTKlmv0U5OTk0bNiwWuvWtoceOoZ3323Fuy0vZ3judL6ZOJHCJk2qvJ1IanNNsTZHB2tz1QwdOnSBqvautKKqhm0COgDLApazy7y/x319DzgpoPwTIKOy7WdkZGh1zZo1q9rr1rbdu1WPPlq11VEFuiO+teqwYar5+VXeTiS1uaZYm6ODtblqgPkawm94bd/F9GPpqSP3dbtbvgloG1CvDbCllmOrsxo3hldfhd37Yjm/wzfoxx/DdddZt+DGmLCq7QTxDnCFO38F8HZA+Sj3bqb+wF5V3VrLsdVp/frBv/4Fn61sybSRk+H55+GVV7wOyxhTj4XzNtdXga+AY0Vkk4iMAcYDp4vIKuB0dxlgBrAWWA08B1wbrrgi2W9/C127wm1LLye73xnOwxIbNngdljGmnvKHa8Oqekk5b50WpK4C14UrlvrC74fHHoPhw4XhDd/m05IONBw1Cj79FHw+r8MzxtQz9iR1hBkyxOl5Y+HyeC7tMg+dPRvGj690PWOMqSpLEBHol790rke8s6gtk/s+CXffDZ9/7nVYxph6xhJEhLrhBhg4EG7I+h2rWw+GSy+FHTu8DssYU49YgohQMTHw4ovg9wsjfdPZtb0Yfv1ryMvzOjRjTD1hCSKCdewIb74Ja7Ykcn7HBZR8Pgd+8xsoKfE6NGNMPWAJIsINGQJPPAGZWS146twPYdo0uOYaKC72OjRjTISzBFEPjB4NZ5wBf/zwNJaOeQQmTYJx4+xJa2PMEbEEUQ+IwMSJkJIinPr2DXw7/Da45x7nwrWNIWGMqSZLEPVE69bw2WeQkCAMX/wP1t/6OEyZAgMGwM6dXodnjIlAliDqkS5dYOZMyMsTBrx0HRv+MwtWrYKhQ2HPHq/DM8ZEGEsQ9Ux6uvPM3MGDcPq4wWybNMMZz3r4cPw5OV6HZ4yJIJYg6qETT4T334ctW+CM8UPZ/cLbsHgx3W+5xckcxhgTAksQ9dTAgU6fTd99B5e9PILC5/5D8vffw4gRsG+f1+EZYyKAJYh67MwzYcIE52jiN+9dyje33w9ffgmnngrbt1e+AWNMVLMEUc/99rfwwAPO0cSot29g13/fg6VLnUOMhQu9Ds8YU4dZgogCd94Jb70Fa9Y0ZMgDw/n++dlOn039+8ODD9pT18aYoCxBRImRI2HcuKWsWwcZ1/Zj/uTlTr/ht98Ow4bBxo1eh2iMqWMsQUSRvn33sGIFNG0Kg3+Zwgtn/Y+S5yfB/PnOrU8TJ3odojGmDrEEEWXatYOvvoJu3WD0GOHkSVfxw4xlcMIJcPXVTvcc2dleh2mMqQMsQUShli3hiy/ghRdg2TLoeW573vvjZ07/Ta+95mSRP/4R7ME6Y6KaJYgo5fPBlVfCggXQti2cM9LHhcvvZt1b38IvfgH//KfzWPZbb1mvsMZEKUsQUa5zZ+eU0513wowZcPz5J/DckJfROV9Ao0bwq1/BySc7hxzGmKhiCcKQmOg8K/Hdd9CnD4wdC0PvGkjmo9+iTz8Da9fCSSc5t0JlZXkdrjGmlliCMIe0aeN0Gf7MM7BkCQwd5mPwy2N57q51HLznQcjMhO7d4a67oKDA63CNMWFmCcIcJibGOYLYtMnppmPtWhh7fTydn7mNV8b/QPHIXzmHG4MGwUcf2fUJY+oxSxAmqKQk+P3vnefnZs+GVq3g0mtTOHHFa7z7x9lOBhk+3Lk+8dlnliiMqYcsQZgKiTg5YN48Z4C6oiL45YMnc2Kzzdx/5hdsX7QZhgxxKk2Z4nThYYypFyxBmJDExMBFFzn9/D38MDRKieGvHwykU8xaxgxYwcxVHSm+5FLnUOPXv4a334aSEq/DNsYcAU8ShIisF5GlIrJYROa7ZU1E5CMRWeW+NvYiNlOxuDi46SaYM8d5yG7kSOGNFcdz5vb/0ikth/EdnmZb5ndw7rlOj7H/+591BmhMhPLyCGKoqvZQ1d7u8h3AJ6raBfjEXTZ12AknwEsvwbZtTh7omJ7InYsupM3eZVzSdzWfrG5P4YW/cZ7Eu/VW54q3MSZi1KVTTCOBye78ZOBcD2MxVZCQAOefD59+6jwmceONwntZnRi26zVaNDzARTKVFY9+BMce6/Qg++yzznioxpg6TdSDu09EZB2wB1DgGVV9VkSyVTU1oM4eVf3ZaSYRGQuMBUhLS8uYMmVKtWLIycmhYcOG1Vo3UtVmm3NzY5g/vwmff96ML79sxoEDfto22M7IkmlckfscGSzgYIcO7Dj5ZHaccgoHOnVyrojXMNvP0cHaXDVDhw5dEHD2pnyqWusT0Mp9PQr4FjgFyC5TZ09l28nIyNDqmjVrVrXXjVRetXnrVtVHHlEdOVI1Lq5EQbVz0916fas3dIFkqIJqy5aqY8eqfvyxan5+jX227efoYG2uGmC+hvBb7ckpJlXd4r5uB6YBfYEfRaQlgPtqgybXEy1awI03Ov3+bd0qPP88tDi+MRP3nEeGzqd3+x081Pzv7HzpA2fwohYt4KqrYNo0WL3aubfWGFPraj1BiEgDEUkunQeGA8uAd4Ar3GpXAG/Xdmwm/Jo0gTFj4PPPYfNm+Pe/Ia9hM25ZcgUtC9YzOH07f2//NHOmbmH9eX+gqMtxzkojRjjDoy5aZLfPGlNL/B58ZhowTZzzzX7gFVX9QES+AaaKyBjgB+ACD2IztahxY7j5ZmdasQL++19h5szm/GnxhcCFAMT6iumesIl+c7+g9wcfcuzt0zg+dRupJ3dzRsEbNMjpSDA52dvGGFMP1XqCUNW1QPcg5buA02o7HlM3pKfD+PHOtHUrLF7sHGF8/72Pb75pz+T57XmC3ziVs6HV+z+SPn0pJ+gyesj/6NtyI8cfU4wcewz06wfHHAN9+3rbKGMinBdHEMZUqGVLZwpUXOxcjli9GpYvhxUr0lixrBnPLR/KwTwfbIFGP+bQYfY62j+zjk7M47jYKbRsn8/6vh/RtksCvtYt4OijnaldO4iN9aaBxkQISxAmIvh8zmMUxx4LZ599qJT8fFizxrmmsXRpQzasO4H1a47hk/U+Dub7YTWwGnwUkcaPtGAbxzCPY2Ne4ag0aN2imLRmJTRpGU/Tzo1JPSoOX3ISpKU5/Z+3aOEMnBSGW3CNqessQZiIFh/vnJ5KTy8tiQHiKSlxOpz93/8W06hRD9avEbatS2HzukS+XncMU3Ymw1acqYw0ttGKLbRhJQl8y1ExO0lLyqF5o3xaNi+iyVF+mjUXmrWOp8mxzfG1PApat3aug6SkOJPf/muZyGf/ik29FBPjnEXKyMhmyBAAH/DTQ0WFhbBzp/NA948/wu5dyu7Nuez6sYgtmxLYsrE9G7YcTX6esmN/ArtzkiAHKPMAuFBCY/bQnB00YwvNWEIzdtIsdi+NE/NJSlJSUoTUxkJK4xhSU5TU5rGktk0muU0KMUc1g+bNoVkzZ4qLq70vyZhKWIIwUSk2tuy1DgGSyq1fUADbtzvT7t1Octm5tZCd63PYubmQHVsasDO7Eev2Hss3++PZcSCJwn1+2AdsC75NoYRG7COFvTQgm1TWkyL7SY09QGp8LolxRcQn+YhLiCEuwUd8Az9xDeOIS44nvlE8ccnxxCX6iE+OI65RglOemsiq9TE0Ti4ivoGfBg2co6zUVMs9puosQRgTgrg455JEmzaBpbFA8E6HVSE3Fw4ehL17ITvbmfbuhexdxWRvPUj21jyyt+ezb7eSszeevftasPNAW9bkxrEnL5G83Fjyd/sp1KpeTE8LWpri20/LhD2kxucSH1tCQmwJifHFJMYriQnOlJQYMJ8EiUlyaIpPjCE20U9sgg9/gp+4JB/xSX7iG8YS1yCWuCS/8+omrrg4JxHb5ZvIZQnCmDAQcUblS0pyzhwdzgcku1PlVJ1TYgUFkJ+nFGQfpGDHXvJ3H6Agp4CC/fnkZ+dSsD+fgv35rP9+PU2Sm5OfU8iBnBLyDxSzOyeWHfsT2