DR_prognosis.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "DR-prognosis.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "qNHOhi54V_-A"
      },
      "source": [
        "!pip install scikit-survival"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "D8qSldV6e8qf"
      },
      "source": [
        "#All the imports\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib.image as mpimg\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import sklearn as sk\n",
        "\n",
        "from sksurv.linear_model import CoxnetSurvivalAnalysis\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "\n",
        "#-----------Other-----------\n",
        "# Make numpy printouts easier to read.\n",
        "np.set_printoptions(precision=3, suppress=True)\n",
        "\n",
        "#IO for the files\n",
        "from google.colab import files\n",
        "import io\n",
        "\n",
        "#Clear output\n",
        "from IPython.display import clear_output"
      ],
      "execution_count": 7,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0vzc3XX9fAT9"
      },
      "source": [
        "#Reading csv from upload\n",
        "uploaded = files.upload()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XQMtfIEgfCDZ"
      },
      "source": [
        "#All the column names\n",
        "column_names = ['ID', 'Laser Type', 'Eye', 'Age', 'Type', 'Treated Group', 'Treated Status', 'Treated Time', 'Untreated Group', 'Untreated Status', 'Untreated Time']\n",
        "raw_ds = pd.read_csv('drdata.csv', na_values=\"NaN\")\n",
        "raw_ds.columns = column_names"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GmSco9stoTdR"
      },
      "source": [
        "dataset = raw_ds.copy()\n",
        "dataset2 = raw_ds.copy()"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8Ax09_8RoajW"
      },
      "source": [
        "dataset = dataset.drop(columns=['ID','Treated Group', 'Treated Status', 'Treated Time', 'Laser Type'])\n",
        "dataset['Untreated Status'] = (dataset['Untreated Status'] == 1).astype(bool)\n",
        "\n",
