LogReg Breast Cancer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "from sklearn.model_selection import train_test_split\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Categorizing Data:\n",
    "<p> The 'id' column is dropped as it has no use in the process. </p>\n",
    "<p>Categorical data are variables that contain label values rather than numeric values.The number of possible values is often limited to a fixed set.</p>\n",
    "\n",
    "### Splitting Data:\n",
    "<p>The data we use is usually split into training data and test data. The training set contains a known output and the model learns on this data in order to be generalized to other data later on. We have the test dataset (or subset) in order to test our model’s prediction on this subset.</p>\n",
    "\n",
    "#### Finally the Model Function is called."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def init():\n",
    "    \n",
    "    df=pd.read_csv(\"C://Users//Rajwrita//Desktop//ds//cancer.csv\")\n",
    "    \n",
    "    \n",
    "    #Dropping the 'id' column from the dataset\n",
    "    df=df.drop(\"id\",1)\n",
    "    df=df.drop(\"Unnamed: 32\",1)\n",
    "    \n",
    "    #Mapping M to 1 and B to 0 in the output Label DataFrame\n",
    "    df['diagnosis']=df['diagnosis'].map({'M':1,'B':0})\n",
    "    \n",
    "    \n",
    "    #Split Data into training and test (70% and 30%)\n",
    "    train, test = train_test_split(df, test_size = 0.3, random_state=1)\n",
    "        \n",
    "    \n",
    "    #Training Data\n",
    "    train_x=train.loc[:,'radius_mean' : 'fractal_dimension_worst']\n",
    "    train_y=train.loc[:,['diagnosis']]\n",
    "    \n",
    "    #Testing Data\n",
    "    test_x=test.loc[:,'radius_mean' : 'fractal_dimension_worst']\n",
    "    test_y=test.loc[:,['diagnosis']]\n",
    "    \n",
    "    #Converting Traing and Test Data to numpy array\n",
    "    train_x=np.asarray(train_x)\n",
    "    train_y=np.asarray(train_y)\n",
    "    test_x=np.asarray(test_x)\n",
    "    test_y=np.asarray(test_y)\n",
    "    \n",
    "    #Calling the model function to train a Logistic Regression Model on Training Data\n",
    "    d= model(train_x.T, train_y.T, num_of_iterations=10000, alpha=0.000001)\n",
    "    \n",
    "        \n",
    "    costs=d[\"costs\"]\n",
    "    w=d[\"w\"]\n",
    "    b=d[\"b\"]\n",
    "    \n",
    "    #Drawing the plot between cost and number of iterations\n",
    "    plt.plot(costs)\n",
    "    plt.title(\"Cost vs #Iterations\")\n",
    "    plt.xlabel(\"Number of Iterations ( * 10)\")\n",
    "    plt.ylabel(\"Cost\")\n",
    "    \n",
    "    \n",
    "    #Now, calculating the accuracy on Training and Test Data\n",
    "    Y_prediction_train = predict(train_x.T, w, b)\n",
    "    Y_prediction_test = predict(test_x.T, w, b)\n",
    "    \n",
    "    \n",
    "    \n",
    "    print(\"\\nTrain accuracy: {} %\".format(100 - np.mean(np.abs(Y_prediction_train - train_y.T)) * 100))\n",
    "    \n",
    "    print(\"\\nTest accuracy: {} %\".format(100 - np.mean(np.abs(Y_prediction_test - test_y.T)) * 100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function to initialize the weights and bias\n",
    "def initialize(m):\n",
    "    \n",
    "    w = np.zeros((m,1))\n",
    "    b = 0\n",
    "    \n",
    "    return w , b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function to calculate sigmoid of x    \n",
    "def sigmoid(X):\n",
    "    return 1/(1 + np.exp(- X)) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function for doing forward and back propogation\n",
    "def propogate(X, Y, w, b):\n",
    "    \n",
    "    m = X.shape[1] #Number of training examples\n",
    "\n",
    "    #Forward Propogation, calculating the cost\n",
    "    Z = np.dot(w.T, X) + b;    \n",
    "    A = sigmoid(Z)\n",
    "    cost= -(1/m) * np.sum(Y * np.log(A) + (1-Y) * np.log(1-A))\n",
    "    \n",
    "    #Back Propogation , calculating the gradients\n",
    "    dw = (1/m)* np.dot(X, (A-Y).T)\n",
    "    db = (1/m)* np.sum(A-Y)\n",
    "    \n",
    "    grads= {\"dw\" : dw, \"db\" : db}\n",
    "    \n",
    "    return grads, cost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function for performing Grdient Descent\n",
    "def optimize(X, Y, w, b, num_of_iterations, alpha):\n",
    "    \n",
    "    costs=[] \n",
    "    \n",
    "    for i in range(num_of_iterations):\n",
    " \n",
    "        grads, cost = propogate(X, Y, w, b)\n",
    "        \n",
    "        dw = grads[\"dw\"]\n",
    "        db = grads[\"db\"]\n",
    "        \n",
    "        w = w - alpha * dw\n",
    "        b = b - alpha * db\n",
    "        \n",
    "        #Storing tthe cost at interval of every 10 iterations\n",
    "        if i% 10 == 0:\n",
    "            costs.append(cost)\n",
    "            print(\"cost after %i iteration : %f\" % (i, cost))\n",
    "            \n",
    "            \n",
    "    parameters = {\"w\":w, \"b\":b}\n",
    "    grads = {\"dw\":dw, \"db\":db}\n",
    "    \n",
    "    \n",
    "    return parameters, grads, costs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function for doing the predictions on the data set (mapping probabilities to 0 or 1)\n",
    "def predict(X, w, b):\n",
    "    \n",
    "    m = X.shape[1] #Number of training examples\n",
    "    \n",
    "    y_prediction =  np.zeros((1,m))\n",
    "    \n",
    "    w = w.reshape(X.shape[0], 1)\n",
    "    \n",
    "    A=sigmoid(np.dot(w.T, X)+b)\n",
    "    \n",
    "    \n",
    "    for i in range(A.shape[1]):\n",
    "        \n",
    "        if(A[0,i]<0.5):\n",
    "            y_prediction[0,i]=0\n",
    "        else:\n",
    "            y_prediction[0,i]=1\n",
    "            \n",
    "    \n",
    "    return y_prediction\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function for calculating the Logistic Regression Model\n",
    "def model(Xtrain, Ytrain, num_of_iterations, alpha):\n",
    "    \n",
    "    dim = Xtrain.shape[0] #Number of features\n",
    "    \n",
    "    w,b = initialize(dim)\n",
    "    \n",
    "    parameters, grads, costs = optimize(Xtrain, Ytrain, w, b, num_of_iterations, alpha) \n",
    "    \n",
    "    w = parameters[\"w\"]\n",
    "    b = parameters[\"b\"]\n",
    "        \n",