XqgEfvy48k/6GdPUSxbSuLILYknl0QOkkQuieSRWKPfhZ9C4iRgiikKmIqJ87mTv5g4Xwmx/hLi/Irfr/h9zuUcf6z89BoL/tgYd16IjRd2Z8Pc/67EHxeDP97nTjH44vz44pxlX7wff4L/p9c4H36/cwNETb7G1KUuUI+QJQhj6jgR5wgmLg4aNhRo1gA6Nyi3fmZmJkOcCy+hKyx0DncKcijJ30N+TiEH9xaSu7+I3JxicnOKKThQSH5uMcX5xRTlF5F/sIT83BLy85TC/BIK8kooyFcKCpSCfJzXAqGgAAoKhYJCobAICgpjKChyp+IYCop9FBT5KMj3sa/YT0GJnwL1U1wSQ5HGUKQ+ComlCH85U927XdkfU4wvRvH79KdXHz+9HpZYnKTn8wt+v+BzE5/PH+OU+yRoIurc+Siqupur3I7wbt4YExFiY527r3DuA0t0pzqj9DAqP985lCrIcQ+p8tH8Ar7+/Gu6p59I0cECCnOLKM4rpDivkKI8Z74ov5ji/CJnOb/IWS4o/qm8oITiwmKKC0ooKnBfC5XiwhLnvSJ1louUoiKc8iIoLlaKisQpLxGK8VGEn+ISH0UlfoqL3GWq95qHjyKJo1j8FEksxeJzXvHTLuM7uD+98u/uCFiCMMbUfYGHUWXfAvJ37SRpiMdPzh92LrA0kRWEtlyQW+V1VnTuDJwX1iZZgjDGmJpw+LnAsH/c9sxMwnv8ULdGlDPGGFOHWIIwxhgTlCUIY4wxQVmCMMYYE5QlCGOMMUFZgjDGGBOUJQhjjDFBWYIwxhgTlKiq1zFUm4jsADZUc/VmwM4aDCcSWJujg7U5OhxJm9uravPKKkV0gjgSIjJfVXt7HUdtsjZHB2tzdKiNNtspJmOMMUFZgjDGGBNUNCeIZ70OwAPW5uhgbY4OYW9z1F6DMMYYU7FoPoIwxhhTAUsQxhhjgorKBCEiZ4rIShFZLSJ3eB1PdYlIWxGZJSJZIrJcRG50y5uIyEcissp9beyWi4hMcNu9RER6BWzrCrf+KhG5wqs2hUpEfCKySESmu8tHi8hcN/7XRCTOLY93l1e773cI2MadbvlKETnDm5aERkRSReR1EfnO3d8D6vt+FpE/uP+ul4nIqyKSUN/2s4hMEpHtIrIsoKzG9quIZIjIUnedCSIiVQpQVaNqAnzAGqAjEAd8C6R7HVc129IS6OXOJwPfA+nAg8AdbvkdwD/c+bOA93FGaewPzHXLmwBr3dfG7nxjr9tXSdtvBl4BprvLU4GL3fmngf9z568FnnbnLwZec+fT3X0fDxzt/pvwed2uCto7GbjanY8DUuvzfgZaA+uAxID9e2V928/AKUAvYFlAWY3tV2AeMMBd531gRJXi8/oL8mCHDABmBizfCdzpdVw11La3gdOBlUBLt6wlsNKdfwa4JKD+Svf9S4BnAsoPq1fXJqAN8AlwKjDd/ce/E/CX3cfATGCAO+9360nZ/R5Yr65NQCP3x1LKlNfb/ewmiI3uj57f3c9n1Mf9DHQokyBqZL+6730XUH5YvVCmaDzFVPoPr9QmtyyiuYfUPYG5QJqqbgVwX49yq5XX9kj7Th4B/giUuMtNgWxVLXKXA+M/1Db3/b1u/Uhqc0dgB/CCe1rteRFpQD3ez6q6GfgX8AOwFWe/LaB+7+dSNbVfW7vzZctDFo0JItg5uIi+11dEGgJvADep6r6KqgYp0wrK6xwROQfYrqoLAouDVNVK3ouYNuP8RdwLeEpVewIHcE49lCfi2+yedx+Jc1qoFdAAGBGkan3az5WpahuPuO3RmCA2AW0DltsAWzyK5YiJSCxOcnhZVd90i38UkZbu+y2B7W55eW2PpO9kEPBLEVkPTME5zfQIkCoifrdOYPyH2ua+nwLsJrLavAnYpKpz3eXXcRJGfd7Pw4B1qrpDVQuBN4GB1O/9XKqm9usmd75seciiMUF8A3Rx74aIw7mg9Y7HMVWLe0fCRCBLVR8KeOsdoPROhitwrk2Ulo9y74boD+x1D2FnAsNFpLH7l9twt6zOUdU7VbWNqnbA2XefquqlwCzgfLda2TaXfhfnu/XVLb/YvfvlaKALzgW9OkdVtwEbReRYt+g0YAX1eD/jnFrqLyJJ7r/z0jbX2/0coEb2q/vefhHp736HowK2FRqvL9B4dFHoLJw7ftYAd3kdzxG04yScQ8YlwGJ3Ogvn3OsnwCr3tYlbX4An3HYvBXoHbGs0sNqdrvK6bSG2fwg/3cXUEec//mrgf0C8W57gLq923+8YsP5d7nexkire3eFBW3sA8919/RbO3Sr1ej8D9wLfAcuAF3HuRKpX+xl4FecaSyHOX/xjanK/Ar3d728N8DhlbnSobLKuNowxxgQVjaeYjDHGhMAShDHGmKAsQRhjjAnKEoQxxpigLEEYY4wJyhKEqZfc3k+vrea6M0QktZI694nIsOpF97Nt/anM8pc1sV1jjpTd5mrqJbdvqumq2jXIez5VLa71oMohIjmq2tDrOIwpy44gTH01HugkIotF5J8iMkScsTNewXnICBF5S0QWuGMOjC1dUUTWi0gzEekgztgLz7l1PhSRRLfOf0Tk/ID694rIQrfv/ePc8uZuf/4LReQZEdkgIs0CgxSR8UCiG+fLblmO+zpERD4Tkaki8r2IjBeRS0Vknvs5nQI+5w0R+cadBoX92zVRwRKEqa/uANaoag9Vvc0t64vz5Hy6uzxaVTNwnja9QUSaBtlOF+AJVT0ByAZ+Xc7n7VTVXsBTwK1u2d04XT70AqYB7cqupKp3ALlunJcG2W534EagG3A5cIyq9gWeB37v1nkUeFhV+7jxPV9OjMZUib/yKsbUG/NUdV3A8g0i8it3vi1OMthVZp11qrrYnV+A03d/MG8G1DnPnT8J+BWAqn4gInuqEfM36nb9LCJrgA/d8qXAUHd+GJAeMFhYIxFJVtX91fg8Yw6xBGGiyYHSGREZgvPDOkBVD4pIJk5/PmXlB8wXA4nlbDs/oE7p/6uqDe9Y8XbBGf8iP2C+9HNicNqRWwOfZ8whdorJ1Ff7cYZhLU8KsMdNDsfhDOFY0+YAFwKIyHCcDvaCKXS7ba+uD4HrSxdEpMcRbMuYQyxBmHpJVXcBX4gz4P0/g1T5APCLyBLgfuDrMIRxL043zAtxBrvZipO4ynoWWFJ6kboabgB6izOQ/Qrgd9XcjjGHsdtcjQkTEYkHilW1SEQG4IwIZ3/dm4hh1yCMCZ92wFQRiQEKgGs8jseYKrEjCGOMMUHZNQhjjDFBWYIwxhgTlCUIY4wxQVmCMMYYE5QlCGOMMUH9P9V3aKvg325OAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "loss_train =  29.11080105299873\n",
      "loss_val =  30.230531139890644\n"
     ]
    }
   ],
   "source": [
    "plt.title('relation between\\nloss and training times')\n",
    "plt.xlabel('training time')\n",
    "plt.ylabel('loss')\n",
    "plt.plot(train_loss_tuple,label= \"Train loss\",color = 'red')\n",
    "plt.plot(test_loss_tuple,label= \"Test loss\",color = 'blue')\n",
    "plt.legend(loc='upper right')\n",
    "plt.grid()\n",
    "plt.show()\n",
    "print('loss_train = ',loss_train1)\n",
    "print('loss_val = ',loss_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}