        "dataset2 = dataset2.drop(columns=['ID','Untreated Group', 'Untreated Status', 'Untreated Time'])\n",
        "dataset2['Treated Status'] = (dataset2['Treated Status'] == 1).astype(bool)"
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hNUUjHwap1kp"
      },
      "source": [
        "dataset2"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NgZvKA0Tp7YF"
      },
      "source": [
        "X = dataset.iloc[:,:-2]\n",
        "Y = dataset.iloc[:,-2:]\n",
        "Y = Y.to_records(index=False)\n",
        "\n",
        "X2 = dataset2.iloc[:,:-2]\n",
        "Y2 = dataset2.iloc[:,-2:]\n",
        "Y2 = Y2.to_records(index=False)"
      ],
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a8jE4KkiqWFL"
      },
      "source": [
        "# One Hot Encodings\n",
        "non_dummy_cols = ['Age'] \n",
        "dummy_cols = list(set(X.columns) - set(non_dummy_cols))\n",
        "X = pd.get_dummies(X, columns=dummy_cols)\n",
        "\n",
        "dummy_cols2 = list(set(X2.columns) - set(non_dummy_cols))\n",
        "X2 = pd.get_dummies(X2, columns=dummy_cols2)"
      ],
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "plPv3N2zoUsh"
      },
      "source": [
        "X2"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jG__QlGiqeP6"
      },
      "source": [
        "random_state = 20\n",
        "\n",
        "X_train, X_test, Y_train, Y_test = train_test_split(\n",
        "    X, Y, test_size=0.2, random_state=random_state)\n",
        "\n",
        "random_state = 20\n",
        "\n",
        "X2_train, X2_test, Y2_train, Y2_test = train_test_split(\n",
        "    X2, Y2, test_size=0.2, random_state=random_state)"
      ],
      "execution_count": 30,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ra4UBaFJqltE"
      },
      "source": [
        "coxnetUT = CoxnetSurvivalAnalysis(l1_ratio=0.9, fit_baseline_model=True)\n",
        "coxnetUT.fit(X_train, Y_train)\n",
        "\n",
        "coxnetTR = CoxnetSurvivalAnalysis(l1_ratio=0.9, fit_baseline_model=True)\n",
        "coxnetTR.fit(X2_train, Y2_train)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mulXC2jAqnvU"