    "    \n",
    "    d={\"w\":w, \"b\":b, \"costs\": costs}\n",
    "    \n",
    "    return d"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cost after 0 iteration : 0.693147\n",
      "cost after 10 iteration : 0.671487\n",
      "cost after 20 iteration : 0.664807\n",
      "cost after 30 iteration : 0.658434\n",
      "cost after 40 iteration : 0.652346\n",
      "cost after 50 iteration : 0.646521\n",
      "cost after 60 iteration : 0.640940\n",
      "cost after 70 iteration : 0.635585\n",
      "cost after 80 iteration : 0.630439\n",
      "cost after 90 iteration : 0.625489\n",
      "cost after 100 iteration : 0.620721\n",
      "cost after 110 iteration : 0.616123\n",
      "cost after 120 iteration : 0.611684\n",
      "cost after 130 iteration : 0.607394\n",
      "cost after 140 iteration : 0.603243\n",
      "cost after 150 iteration : 0.599223\n",
      "cost after 160 iteration : 0.595328\n",
      "cost after 170 iteration : 0.591549\n",
      "cost after 180 iteration : 0.587880\n",
      "cost after 190 iteration : 0.584316\n",
      "cost after 200 iteration : 0.580850\n",
      "cost after 210 iteration : 0.577478\n",
      "cost after 220 iteration : 0.574196\n",
      "cost after 230 iteration : 0.570998\n",
      "cost after 240 iteration : 0.567881\n",
      "cost after 250 iteration : 0.564841\n",
      "cost after 260 iteration : 0.561875\n",
      "cost after 270 iteration : 0.558979\n",
      "cost after 280 iteration : 0.556150\n",
      "cost after 290 iteration : 0.553386\n",
      "cost after 300 iteration : 0.550683\n",
      "cost after 310 iteration : 0.548040\n",
      "cost after 320 iteration : 0.545453\n",
      "cost after 330 iteration : 0.542921\n",
      "cost after 340 iteration : 0.540442\n",
      "cost after 350 iteration : 0.538013\n",
      "cost after 360 iteration : 0.535633\n",
      "cost after 370 iteration : 0.533300\n",
      "cost after 380 iteration : 0.531012\n",
      "cost after 390 iteration : 0.528768\n",
      "cost after 400 iteration : 0.526566\n",
      "cost after 410 iteration : 0.524405\n",
      "cost after 420 iteration : 0.522284\n",
      "cost after 430 iteration : 0.520200\n",
      "cost after 440 iteration : 0.518154\n",
      "cost after 450 iteration : 0.516143\n",
      "cost after 460 iteration : 0.514167\n",
      "cost after 470 iteration : 0.512224\n",
      "cost after 480 iteration : 0.510314\n",
      "cost after 490 iteration : 0.508435\n",
      "cost after 500 iteration : 0.506587\n",
      "cost after 510 iteration : 0.504769\n",
      "cost after 520 iteration : 0.502980\n",
      "cost after 530 iteration : 0.501219\n",
      "cost after 540 iteration : 0.499485\n",
      "cost after 550 iteration : 0.497777\n",
      "cost after 560 iteration : 0.496096\n",
      "cost after 570 iteration : 0.494440\n",
      "cost after 580 iteration : 0.492808\n",
      "cost after 590 iteration : 0.491199\n",
      "cost after 600 iteration : 0.489614\n",
      "cost after 610 iteration : 0.488052\n",
      "cost after 620 iteration : 0.486512\n",
      "cost after 630 iteration : 0.484993\n",
      "cost after 640 iteration : 0.483495\n",
      "cost after 650 iteration : 0.482018\n",
      "cost after 660 iteration : 0.480560\n",
      "cost after 670 iteration : 0.479122\n",
      "cost after 680 iteration : 0.477703\n",
      "cost after 690 iteration : 0.476302\n",
      "cost after 700 iteration : 0.474920\n",
      "cost after 710 iteration : 0.473555\n",
      "cost after 720 iteration : 0.472208\n",
      "cost after 730 iteration : 0.470878\n",
      "cost after 740 iteration : 0.469564\n",
      "cost after 750 iteration : 0.468266\n",
      "cost after 760 iteration : 0.466984\n",
      "cost after 770 iteration : 0.465718\n",
      "cost after 780 iteration : 0.464467\n",
      "cost after 790 iteration : 0.463230\n",
      "cost after 800 iteration : 0.462008\n",
      "cost after 810 iteration : 0.460801\n",
      "cost after 820 iteration : 0.459607\n",
      "cost after 830 iteration : 0.458427\n",
      "cost after 840 iteration : 0.457260\n",
      "cost after 850 iteration : 0.456106\n",
      "cost after 860 iteration : 0.454965\n",
      "cost after 870 iteration : 0.453837\n",
      "cost after 880 iteration : 0.452721\n",
      "cost after 890 iteration : 0.451617\n",
      "cost after 900 iteration : 0.450525\n",
      "cost after 910 iteration : 0.449445\n",
      "cost after 920 iteration : 0.448376\n",
      "cost after 930 iteration : 0.447318\n",
      "cost after 940 iteration : 0.446272\n",
      "cost after 950 iteration : 0.445236\n",
      "cost after 960 iteration : 0.444210\n",
      "cost after 970 iteration : 0.443195\n",
      "cost after 980 iteration : 0.442191\n",
      "cost after 990 iteration : 0.441196\n",
      "cost after 1000 iteration : 0.440211\n",
      "cost after 1010 iteration : 0.439236\n",
      "cost after 1020 iteration : 0.438270\n",
      "cost after 1030 iteration : 0.437314\n",
      "cost after 1040 iteration : 0.436366\n",
      "cost after 1050 iteration : 0.435428\n",
      "cost after 1060 iteration : 0.434499\n",
      "cost after 1070 iteration : 0.433578\n",
      "cost after 1080 iteration : 0.432666\n",
      "cost after 1090 iteration : 0.431763\n",
      "cost after 1100 iteration : 0.430868\n",
      "cost after 1110 iteration : 0.429981\n",
      "cost after 1120 iteration : 0.429101\n",
      "cost after 1130 iteration : 0.428230\n",
      "cost after 1140 iteration : 0.427367\n",
      "cost after 1150 iteration : 0.426511\n",
      "cost after 1160 iteration : 0.425663\n",
      "cost after 1170 iteration : 0.424822\n",
      "cost after 1180 iteration : 0.423989\n",
      "cost after 1190 iteration : 0.423162\n",
      "cost after 1200 iteration : 0.422343\n",
      "cost after 1210 iteration : 0.421531\n",
      "cost after 1220 iteration : 0.420725\n",
      "cost after 1230 iteration : 0.419927\n",
      "cost after 1240 iteration : 0.419135\n",
      "cost after 1250 iteration : 0.418349\n",
      "cost after 1260 iteration : 0.417570\n",
      "cost after 1270 iteration : 0.416798\n",
      "cost after 1280 iteration : 0.416031\n",
      "cost after 1290 iteration : 0.415271\n",
      "cost after 1300 iteration : 0.414517\n",
      "cost after 1310 iteration : 0.413769\n",
      "cost after 1320 iteration : 0.413027\n",
      "cost after 1330 iteration : 0.412290\n",
      "cost after 1340 iteration : 0.411560\n",
      "cost after 1350 iteration : 0.410835\n",
      "cost after 1360 iteration : 0.410116\n",
      "cost after 1370 iteration : 0.409402\n",
      "cost after 1380 iteration : 0.408694\n",
      "cost after 1390 iteration : 0.407991\n",
      "cost after 1400 iteration : 0.407293\n",
      "cost after 1410 iteration : 0.406601\n",
      "cost after 1420 iteration : 0.405914\n",
      "cost after 1430 iteration : 0.405232\n",
      "cost after 1440 iteration : 0.404554\n",
      "cost after 1450 iteration : 0.403882\n",
      "cost after 1460 iteration : 0.403215\n",
      "cost after 1470 iteration : 0.402553\n",
      "cost after 1480 iteration : 0.401895\n",
      "cost after 1490 iteration : 0.401242\n",
      "cost after 1500 iteration : 0.400594\n",
      "cost after 1510 iteration : 0.399950\n",
      "cost after 1520 iteration : 0.399311\n",
      "cost after 1530 iteration : 0.398676\n",
      "cost after 1540 iteration : 0.398046\n",
      "cost after 1550 iteration : 0.397420\n",
      "cost after 1560 iteration : 0.396799\n",
      "cost after 1570 iteration : 0.396181\n",
      "cost after 1580 iteration : 0.395568\n",
      "cost after 1590 iteration : 0.394959\n",
      "cost after 1600 iteration : 0.394355\n",
      "cost after 1610 iteration : 0.393754\n",
      "cost after 1620 iteration : 0.393157\n",
      "cost after 1630 iteration : 0.392565\n",
      "cost after 1640 iteration : 0.391976\n",
      "cost after 1650 iteration : 0.391391\n",
      "cost after 1660 iteration : 0.390810\n",
      "cost after 1670 iteration : 0.390232\n",
      "cost after 1680 iteration : 0.389659\n",
      "cost after 1690 iteration : 0.389089\n",
      "cost after 1700 iteration : 0.388523\n",
      "cost after 1710 iteration : 0.387960\n",
      "cost after 1720 iteration : 0.387401\n",
      "cost after 1730 iteration : 0.386846\n",
      "cost after 1740 iteration : 0.386294\n",
      "cost after 1750 iteration : 0.385745\n",
      "cost after 1760 iteration : 0.385200\n",
      "cost after 1770 iteration : 0.384658\n",
      "cost after 1780 iteration : 0.384120\n",
      "cost after 1790 iteration : 0.383585\n",
      "cost after 1800 iteration : 0.383053\n",
      "cost after 1810 iteration : 0.382525\n",
      "cost after 1820 iteration : 0.381999\n",
      "cost after 1830 iteration : 0.381477\n",
      "cost after 1840 iteration : 0.380958\n",
      "cost after 1850 iteration : 0.380442\n",
      "cost after 1860 iteration : 0.379930\n",
      "cost after 1870 iteration : 0.379420\n",
      "cost after 1880 iteration : 0.378913\n",
      "cost after 1890 iteration : 0.378409\n",
      "cost after 1900 iteration : 0.377909\n",
      "cost after 1910 iteration : 0.377411\n",
      "cost after 1920 iteration : 0.376916\n",
      "cost after 1930 iteration : 0.376424\n",
      "cost after 1940 iteration : 0.375935\n",
      "cost after 1950 iteration : 0.375448\n",
      "cost after 1960 iteration : 0.374964\n",
      "cost after 1970 iteration : 0.374484\n",
      "cost after 1980 iteration : 0.374005\n",
      "cost after 1990 iteration : 0.373530\n",
      "cost after 2000 iteration : 0.373057\n",
      "cost after 2010 iteration : 0.372587\n",
      "cost after 2020 iteration : 0.372119\n",
      "cost after 2030 iteration : 0.371654\n",
      "cost after 2040 iteration : 0.371192\n",
      "cost after 2050 iteration : 0.370732\n",
      "cost after 2060 iteration : 0.370275\n",
      "cost after 2070 iteration : 0.369820\n",
      "cost after 2080 iteration : 0.369368\n",
      "cost after 2090 iteration : 0.368918\n",
      "cost after 2100 iteration : 0.368471\n",
      "cost after 2110 iteration : 0.368026\n",
      "cost after 2120 iteration : 0.367583\n",
      "cost after 2130 iteration : 0.367143\n",
      "cost after 2140 iteration : 0.366705\n",
      "cost after 2150 iteration : 0.366270\n",
      "cost after 2160 iteration : 0.365836\n",
      "cost after 2170 iteration : 0.365405\n",
      "cost after 2180 iteration : 0.364977\n",
      "cost after 2190 iteration : 0.364550\n",
      "cost after 2200 iteration : 0.364126\n",
      "cost after 2210 iteration : 0.363704\n",
      "cost after 2220 iteration : 0.363284\n",
      "cost after 2230 iteration : 0.362867\n",
      "cost after 2240 iteration : 0.362451\n",
      "cost after 2250 iteration : 0.362038\n",
      "cost after 2260 iteration : 0.361627\n",
      "cost after 2270 iteration : 0.361218\n",
      "cost after 2280 iteration : 0.360811\n",
      "cost after 2290 iteration : 0.360406\n",
      "cost after 2300 iteration : 0.360003\n",
      "cost after 2310 iteration : 0.359602\n",
      "cost after 2320 iteration : 0.359203\n",
      "cost after 2330 iteration : 0.358806\n",
      "cost after 2340 iteration : 0.358411\n",
      "cost after 2350 iteration : 0.358018\n",
      "cost after 2360 iteration : 0.357627\n",
      "cost after 2370 iteration : 0.357238\n",
      "cost after 2380 iteration : 0.356851\n",
      "cost after 2390 iteration : 0.356465\n",
      "cost after 2400 iteration : 0.356082\n",
      "cost after 2410 iteration : 0.355700\n",
      "cost after 2420 iteration : 0.355321\n",
      "cost after 2430 iteration : 0.354943\n",
      "cost after 2440 iteration : 0.354567\n",
      "cost after 2450 iteration : 0.354193\n",
      "cost after 2460 iteration : 0.353820\n",
      "cost after 2470 iteration : 0.353449\n",
      "cost after 2480 iteration : 0.353081\n",
      "cost after 2490 iteration : 0.352714\n",
      "cost after 2500 iteration : 0.352348\n",
      "cost after 2510 iteration : 0.351985\n",
      "cost after 2520 iteration : 0.351623\n",
      "cost after 2530 iteration : 0.351262\n",
      "cost after 2540 iteration : 0.350904\n",
      "cost after 2550 iteration : 0.350547\n",
      "cost after 2560 iteration : 0.350192\n",
      "cost after 2570 iteration : 0.349838\n",
      "cost after 2580 iteration : 0.349486\n",
      "cost after 2590 iteration : 0.349136\n",
      "cost after 2600 iteration : 0.348788\n",
      "cost after 2610 iteration : 0.348440\n",
      "cost after 2620 iteration : 0.348095\n",