      },
      "source": [
        "from joblib import dump, load\n",
        "dump(coxnetUT, 'coxnetUT.joblib') \n",
        "dump(coxnetTR, 'coxnetTR.joblib') "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R5psUoq9q2-4"
      },
      "source": [
        "surv_funcs = {}\n",
        "for alpha in coxnetUT.alphas_[:5]:\n",
        "    surv_funcs[alpha] = coxnetUT.predict_survival_function(\n",
        "        X_train.iloc[150:151], alpha=alpha)\n",
        "    \n",
        "surv_funcs2 = {}\n",
        "for alpha in coxnetTR.alphas_[:5]:\n",
        "    surv_funcs2[alpha] = coxnetTR.predict_survival_function(\n",
        "        X2_train.iloc[150:151], alpha=alpha)"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 285
        },
        "id": "EVeInEIeo59o",
        "outputId": "f76f4880-1dc2-4fa5-fd03-b65eb1861d48"
      },
      "source": [
        "for alpha, surv_alpha in surv_funcs.items():\n",
        "    for fn in surv_alpha:\n",
        "        plt.step(fn.x, fn(fn.x), where=\"post\")\n",
        "\n",
        "for alpha, surv_alpha in surv_funcs2.items():\n",
        "    for fn in surv_alpha:\n",
        "        plt.step(fn.x, fn(fn.x), where=\"post\")\n",
        "\n",
        "plt.ylim(0, 1)\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "No handles with labels found to put in legend.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD8CAYAAABn919SAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3de3iU9Z338fc3k5CQAwQBgSSEw2NAoVKpgSgVSj10wR641kofZXWlS6G7rfSwtLt299FVe+1eba2lrmuttHhofVALj1SsB0SLFUWQoJEz4aBAwiHhEOQUSMLv+WNm6BBymCQzmXvufF7XlcvMPffc8y1TP/z8zu/+/cw5h4iIJL+URBcgIiKxoUAXEfEJBbqIiE8o0EVEfEKBLiLiEwp0ERGfaDXQzexxM6sysw3NPG9m9t9mtt3M1pnZZ2JfpoiItCaaEfqTwKQWnp8MFIV+ZgGPdrwsERFpq1YD3Tn3FnC4hVOmAL9zQauAXDMbEKsCRUQkOqkxuEY+sCficUXo2L7GJ5rZLIKjeLKysq689NJLY/D2sPWTIwAM79ErJtcTEfGqtWvXHnTO9W3quVgEetScc/OAeQDFxcWutLS0zdd4/Ov34872AyCQu4vpD/0XE15bSHlaEdcfWclPbvpWTGsWEfESM9vV3HOxmOVSCQyMeFwQOhZXdWn5NNQMAmDc8WoAVmY3+ZeWiEiXEItAXwL8fWi2y1XAUefcBe2WWPmHJ+5hxlPfJK2ukrq0fObf8RiX/qWCYXXb4vWWIiJJodWWi5k9A0wE+phZBfAfQBqAc+7XwMvAjcB24CTw9XgVGymQuwtqgiN1ajrjHUVEvM0StXxue3vojc2/4zFquxfx3E2HKE8rYrr66CKSJOrq6qioqKC2tvaC5zIyMigoKCAtLe2842a21jlX3NT1OvVL0XgI5O6C00WM3r+b8oFFPNlrHCtfW9jia8Ydr1boi0jCVVRUkJOTw+DBgzGzc8edcxw6dIiKigqGDBkS9fWS/tb/6Q/9FxmntjFs5WhuqXi91V56RWqevjwVEU+ora2ld+/e54U5gJnRu3fvJkfuLUn6ETrA6cyd4Iq4/PUcfvnU1BbPnfDaQipS85jQxCheI3cR6WyNw7y14y1J+hE6QO70b5NxMrpZLuOOV1NQv/eC4xq5i0iy88UIfVpJIfN/Fd25zY3AJ7y2EIfjpYef5ouzb4thdSIincMXI/RYSEkxtqUN45XeBxJdioh0Ic3NNGzPDEQFeshXB2QBsGjAddz1fJTDfRGRDsjIyODQoUMXhHd4lktGRkabrueLlksszB45mcqtvwpOe1QvXUQ6QUFBARUVFVRXV1/wXHgeelsk/Y1FYfPveIy6tHzS6irPLdrVHuHZL/+TksOo61taBl5EpPO1dGORb1ougdxdpNVVUtu9iBOnr2f+HY+x6If3tOta5WlFPFm1McYViojEl29G6GG/+qd/Jf3EUGoziwAYO/wYY74/JerX3/3KE/wmYzTAuZuUND9dRLzC17f+N5Y7/du8UFbJmLIn6O5uYWNZNWPa8PofT/46dc//6lwfvSI1j3eyHY/c88B55w3uO0DTG0XEU3wX6NNKCplWUsjPnnmCtFfLaaDtd1tFjsYnvr6I3an5PHdNxDoLOD59cA9fjEnFIiKx4buWS6TIL0oBel5cyc0P3N+mazy88RWWHKo779j6hkKAJteNUXtGROKpS7VcIp3O3En6CcCgtnsRtceKeGbmAgAuGZEVVW999sjJzG507K6IlkykitQ8Vuee5sXnryHtZDGTbvtlDP5XiIhEx9cj9AWrd/NCWXB0PqbsCXqe/AwpGLXdg1+Y5hXlMmxsP0aOz4/J+33u1VfZndaDQXzMmFMf8sCXfhqT64qIhHXZEXq4nw7ws2eeYNmxXwPwpaXF0K2Y/VsKqNqxnpHj74jJ+31jVDHPHzhC2WE42y2t9ReIiMSQr0fozXn2B1+mR9kOjvT5DsezC0it30/PIae4+T8aN1faZ8JrCwmknuZf95aq7SIiMdVlR+jNueXnLwLwu1n/SGot1HUr4OjOC5fUbS87m8NmK+KN3A0M/Lc/tnp+6rBshk+/PmbvLyJdU5cM9LDhNxSRve1pVm25jTPpBSx+8P2Y9NS/MaqYH26t4PdZt7B6YuvrtF91cDM/Q4EuIh3TpQO9ZOocYA7rb5pJWh3s3wJHPt7IyPG3d+i6t+f1YceOV1l+OgNSWp4HvydlAKv6dOjtRESALtpDb+zRb01i0NbdHOn7HWpyh9FvwMaY9dNbE14M7Dd7e6ntIiKt6hKLc3VEzzvm8dTX5lOXXQbAgX0jeWbmAtbMfSH+b27BxcDmd9sc//cSEV9ToBOc3vjcN6/msr8pot6eptupcg4H+vPe1hw2rqiM63tfm3EagN8NGK+NNUSkQxToEUqmzuH0hFSWffbX5FQH7yhd/Yc/x/U97x1/G9OPrATgnew+vPTw03F9PxHxLwV6I/9y6zz+MKuME/0qyK0p51RdPovueziu7/mTm77FpQ3btaepiHSIAr0ZPe+Y16k99fCepuv6FMbtPUTE3xTozZhWUtipPfXZIyc3uXqjiEi0FOgtSERPHWDd66/G/T1ExH8U6K1o3FOvO9WL+dN/G7e+uvYzFZH2UqBHKdxTTz1TQW3G0Lj01T/XcByAD/r349Wnvxez64pI16BAj1K4p37VpU9f0Fdf/OD7Memt/3jy1xl+ej+b7VO8kZseg6pFpCtRoLdBydQ5jPy3t9lxeR1Lr36UnOoF5BwtZ/+WvTHrrX9jVPCO3tXdmryzV0SkWV16ca72uurK/+SFskoO9LybkbtWEnDfoSYwjEX