      "cost after 2630 iteration : 0.347751\n",
      "cost after 2640 iteration : 0.347409\n",
      "cost after 2650 iteration : 0.347068\n",
      "cost after 2660 iteration : 0.346729\n",
      "cost after 2670 iteration : 0.346391\n",
      "cost after 2680 iteration : 0.346055\n",
      "cost after 2690 iteration : 0.345721\n",
      "cost after 2700 iteration : 0.345387\n",
      "cost after 2710 iteration : 0.345056\n",
      "cost after 2720 iteration : 0.344726\n",
      "cost after 2730 iteration : 0.344397\n",
      "cost after 2740 iteration : 0.344070\n",
      "cost after 2750 iteration : 0.343744\n",
      "cost after 2760 iteration : 0.343420\n",
      "cost after 2770 iteration : 0.343097\n",
      "cost after 2780 iteration : 0.342775\n",
      "cost after 2790 iteration : 0.342455\n",
      "cost after 2800 iteration : 0.342137\n",
      "cost after 2810 iteration : 0.341819\n",
      "cost after 2820 iteration : 0.341504\n",
      "cost after 2830 iteration : 0.341189\n",
      "cost after 2840 iteration : 0.340876\n",
      "cost after 2850 iteration : 0.340564\n",
      "cost after 2860 iteration : 0.340254\n",
      "cost after 2870 iteration : 0.339945\n",
      "cost after 2880 iteration : 0.339637\n",
      "cost after 2890 iteration : 0.339331\n",
      "cost after 2900 iteration : 0.339025\n",
      "cost after 2910 iteration : 0.338722\n",
      "cost after 2920 iteration : 0.338419\n",
      "cost after 2930 iteration : 0.338118\n",
      "cost after 2940 iteration : 0.337818\n",
      "cost after 2950 iteration : 0.337519\n",
      "cost after 2960 iteration : 0.337222\n",
      "cost after 2970 iteration : 0.336926\n",
      "cost after 2980 iteration : 0.336631\n",
      "cost after 2990 iteration : 0.336337\n",
      "cost after 3000 iteration : 0.336045\n",
      "cost after 3010 iteration : 0.335753\n",
      "cost after 3020 iteration : 0.335463\n",
      "cost after 3030 iteration : 0.335175\n",
      "cost after 3040 iteration : 0.334887\n",
      "cost after 3050 iteration : 0.334601\n",
      "cost after 3060 iteration : 0.334315\n",
      "cost after 3070 iteration : 0.334031\n",
      "cost after 3080 iteration : 0.333748\n",
      "cost after 3090 iteration : 0.333467\n",
      "cost after 3100 iteration : 0.333186\n",
      "cost after 3110 iteration : 0.332907\n",
      "cost after 3120 iteration : 0.332628\n",
      "cost after 3130 iteration : 0.332351\n",
      "cost after 3140 iteration : 0.332075\n",
      "cost after 3150 iteration : 0.331800\n",
      "cost after 3160 iteration : 0.331526\n",
      "cost after 3170 iteration : 0.331254\n",
      "cost after 3180 iteration : 0.330982\n",
      "cost after 3190 iteration : 0.330712\n",
      "cost after 3200 iteration : 0.330442\n",
      "cost after 3210 iteration : 0.330174\n",
      "cost after 3220 iteration : 0.329907\n",
      "cost after 3230 iteration : 0.329641\n",
      "cost after 3240 iteration : 0.329375\n",
      "cost after 3250 iteration : 0.329111\n",
      "cost after 3260 iteration : 0.328848\n",
      "cost after 3270 iteration : 0.328586\n",
      "cost after 3280 iteration : 0.328325\n",
      "cost after 3290 iteration : 0.328066\n",
      "cost after 3300 iteration : 0.327807\n",
      "cost after 3310 iteration : 0.327549\n",
      "cost after 3320 iteration : 0.327292\n",
      "cost after 3330 iteration : 0.327036\n",
      "cost after 3340 iteration : 0.326781\n",
      "cost after 3350 iteration : 0.326527\n",
      "cost after 3360 iteration : 0.326275\n",
      "cost after 3370 iteration : 0.326023\n",
      "cost after 3380 iteration : 0.325772\n",
      "cost after 3390 iteration : 0.325522\n",
      "cost after 3400 iteration : 0.325273\n",
      "cost after 3410 iteration : 0.325025\n",
      "cost after 3420 iteration : 0.324778\n",
      "cost after 3430 iteration : 0.324532\n",
      "cost after 3440 iteration : 0.324287\n",
      "cost after 3450 iteration : 0.324042\n",
      "cost after 3460 iteration : 0.323799\n",
      "cost after 3470 iteration : 0.323557\n",
      "cost after 3480 iteration : 0.323315\n",
      "cost after 3490 iteration : 0.323075\n",
      "cost after 3500 iteration : 0.322835\n",
      "cost after 3510 iteration : 0.322597\n",
      "cost after 3520 iteration : 0.322359\n",
      "cost after 3530 iteration : 0.322122\n",
      "cost after 3540 iteration : 0.321886\n",
      "cost after 3550 iteration : 0.321651\n",
      "cost after 3560 iteration : 0.321417\n",
      "cost after 3570 iteration : 0.321183\n",
      "cost after 3580 iteration : 0.320951\n",
      "cost after 3590 iteration : 0.320719\n",
      "cost after 3600 iteration : 0.320489\n",
      "cost after 3610 iteration : 0.320259\n",
      "cost after 3620 iteration : 0.320030\n",
      "cost after 3630 iteration : 0.319802\n",
      "cost after 3640 iteration : 0.319574\n",
      "cost after 3650 iteration : 0.319348\n",
      "cost after 3660 iteration : 0.319122\n",
      "cost after 3670 iteration : 0.318897\n",
      "cost after 3680 iteration : 0.318673\n",
      "cost after 3690 iteration : 0.318450\n",
      "cost after 3700 iteration : 0.318228\n",
      "cost after 3710 iteration : 0.318006\n",
      "cost after 3720 iteration : 0.317786\n",
      "cost after 3730 iteration : 0.317566\n",
      "cost after 3740 iteration : 0.317347\n",
      "cost after 3750 iteration : 0.317129\n",
      "cost after 3760 iteration : 0.316911\n",
      "cost after 3770 iteration : 0.316694\n",
      "cost after 3780 iteration : 0.316478\n",
      "cost after 3790 iteration : 0.316263\n",
      "cost after 3800 iteration : 0.316049\n",
      "cost after 3810 iteration : 0.315835\n",
      "cost after 3820 iteration : 0.315622\n",
      "cost after 3830 iteration : 0.315410\n",
      "cost after 3840 iteration : 0.315199\n",
      "cost after 3850 iteration : 0.314989\n",
      "cost after 3860 iteration : 0.314779\n",
      "cost after 3870 iteration : 0.314570\n",
      "cost after 3880 iteration : 0.314361\n",
      "cost after 3890 iteration : 0.314154\n",
      "cost after 3900 iteration : 0.313947\n",
      "cost after 3910 iteration : 0.313741\n",
      "cost after 3920 iteration : 0.313536\n",
      "cost after 3930 iteration : 0.313331\n",
      "cost after 3940 iteration : 0.313127\n",
      "cost after 3950 iteration : 0.312924\n",
      "cost after 3960 iteration : 0.312721\n",
      "cost after 3970 iteration : 0.312520\n",
      "cost after 3980 iteration : 0.312319\n",
      "cost after 3990 iteration : 0.312118\n",
      "cost after 4000 iteration : 0.311919\n",
      "cost after 4010 iteration : 0.311720\n",
      "cost after 4020 iteration : 0.311521\n",
      "cost after 4030 iteration : 0.311324\n",
      "cost after 4040 iteration : 0.311127\n",
      "cost after 4050 iteration : 0.310931\n",
      "cost after 4060 iteration : 0.310735\n",
      "cost after 4070 iteration : 0.310540\n",
      "cost after 4080 iteration : 0.310346\n",
      "cost after 4090 iteration : 0.310153\n",
      "cost after 4100 iteration : 0.309960\n",
      "cost after 4110 iteration : 0.309768\n",
      "cost after 4120 iteration : 0.309576\n",
      "cost after 4130 iteration : 0.309385\n",
      "cost after 4140 iteration : 0.309195\n",
      "cost after 4150 iteration : 0.309005\n",
      "cost after 4160 iteration : 0.308816\n",
      "cost after 4170 iteration : 0.308628\n",
      "cost after 4180 iteration : 0.308441\n",
      "cost after 4190 iteration : 0.308254\n",
      "cost after 4200 iteration : 0.308067\n",
      "cost after 4210 iteration : 0.307881\n",
      "cost after 4220 iteration : 0.307696\n",
      "cost after 4230 iteration : 0.307512\n",
      "cost after 4240 iteration : 0.307328\n",
      "cost after 4250 iteration : 0.307145\n",
      "cost after 4260 iteration : 0.306962\n",
      "cost after 4270 iteration : 0.306780\n",
      "cost after 4280 iteration : 0.306599\n",
      "cost after 4290 iteration : 0.306418\n",
      "cost after 4300 iteration : 0.306238\n",
      "cost after 4310 iteration : 0.306058\n",
      "cost after 4320 iteration : 0.305879\n",
      "cost after 4330 iteration : 0.305701\n",
      "cost after 4340 iteration : 0.305523\n",
      "cost after 4350 iteration : 0.305345\n",
      "cost after 4360 iteration : 0.305169\n",
      "cost after 4370 iteration : 0.304993\n",
      "cost after 4380 iteration : 0.304817\n",
      "cost after 4390 iteration : 0.304642\n",
      "cost after 4400 iteration : 0.304468\n",
      "cost after 4410 iteration : 0.304294\n",
      "cost after 4420 iteration : 0.304121\n",
      "cost after 4430 iteration : 0.303948\n",
      "cost after 4440 iteration : 0.303776\n",
      "cost after 4450 iteration : 0.303605\n",
      "cost after 4460 iteration : 0.303434\n",
      "cost after 4470 iteration : 0.303263\n",
      "cost after 4480 iteration : 0.303094\n",
      "cost after 4490 iteration : 0.302924\n",
      "cost after 4500 iteration : 0.302756\n",
      "cost after 4510 iteration : 0.302587\n",
      "cost after 4520 iteration : 0.302420\n",
      "cost after 4530 iteration : 0.302253\n",
      "cost after 4540 iteration : 0.302086\n",
      "cost after 4550 iteration : 0.301920\n",
      "cost after 4560 iteration : 0.301754\n",
      "cost after 4570 iteration : 0.301589\n",
      "cost after 4580 iteration : 0.301425\n",
      "cost after 4590 iteration : 0.301261\n",
      "cost after 4600 iteration : 0.301098\n",
      "cost after 4610 iteration : 0.300935\n",
      "cost after 4620 iteration : 0.300772\n",
      "cost after 4630 iteration : 0.300610\n",
      "cost after 4640 iteration : 0.300449\n",
      "cost after 4650 iteration : 0.300288\n",
      "cost after 4660 iteration : 0.300128\n",
      "cost after 4670 iteration : 0.299968\n",
      "cost after 4680 iteration : 0.299809\n",
      "cost after 4690 iteration : 0.299650\n",
      "cost after 4700 iteration : 0.299491\n",
      "cost after 4710 iteration : 0.299334\n",
      "cost after 4720 iteration : 0.299176\n",
      "cost after 4730 iteration : 0.299019\n",
      "cost after 4740 iteration : 0.298863\n",
      "cost after 4750 iteration : 0.298707\n",
      "cost after 4760 iteration : 0.298552\n",
      "cost after 4770 iteration : 0.298397\n",
      "cost after 4780 iteration : 0.298242\n",
      "cost after 4790 iteration : 0.298088\n",
      "cost after 4800 iteration : 0.297935\n",
      "cost after 4810 iteration : 0.297782\n",
      "cost after 4820 iteration : 0.297629\n",
      "cost after 4830 iteration : 0.297477\n",
      "cost after 4840 iteration : 0.297325\n",
      "cost after 4850 iteration : 0.297174\n",
      "cost after 4860 iteration : 0.297024\n",
      "cost after 4870 iteration : 0.296873\n",
      "cost after 4880 iteration : 0.296724\n",
      "cost after 4890 iteration : 0.296574\n",
      "cost after 4900 iteration : 0.296425\n",
      "cost after 4910 iteration : 0.296277\n",
      "cost after 4920 iteration : 0.296129\n",
      "cost after 4930 iteration : 0.295982\n",
      "cost after 4940 iteration : 0.295834\n",
      "cost after 4950 iteration : 0.295688\n",
      "cost after 4960 iteration : 0.295542\n",
      "cost after 4970 iteration : 0.295396\n",
      "cost after 4980 iteration : 0.295251\n",
      "cost after 4990 iteration : 0.295106\n",
      "cost after 5000 iteration : 0.294961\n",
      "cost after 5010 iteration : 0.294817\n",
      "cost after 5020 iteration : 0.294674\n",
      "cost after 5030 iteration : 0.294531\n",
      "cost after 5040 iteration : 0.294388\n",
      "cost after 5050 iteration : 0.294245\n",
      "cost after 5060 iteration : 0.294104\n",
      "cost after 5070 iteration : 0.293962\n",
      "cost after 5080 iteration : 0.293821\n",
      "cost after 5090 iteration : 0.293680\n",
      "cost after 5100 iteration : 0.293540\n",
      "cost after 5110 iteration : 0.293400\n",
      "cost after 5120 iteration : 0.293261\n",
      "cost after 5130 iteration : 0.293122\n",
      "cost after 5140 iteration : 0.292983\n",
      "cost after 5150 iteration : 0.292845\n",
      "cost after 5160 iteration : 0.292708\n",
      "cost after 5170 iteration : 0.292570\n",
      "cost after 5180 iteration : 0.292433\n",
      "cost after 5190 iteration : 0.292297\n",
      "cost after 5200 iteration : 0.292160\n",
      "cost after 5210 iteration : 0.292025\n",
      "cost after 5220 iteration : 0.291889\n",
      "cost after 5230 iteration : 0.291754\n",
      "cost after 5240 iteration : 0.291620\n",
      "cost after 5250 iteration : 0.291485\n",