3PdzhNWBuz+vDbzYsJyVQx4vPX9Pm12vrO5GuS4HeDuGdkBZc8ThPlVUyec1TwDCO7oxNm8TO5rAr0IMHsr/bptelBOoY023deWuwa611ka4jqtUWzWwS8BAQAH7rnPtJo+cLgaeA3NA5dznnXm7pml5abbGjVi98kA0vZlGbOYyJfze8w+up/37vQZ4/cKTNr3u35gTAefPZSw7spXjnvgvOveyzExl1/aT2FykiCdGhHYvMLAA8AtwAVABrzGyJc25TxGn/B/iDc+5RMxsBvAwM7nDlSaJk6hw+emYmtZnDWL34XUaOv7lD17s9rw+357V9kfR7Vzx93hrse1IGEMg/zYSsheed5xzs2FalQBfxmWhaLmOB7c65nQBm9iwwBYgMdAf0CP3eE4jdfm5JIrXPGnJrrqT2VOK+Z753/G3cG/H42mXL+CgwlAdyzm/dpKTUMSZ9HcOa2B5PLRqR5BVNoOcDeyIeVwAljc65F3jNzGYDWdD0fmpmNguYBVBY6K81SwbecDsbXoTg323e8PWRo0Otm77nHX+35gRbskY0uT1eyYGNFN/3etTvodaNiHfE6kvRW4EnnXMPmtnVwO/N7FPOubORJznn5gHzINhDj9F7e0LJ1Dls+NNjnOmWz+IH3z93PBZ7lLZXc62bxq2ZsOZaNM1R60bEW6IJ9EpgYMTjgtCxSDOASQDOuXfNLAPoA1TFoshkkXWslLR0x+H1+wGozRjK3m01AAkL9aY0bs2ENdeiaU5zrRu1bUQSI5pAXwMUmdkQgkF+CzCt0Tm7geuAJ83sMiADqI5lockgtc8acnes4KwF++jdGsaxdfi0mHxR2hmaa9E0p7nWzVUHN/OzprtuIhJHrQa6c67ezO4ElhKckvi4c26jmd0PlDrnlgBzgN+Y2fcJNpGnu0TtPp1AA2+4nezBi8893rdzGbk1xRyvL+CZmQvOO/eSEVmM+f6Uzi6xRW2dXdNU62ZPygBWtX2CjojEQFTz0OPBT/PQmzNv7t2cKq8k8+RYUvhr6NV2LwIgryg3oT32eJjwWrD//uibaWq9iMRBh+ahS/tlj5vJm3U/4HSPRwlEjGK/tLQYuhWzd9sw9m6rofy94LZzvgh3g4pAHv80ca9aLyKdTIEeR9NKCoGf80JZ5XmzGbvl/iO5H79N3v5rONCvmNotPfgk5SLqq6uTPtCvzTjN8tP72JOap9aLSCdTyyUBVi98kOxti1lWWcWg8gCZpHCw92xqcocxdvgxz/XW2yOy9dLZ1OoRP1PLxWNKps4B5vDYr27n7cvXkZJiXLpuLf3qh7F53UHGJLrAWIhovXQ2tXqkq9J66Al01ZX/ScB+jblH2V54lLRT5dSdrU90WTFxbcZpBp7dh6VYp/5UpObxbp9+vPTw04n+IxDpdGq5eMSC1bs59cgrYDDjqW8mupykNfH1RWwJXMJX977OhO0HtTSB+E5LLReN0D1iWkkhWOvnScu+OiALgE0D+tP9U2+zY9sfElyRSOdRoHtMbfeimOxP2lXNHjmZEfXVfMxQfpH7LVYOvJitT0a/2JhIMtOXoh6SdayU2u5FSbNUgFeFlzB4tyaTLVkj4Mhf2rSCZGNq20iyUKB7yLk11Wv1H04dEV7C4N4VT/Pr+k+xqt8ABm87065rnT17loa3NijQJSko0D3k8GVfJvABwXVppcPuHX8bf3njeXal5vPc+PZ9QeGcY9TB3QTuu0sjdfE8DQU9JHvcTMDUR4+hm/p353+lHiI9Pb1dP5Wp+Wzqry9YJTlo2qLHPHvTTA5dfCvdMw/zD79QHz3RgmvEZzHIfcyYU+uYvvKqNr1ed61KrOlO0SQS7qOHl9z14jK7XUn4C9ayI0B3mFD88+hf7IzAgVEM112r0kk0QveYyCV3z3QfBsBFDfvbfB39RRBb1735Z3ae7U7h2ehbYWZnGHPqQ6avvLpD761RvkTSCD2JRC65O3LDWAqOXclJa9vHVJfan80bfLImjEdMHzYqtJtT9EtIlh2pbvuovjGN8qUNNEL3oAWrd/NCWSXFx2ezPvs4gfTsNr1+3PKZpJgx48lZcapQotGeUX1jZmcoPvUhY9a2PktHs3C6Bo3Qk8y0kkKmlRQyb+5kZu97ncxugTa9fhWO2owi1sx9QW2XBGrPqL6xssPVNHRLZUjDrhbP03x5AQW6p2WPm8nPy25s8+vGZj5BhhvGe1tzqHjwfX/shJSE2rpHa1M+/8b6qObRO+f49ME9F+zeLl2LAt3DwiP1tnrol98lZ8XZ87a5AxTqSeim/t1ZcugQpKa3eN6O+t5YH2Pjv/2xkypLLH1R3DQFug9dlf+/Odbrv2n4KLjN3dbh09j00gYFehKaPXIys6M4b8KyhVSkJmZDkUTQJiZN052iPlQydQ47R/49F43vxYYRb5FbU05VTRpr5r6Q6NIkThK1oUgifipS81jVp3+i/8g9SbNcfCo8UyZwdC6X7O5Jv/pbAcgrym3yfPXZJVmE96t96wtTE1xJYmiWSxcU7r8vWJ1PxVt/S84qB92KObz+wo/8TLcBHKncqSV7RZKcAt3nppUUMm/ljeT1+h0NH70NKRdOgTzYezbHzwaXGgDdZSqSrBToXUD2uJm8cfAEU/JWkpNx4Ue+ec8ask4aJy1AbcZQ3tsKmSsq1YIRSTIK9C5gWkkhC/gec8ua7jlemXYn7+e8Q21KBl9aWsyxvtO0a5J43nP33ZXoEtrt4kFD+fz02N/JrUDvIlqa07564TS+uG0xAPt6LeNATbF2TRJPczgONvROdBntVrenLi7XVaALJVPnAHMAeGfu3do1STwtEAiwy/q1excqLxhU9wm3xeG6CnQ5T/a4mZz64JVzuyapjy5eE+3ds17Ws2evuFxX89DlAto1ScS7NA9d2iS8a5L66CLJRYEuFzh82ZcJfABn0vJY/OD7rZ6vu0xFvEGBLhfIHjeTlLd+TFodHF7f8vZ3tRlD2buthvL3DijYRRIsqkA3s0nAQ0AA+K1z7idNnPM14F7AAR8657Q0c5KaVlLIor5ryN25ArOWN9cI1F/NgX7FVJUPpr66WoEukkCtBroF/41+BLgBqADWmNkS59ymiHOKgB8Bn3XOHTGzi+NVsHSOw5d9mUF5re+WtLTybYaUrwS+w+GGvy4fAFpCQKSzRTNCHwtsd87tBDCzZ4EpwKaIc2YCjzjnjgA456piXah0rmh3S7oy7U4WX36CwRvWUnCMcxtaawkBkc4XTaDnA3siHlcAJY3OGQZgZu8QbMvc65x7tfGFzGwWMAugsLDtO/FI54l2t6TwXaabhrzEy1lvQkZPAK5bMpqjF9+qjTVEOlGsvhRNBYqAiUAB8JaZXe6cq4k8yTk3D5gHwXnoMXpvSaDwXabvzL2bb+3/a4tm2YC/0PfQlZw6nZbYAkW6kGgCvRIYGPG4IHQsUgWw2jlXB3xkZuUEA35NTKoUz2vcogmkzmXicWhI7XfB1EfNhhGJj2gCfQ1QZGZDCAb5LXDB5uJ/BG4FnjCzPgRbMDtjWah4W+MWzYLV+aS8+2POZJw/9TFymmNLFPoibddqoDvn6s3sTmApwf744865jWZ2P1DqnFsSeu4LZrYJaAB+6Jw7FM/Cxduam/oYnuZYu6VHs6/9JOUiTYEUaQet5SJxM2/u3Yw+ev7Ux6WVBxlSnkJmt+xmX1fVcwYnsgfSb3h/jdRFGtFaLpIQTU19LE6fzeJRxwmkn232dZ9/cQ2ZzrF3W8YF7RkFvEjzFOgSN01NfZw3dzKz97V8w9K+Xsto+Oht8vZfc157Rq0YkZYp0KVTRXPD0tATC5mSt5KVh4J3oWYeC7Zn6nvOoMq0TrtIc9RDF89ZsHo3L5RVUnx8NuuzjxNIDwb651+8gmN9gxOs8opym3292jLiZ+qhS1IJt2oat2f29VoGW2lxlozaMtKVKdDFsxq3Z4aeWEi/9D8xKKIN01h9zxkccgOjWsc9WhrxS7JQoItnNXWzUsVbf2HxqBPNzpIJz5A5vH5vTGo4020ARyp3MnK8tuIT71MPXZJKU3PbI+3buYeGj7pDSsvL/kbrYO/ZHM8uoId9EvVrtGywxJN66OIbrc2SCc+QycmIzf+1N+9ZQ9ZJ42QrG32EhZcNrgi1fNSukc6kQJek0tqyvgtW5zO3bGrM3q84fTYrst85N9OmNdctGY3LKKZ2S46+oJVOp0AXX4l2HfdoRXMjVKSlA1YwpPwdMo9ma968dDoFukgLot25KSxyaYPrlpRCjyJt8iGdRl+KisRQ5Je2SysP0u/Q9zjWo1ALjUnM6EtRkU4SOaIvTp9Nw/5Sso4bVeUp6qdL3KUkugARP5lWUshz37ya5755NT1TJnO4/1tkVz9E5rFdHDrcwOIH32fjisYbfonEhgJdJE6yx83kzOAcFt98FqstJev4HqrKq9j00oZElyY+pUAXiZNpJYXkpt7I7H29qBqw4txIvaomTaN0iQsFukgcZY+byc8H/IKGodnnRuqARukSFwp0kTgK99R7pkw+N1LPrSnnVPW+RJcmPqRZLiKdIDz7pThnNieOw1nrwzMzF7T6Oq0LI22hQBfpBJFrvJ/KfI/MKNaHqUvtz+YNBxnTSTVK8lOgi3Si7HEzqan/W9b2WNnq+jDjls/kZFp+qyN5jeIlTIEu0ommlRQyb+WNUa0PsyzzPbJOjm1xJK9RvERSoIt0smjXh4lmpcdxy2dSZxbL8iSJKdBFOlm0K0JGs9LjKhwkZjkm8SAFuohHRTOS/4Jbx5n0/JjuodqYFhVLHgp0EY+KZiT/7NLH6VbnqN0S/RZ5baFNOpKLAl0kie3KW8WwrW+T+UlOXK5f33MGp+rT43JtiT0FukgSy7ikO4sv/4RA+tm4XH/ccqgP9I1bS0ftnNhSoIskse7ub9q0RV5bbTz2HqfPOGq3HI35tdXOiT0FukgSa+sWeW01esBMhpevjEtLR+2c2FOgiySxWG+K3dhDv8xk8aj4tHTGLYeUhjMxv25XpkAXkWbFs6WjOfSxp0AXkWbFs6XzBbcOdJNrTJlzifkrsri42JWWlibkvUUk8ebf8Rh1afn0tObn0GvhsQuZ2VrnXHFTz0W1wYWZTTKzrWa23czuauG8r5qZM7Mm30xEJKwmay2BugpO1p1s8ueo68HmDUcSXWZSabXlYmYB4BHgBqACWGNmS5xzmxqdlwN8F1gdj0JFxF9Sh5WxovtbzS4+poXH2i6aHvpYYLtzbieAmT0LTAE2NTrvx8BPgR/GtEIR8aXWvnDVl6ZtF02g5wN7Ih5XACWRJ5jZZ4CBzrmXzKzZQDezWcAsgMLC+E21EhHva+0LV31p2nYdnuViZinAL4DprZ3rnJsHzIPgl6IdfW8RSV6tzaGfv3IdZ7rFdyXJROgzMJvxXxsWl2tHE+iVwMCIxwWhY2E5wKeANy3Y7+oPLDGzrzjnNI1FRNol61gpaemOw+v3J7qUmDq24xPGf+2f43LtaAJ9DVBkZkMIBvktwLTwk865o0Cf8GMzexP4gcJcRDoitc8acneuwFrZTDvZ1OXnAAkKdOdcvZndCSwFAsDjzrmNZnY/UOqcWxKXykSkSzt82ZcZlBe/hccS5VjuZXG7dlQ9dOfcy8DLjY7d08y5Ezteloh0dfFeeCxRRuT14Ko4XVu3/ouIJ8V74TE/iupOURER8T4FuoiITyjQRUR8QoEuIuITCnQREZ9QoIuI+IQCXUTEJxToIiI+oUAXEfEJBbqIiE8o0EVEfEKBLiLiEwp0ERGfUKCLiPiEAl1ExCcU6CIiPqFAFxHxCQW6iIhPKNBFRHxCgS4i4hMKdBERn1Cgi4j4hAJdRMQnFOgiIj6hQBcR8QkFuoiITyjQRUR8QoEuIuITCnQREZ9QoIuI+IQCXUTEJxToIiI+oUAXEfGJqALdzCaZ2VYz225mdzXx/D+b2SYzW2dmb5jZoNiXKiIiLWk10M0sADwCTAZGALea2YhGp30AFDvnRgGLgJ/FulAREWlZNCP0scB259xO59wZ4FlgSuQJzrnlzrmToYergILYlikiIq2JJtDzgT0RjytCx5ozA3ilqSfMbJaZlZpZaXV1dfRViohIq2L6paiZ3QYUAw809bxzbp5zrtg5V9y3b99YvrWISJeXGsU5lcDAiMcFoWPnMbPrgX8HPuecOx2b8kREJFrRjNDXAEVmNsTMugG3AEsiTzCz0cBjwFecc1WxL1NERFrTaqA75+qBO4GlwGbgD865jWZ2v5l9JXTaA0A2sNDMysxsSTOXExGROImm5YJz7mXg5UbH7on4/foY1yUiIm2kO0VFRHxCgS4i4hMKdBERn1Cgi4j4hAJdRMQnFOgiIj6hQBcR8QkFuoiITyjQRUR8QoEuIuITCnQREZ9QoIuI+IQCXUTEJxToIiI+oUAXEfEJBbqIiE8o0EVEfEKBLiLiEwp0ERGfUKCLiPiEAl1ExCcU6CIiPqFAFxHxCQW6iIhPKNBFRHxCgS4i4hMKdBERn1Cgi4j4hAJdRMQnFOgiIj6hQBcR8QkFuoiITyjQRUR8QoEuIuITCnQREZ+IKtDNbJKZbTWz7WZ2VxPPp5vZc6HnV5vZ4FgXKiIiLWs10M0sADwCTAZGALea2YhGp80AjjjnLgHmAj+NdaEiItKyaEboY4HtzrmdzrkzwLPAlEbnTAGeCv2+CLjOzCx2ZYqISGtSozgnH9gT8bgCKGnuHOdcvZkdBXoDByNPMrNZwKzQw+NmtrUdNfdpfF0PUo0d5/X6QDXGitdr9Fp9g5p7IppAjxnn3DxgXkeuYWalzrniGJUUF6qx47xeH6jGWPF6jV6vL1I0LZdKYGDE44LQsSbPMbNUoCdwKBYFiohIdKIJ9DVAkZkNMbNuwC3AkkbnLAHuCP1+M/Bn55yLXZkiItKaVlsuoZ74ncBSIAA87pzbaGb3A6XOuSXAfOD3ZrYdOEww9OOlQy2bTqIaO87r9YFqjBWv1+j1+s4xDaRFRPxBd4qKiPiEAl1ExCeSKtBbW4IgEczscTOrMrMNEccuMrNlZrYt9M9eCaxvoJktN7NNZrbRzL7rwRozzOw9M/swVON9oeNDQktJbA8tLdEtUTWG6gmY2Qdm9ieP1vexma03szIzKw0d88znHKon18wWmdkWM9tsZld7qUYzGx768wv/fGJm3/NSjS1JmkCPcgmCRHgSmNTo2F3AG865IuCN0ONEqQfmOOdGAFcB3w79uXmpxtPAtc65TwNXAJPM7CqCS0jMDS0pcYTgEhOJ9F1gc8Rjr9UH8Hnn3BUR86a99DkDPAS86py7FPg0wT9Pz9TonNsa+vO7ArgSOAks9lKNLXLOJcUPcDWwNOLxj4AfJbquUC2DgQ0Rj7cCA0K/DwC2JrrGiNpeAG7wao1AJvA+wbuRDwKpTX3+CairgOC/yNcCfwLMS/WFavgY6NPomGc+Z4L3p3xEaDKGF2tsVNcXgHe8XGPjn6QZodP0EgT5CaqlNf2cc/tCv+8H+iWymLDQKpijgdV4rMZQO6MMqAKWATuAGudcfeiURH/evwT+BTgbetwbb9UH4IDXzGxtaJkN8NbnPASoBp4Ita5+a2ZZeKvGSLcAz4R+92qN50mmQE9KLvhXesLnhppZNvD/gO855z6JfM4LNTrnGlzwP3MLCC4Id2ki64lkZl8CqpxzaxNdSyuucc59hmBb8ttmNiHySQ98zqnAZ4BHnXOjgRM0al14oEYAQt+HfAVY2Pg5r9TYlGQK9GiWIPCKA2Y2ACD0z6pEFmNmaQTD/P86554PHfZUjWHOuRpgOcEWRm5oKQlI7Of9WeArZvYxwdVGryXYC/ZKfQA45ypD/6wi2Pcdi7c+5wqgwjm3OvR4EcGA91KNYZOB951zB0KPvVjjBZIp0KNZgsArIpdCuINg3zohQssYzwc2O+d+EfGUl2rsa2a5od+7E+zxbyYY7DeHTktYjc65HznnCpxzgwn+/+7Pzrm/80p9AGaWZWY54d8J9n834KHP2Tm3H9hjZsNDh64DNuGhGiPcyl/bLeDNGi+U6CZ+G7+kuBEoJ9hf/fdE1xOq6RlgH1BHcAQyg2B/9Q1gG/A6cFEC67uG4H8ergPKQj83eqzGUcAHoRo3APeEjg8F3gO2E/xP33QPfN4TgT95rb5QLR+GfjaG//3w0uccqucKoDT0Wf8R6OXBGrMILi7YM+KYp2ps7ke3/ouI+EQytVxERKQFCnQREZ9QoIuI+IQCXUTEJwQ4gdkAAAAWSURBVBToIiI+oUAXEfEJBbqIiE/8f09T56/0abfuAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}