      "cost after 5260 iteration : 0.291352\n",
      "cost after 5270 iteration : 0.291218\n",
      "cost after 5280 iteration : 0.291085\n",
      "cost after 5290 iteration : 0.290952\n",
      "cost after 5300 iteration : 0.290820\n",
      "cost after 5310 iteration : 0.290688\n",
      "cost after 5320 iteration : 0.290557\n",
      "cost after 5330 iteration : 0.290425\n",
      "cost after 5340 iteration : 0.290295\n",
      "cost after 5350 iteration : 0.290164\n",
      "cost after 5360 iteration : 0.290034\n",
      "cost after 5370 iteration : 0.289904\n",
      "cost after 5380 iteration : 0.289775\n",
      "cost after 5390 iteration : 0.289646\n",
      "cost after 5400 iteration : 0.289518\n",
      "cost after 5410 iteration : 0.289389\n",
      "cost after 5420 iteration : 0.289261\n",
      "cost after 5430 iteration : 0.289134\n",
      "cost after 5440 iteration : 0.289007\n",
      "cost after 5450 iteration : 0.288880\n",
      "cost after 5460 iteration : 0.288753\n",
      "cost after 5470 iteration : 0.288627\n",
      "cost after 5480 iteration : 0.288502\n",
      "cost after 5490 iteration : 0.288376\n",
      "cost after 5500 iteration : 0.288251\n",
      "cost after 5510 iteration : 0.288126\n",
      "cost after 5520 iteration : 0.288002\n",
      "cost after 5530 iteration : 0.287878\n",
      "cost after 5540 iteration : 0.287754\n",
      "cost after 5550 iteration : 0.287631\n",
      "cost after 5560 iteration : 0.287508\n",
      "cost after 5570 iteration : 0.287385\n",
      "cost after 5580 iteration : 0.287263\n",
      "cost after 5590 iteration : 0.287141\n",
      "cost after 5600 iteration : 0.287019\n",
      "cost after 5610 iteration : 0.286898\n",
      "cost after 5620 iteration : 0.286777\n",
      "cost after 5630 iteration : 0.286656\n",
      "cost after 5640 iteration : 0.286536\n",
      "cost after 5650 iteration : 0.286416\n",
      "cost after 5660 iteration : 0.286296\n",
      "cost after 5670 iteration : 0.286177\n",
      "cost after 5680 iteration : 0.286058\n",
      "cost after 5690 iteration : 0.285939\n",
      "cost after 5700 iteration : 0.285821\n",
      "cost after 5710 iteration : 0.285703\n",
      "cost after 5720 iteration : 0.285585\n",
      "cost after 5730 iteration : 0.285467\n",
      "cost after 5740 iteration : 0.285350\n",
      "cost after 5750 iteration : 0.285233\n",
      "cost after 5760 iteration : 0.285117\n",
      "cost after 5770 iteration : 0.285001\n",
      "cost after 5780 iteration : 0.284885\n",
      "cost after 5790 iteration : 0.284769\n",
      "cost after 5800 iteration : 0.284654\n",
      "cost after 5810 iteration : 0.284539\n",
      "cost after 5820 iteration : 0.284424\n",
      "cost after 5830 iteration : 0.284310\n",
      "cost after 5840 iteration : 0.284196\n",
      "cost after 5850 iteration : 0.284082\n",
      "cost after 5860 iteration : 0.283969\n",
      "cost after 5870 iteration : 0.283855\n",
      "cost after 5880 iteration : 0.283743\n",
      "cost after 5890 iteration : 0.283630\n",
      "cost after 5900 iteration : 0.283518\n",
      "cost after 5910 iteration : 0.283406\n",
      "cost after 5920 iteration : 0.283294\n",
      "cost after 5930 iteration : 0.283183\n",
      "cost after 5940 iteration : 0.283072\n",
      "cost after 5950 iteration : 0.282961\n",
      "cost after 5960 iteration : 0.282850\n",
      "cost after 5970 iteration : 0.282740\n",
      "cost after 5980 iteration : 0.282630\n",
      "cost after 5990 iteration : 0.282520\n",
      "cost after 6000 iteration : 0.282411\n",
      "cost after 6010 iteration : 0.282302\n",
      "cost after 6020 iteration : 0.282193\n",
      "cost after 6030 iteration : 0.282084\n",
      "cost after 6040 iteration : 0.281976\n",
      "cost after 6050 iteration : 0.281868\n",
      "cost after 6060 iteration : 0.281760\n",
      "cost after 6070 iteration : 0.281653\n",
      "cost after 6080 iteration : 0.281546\n",
      "cost after 6090 iteration : 0.281439\n",
      "cost after 6100 iteration : 0.281332\n",
      "cost after 6110 iteration : 0.281226\n",
      "cost after 6120 iteration : 0.281120\n",
      "cost after 6130 iteration : 0.281014\n",
      "cost after 6140 iteration : 0.280908\n",
      "cost after 6150 iteration : 0.280803\n",
      "cost after 6160 iteration : 0.280698\n",
      "cost after 6170 iteration : 0.280593\n",
      "cost after 6180 iteration : 0.280489\n",
      "cost after 6190 iteration : 0.280384\n",
      "cost after 6200 iteration : 0.280280\n",
      "cost after 6210 iteration : 0.280177\n",
      "cost after 6220 iteration : 0.280073\n",
      "cost after 6230 iteration : 0.279970\n",
      "cost after 6240 iteration : 0.279867\n",
      "cost after 6250 iteration : 0.279764\n",
      "cost after 6260 iteration : 0.279662\n",
      "cost after 6270 iteration : 0.279560\n",
      "cost after 6280 iteration : 0.279458\n",
      "cost after 6290 iteration : 0.279356\n",
      "cost after 6300 iteration : 0.279255\n",
      "cost after 6310 iteration : 0.279154\n",
      "cost after 6320 iteration : 0.279053\n",
      "cost after 6330 iteration : 0.278952\n",
      "cost after 6340 iteration : 0.278852\n",
      "cost after 6350 iteration : 0.278751\n",
      "cost after 6360 iteration : 0.278651\n",
      "cost after 6370 iteration : 0.278552\n",
      "cost after 6380 iteration : 0.278452\n",
      "cost after 6390 iteration : 0.278353\n",
      "cost after 6400 iteration : 0.278254\n",
      "cost after 6410 iteration : 0.278155\n",
      "cost after 6420 iteration : 0.278057\n",
      "cost after 6430 iteration : 0.277959\n",
      "cost after 6440 iteration : 0.277861\n",
      "cost after 6450 iteration : 0.277763\n",
      "cost after 6460 iteration : 0.277665\n",
      "cost after 6470 iteration : 0.277568\n",
      "cost after 6480 iteration : 0.277471\n",
      "cost after 6490 iteration : 0.277374\n",
      "cost after 6500 iteration : 0.277278\n",
      "cost after 6510 iteration : 0.277181\n",
      "cost after 6520 iteration : 0.277085\n",
      "cost after 6530 iteration : 0.276989\n",
      "cost after 6540 iteration : 0.276893\n",
      "cost after 6550 iteration : 0.276798\n",
      "cost after 6560 iteration : 0.276703\n",
      "cost after 6570 iteration : 0.276608\n",
      "cost after 6580 iteration : 0.276513\n",
      "cost after 6590 iteration : 0.276418\n",
      "cost after 6600 iteration : 0.276324\n",
      "cost after 6610 iteration : 0.276230\n",
      "cost after 6620 iteration : 0.276136\n",
      "cost after 6630 iteration : 0.276042\n",
      "cost after 6640 iteration : 0.275949\n",
      "cost after 6650 iteration : 0.275856\n",
      "cost after 6660 iteration : 0.275763\n",
      "cost after 6670 iteration : 0.275670\n",
      "cost after 6680 iteration : 0.275577\n",
      "cost after 6690 iteration : 0.275485\n",
      "cost after 6700 iteration : 0.275393\n",
      "cost after 6710 iteration : 0.275301\n",
      "cost after 6720 iteration : 0.275209\n",
      "cost after 6730 iteration : 0.275118\n",
      "cost after 6740 iteration : 0.275027\n",
      "cost after 6750 iteration : 0.274936\n",
      "cost after 6760 iteration : 0.274845\n",
      "cost after 6770 iteration : 0.274754\n",
      "cost after 6780 iteration : 0.274664\n",
      "cost after 6790 iteration : 0.274573\n",
      "cost after 6800 iteration : 0.274483\n",
      "cost after 6810 iteration : 0.274394\n",
      "cost after 6820 iteration : 0.274304\n",
      "cost after 6830 iteration : 0.274215\n",
      "cost after 6840 iteration : 0.274125\n",
      "cost after 6850 iteration : 0.274036\n",
      "cost after 6860 iteration : 0.273948\n",
      "cost after 6870 iteration : 0.273859\n",
      "cost after 6880 iteration : 0.273771\n",
      "cost after 6890 iteration : 0.273682\n",
      "cost after 6900 iteration : 0.273594\n",
      "cost after 6910 iteration : 0.273507\n",
      "cost after 6920 iteration : 0.273419\n",
      "cost after 6930 iteration : 0.273332\n",
      "cost after 6940 iteration : 0.273245\n",
      "cost after 6950 iteration : 0.273158\n",
      "cost after 6960 iteration : 0.273071\n",
      "cost after 6970 iteration : 0.272984\n",
      "cost after 6980 iteration : 0.272898\n",
      "cost after 6990 iteration : 0.272812\n",
      "cost after 7000 iteration : 0.272726\n",
      "cost after 7010 iteration : 0.272640\n",
      "cost after 7020 iteration : 0.272554\n",
      "cost after 7030 iteration : 0.272469\n",
      "cost after 7040 iteration : 0.272383\n",
      "cost after 7050 iteration : 0.272298\n",
      "cost after 7060 iteration : 0.272214\n",
      "cost after 7070 iteration : 0.272129\n",
      "cost after 7080 iteration : 0.272044\n",
      "cost after 7090 iteration : 0.271960\n",
      "cost after 7100 iteration : 0.271876\n",
      "cost after 7110 iteration : 0.271792\n",
      "cost after 7120 iteration : 0.271708\n",
      "cost after 7130 iteration : 0.271625\n",
      "cost after 7140 iteration : 0.271541\n",
      "cost after 7150 iteration : 0.271458\n",
      "cost after 7160 iteration : 0.271375\n",
      "cost after 7170 iteration : 0.271292\n",
      "cost after 7180 iteration : 0.271210\n",
      "cost after 7190 iteration : 0.271127\n",
      "cost after 7200 iteration : 0.271045\n",
      "cost after 7210 iteration : 0.270963\n",
      "cost after 7220 iteration : 0.270881\n",
      "cost after 7230 iteration : 0.270799\n",
      "cost after 7240 iteration : 0.270718\n",
      "cost after 7250 iteration : 0.270636\n",
      "cost after 7260 iteration : 0.270555\n",
      "cost after 7270 iteration : 0.270474\n",
      "cost after 7280 iteration : 0.270393\n",
      "cost after 7290 iteration : 0.270312\n",
      "cost after 7300 iteration : 0.270232\n",
      "cost after 7310 iteration : 0.270152\n",
      "cost after 7320 iteration : 0.270071\n",
      "cost after 7330 iteration : 0.269991\n",
      "cost after 7340 iteration : 0.269912\n",
      "cost after 7350 iteration : 0.269832\n",
      "cost after 7360 iteration : 0.269752\n",
      "cost after 7370 iteration : 0.269673\n",
      "cost after 7380 iteration : 0.269594\n",
      "cost after 7390 iteration : 0.269515\n",
      "cost after 7400 iteration : 0.269436\n",
      "cost after 7410 iteration : 0.269358\n",
      "cost after 7420 iteration : 0.269279\n",
      "cost after 7430 iteration : 0.269201\n",
      "cost after 7440 iteration : 0.269123\n",
      "cost after 7450 iteration : 0.269045\n",
      "cost after 7460 iteration : 0.268967\n",
      "cost after 7470 iteration : 0.268889\n",
      "cost after 7480 iteration : 0.268812\n",
      "cost after 7490 iteration : 0.268734\n",
      "cost after 7500 iteration : 0.268657\n",
      "cost after 7510 iteration : 0.268580\n",
      "cost after 7520 iteration : 0.268503\n",
      "cost after 7530 iteration : 0.268427\n",
      "cost after 7540 iteration : 0.268350\n",
      "cost after 7550 iteration : 0.268274\n",
      "cost after 7560 iteration : 0.268198\n",
      "cost after 7570 iteration : 0.268122\n",
      "cost after 7580 iteration : 0.268046\n",
      "cost after 7590 iteration : 0.267970\n",
      "cost after 7600 iteration : 0.267895\n",
      "cost after 7610 iteration : 0.267819\n",
      "cost after 7620 iteration : 0.267744\n",
      "cost after 7630 iteration : 0.267669\n",
      "cost after 7640 iteration : 0.267594\n",
      "cost after 7650 iteration : 0.267519\n",
      "cost after 7660 iteration : 0.267444\n",
      "cost after 7670 iteration : 0.267370\n",
      "cost after 7680 iteration : 0.267296\n",
      "cost after 7690 iteration : 0.267222\n",
      "cost after 7700 iteration : 0.267148\n",
      "cost after 7710 iteration : 0.267074\n",
      "cost after 7720 iteration : 0.267000\n",
      "cost after 7730 iteration : 0.266926\n",
      "cost after 7740 iteration : 0.266853\n",
      "cost after 7750 iteration : 0.266780\n",
      "cost after 7760 iteration : 0.266707\n",
      "cost after 7770 iteration : 0.266634\n",
      "cost after 7780 iteration : 0.266561\n",
      "cost after 7790 iteration : 0.266488\n",
      "cost after 7800 iteration : 0.266416\n",
      "cost after 7810 iteration : 0.266343\n",
      "cost after 7820 iteration : 0.266271\n",
      "cost after 7830 iteration : 0.266199\n",
      "cost after 7840 iteration : 0.266127\n",
      "cost after 7850 iteration : 0.266055\n",
      "cost after 7860 iteration : 0.265984\n",
      "cost after 7870 iteration : 0.265912\n",
      "cost after 7880 iteration : 0.265841\n",
      "cost after 7890 iteration : 0.265770\n",
      "cost after 7900 iteration : 0.265699\n",
      "cost after 7910 iteration : 0.265628\n",
      "cost after 7920 iteration : 0.265557\n",
      "cost after 7930 iteration : 0.265486\n",
      "cost after 7940 iteration : 0.265416\n",
      "cost after 7950 iteration : 0.265345\n",
      "cost after 7960 iteration : 0.265275\n",
      "cost after 7970 iteration : 0.265205\n",
      "cost after 7980 iteration : 0.265135\n",
      "cost after 7990 iteration : 0.265065\n",
      "cost after 8000 iteration : 0.264996\n",
      "cost after 8010 iteration : 0.264926\n",
      "cost after 8020 iteration : 0.264857\n",
      "cost after 8030 iteration : 0.264787\n",
      "cost after 8040 iteration : 0.264718\n",
      "cost after 8050 iteration : 0.264649\n",
      "cost after 8060 iteration : 0.264580\n",
      "cost after 8070 iteration : 0.264512\n",
      "cost after 8080 iteration : 0.264443\n",
      "cost after 8090 iteration : 0.264375\n",
      "cost after 8100 iteration : 0.264306\n",
      "cost after 8110 iteration : 0.264238\n",
      "cost after 8120 iteration : 0.264170\n",
      "cost after 8130 iteration : 0.264102\n",
      "cost after 8140 iteration : 0.264035\n",
      "cost after 8150 iteration : 0.263967\n",
      "cost after 8160 iteration : 0.263899\n",
      "cost after 8170 iteration : 0.263832\n",
      "cost after 8180 iteration : 0.263765\n",
      "cost after 8190 iteration : 0.263698\n",
      "cost after 8200 iteration : 0.263631\n",
      "cost after 8210 iteration : 0.263564\n",
      "cost after 8220 iteration : 0.263497\n",
      "cost after 8230 iteration : 0.263430\n",
      "cost after 8240 iteration : 0.263364\n",
      "cost after 8250 iteration : 0.263298\n",
      "cost after 8260 iteration : 0.263231\n",
      "cost after 8270 iteration : 0.263165\n",
      "cost after 8280 iteration : 0.263099\n",
      "cost after 8290 iteration : 0.263033\n",
      "cost after 8300 iteration : 0.262968\n",
      "cost after 8310 iteration : 0.262902\n",
      "cost after 8320 iteration : 0.262837\n",
      "cost after 8330 iteration : 0.262771\n",
      "cost after 8340 iteration : 0.262706\n",
      "cost after 8350 iteration : 0.262641\n",
      "cost after 8360 iteration : 0.262576\n",
      "cost after 8370 iteration : 0.262511\n",
      "cost after 8380 iteration : 0.262446\n",
      "cost after 8390 iteration : 0.262382\n",
      "cost after 8400 iteration : 0.262317\n",
      "cost after 8410 iteration : 0.262253\n",
      "cost after 8420 iteration : 0.262189\n",
      "cost after 8430 iteration : 0.262125\n",
      "cost after 8440 iteration : 0.262061\n",
      "cost after 8450 iteration : 0.261997\n",
      "cost after 8460 iteration : 0.261933\n",
      "cost after 8470 iteration : 0.261869\n",
      "cost after 8480 iteration : 0.261806\n",
      "cost after 8490 iteration : 0.261742\n",
      "cost after 8500 iteration : 0.261679\n",
      "cost after 8510 iteration : 0.261616\n",
      "cost after 8520 iteration : 0.261553\n",
      "cost after 8530 iteration : 0.261490\n",
      "cost after 8540 iteration : 0.261427\n",
      "cost after 8550 iteration : 0.261365\n",
      "cost after 8560 iteration : 0.261302\n",
      "cost after 8570 iteration : 0.261240\n",
      "cost after 8580 iteration : 0.261177\n",
      "cost after 8590 iteration : 0.261115\n",
      "cost after 8600 iteration : 0.261053\n",
      "cost after 8610 iteration : 0.260991\n",
      "cost after 8620 iteration : 0.260929\n",
      "cost after 8630 iteration : 0.260867\n",
      "cost after 8640 iteration : 0.260806\n",
      "cost after 8650 iteration : 0.260744\n",
      "cost after 8660 iteration : 0.260683\n",
      "cost after 8670 iteration : 0.260621\n",
      "cost after 8680 iteration : 0.260560\n",
      "cost after 8690 iteration : 0.260499\n",
      "cost after 8700 iteration : 0.260438\n",
      "cost after 8710 iteration : 0.260377\n",
      "cost after 8720 iteration : 0.260316\n",
      "cost after 8730 iteration : 0.260256\n",
      "cost after 8740 iteration : 0.260195\n",
      "cost after 8750 iteration : 0.260135\n",
      "cost after 8760 iteration : 0.260074\n",
      "cost after 8770 iteration : 0.260014\n",
      "cost after 8780 iteration : 0.259954\n",
      "cost after 8790 iteration : 0.259894\n",
      "cost after 8800 iteration : 0.259834\n",
      "cost after 8810 iteration : 0.259774\n",
      "cost after 8820 iteration : 0.259715\n",
      "cost after 8830 iteration : 0.259655\n",
      "cost after 8840 iteration : 0.259596\n",
      "cost after 8850 iteration : 0.259536\n",
      "cost after 8860 iteration : 0.259477\n",
      "cost after 8870 iteration : 0.259418\n",
      "cost after 8880 iteration : 0.259359\n",
      "cost after 8890 iteration : 0.259300\n",
      "cost after 8900 iteration : 0.259241\n",
      "cost after 8910 iteration : 0.259182\n",
      "cost after 8920 iteration : 0.259124\n",
      "cost after 8930 iteration : 0.259065\n",
      "cost after 8940 iteration : 0.259007\n",
      "cost after 8950 iteration : 0.258949\n",
      "cost after 8960 iteration : 0.258890\n",
      "cost after 8970 iteration : 0.258832\n",
      "cost after 8980 iteration : 0.258774\n",
      "cost after 8990 iteration : 0.258717\n",
      "cost after 9000 iteration : 0.258659\n",
      "cost after 9010 iteration : 0.258601\n",
      "cost after 9020 iteration : 0.258543\n",
      "cost after 9030 iteration : 0.258486\n",
      "cost after 9040 iteration : 0.258429\n",
      "cost after 9050 iteration : 0.258371\n",
      "cost after 9060 iteration : 0.258314\n",
      "cost after 9070 iteration : 0.258257\n",
      "cost after 9080 iteration : 0.258200\n",
      "cost after 9090 iteration : 0.258143\n",
      "cost after 9100 iteration : 0.258086\n",
      "cost after 9110 iteration : 0.258030\n",
      "cost after 9120 iteration : 0.257973\n",
      "cost after 9130 iteration : 0.257917\n",
      "cost after 9140 iteration : 0.257860\n",
      "cost after 9150 iteration : 0.257804\n",
      "cost after 9160 iteration : 0.257748\n",
      "cost after 9170 iteration : 0.257692\n",
      "cost after 9180 iteration : 0.257636\n",
      "cost after 9190 iteration : 0.257580\n",
      "cost after 9200 iteration : 0.257524\n",
      "cost after 9210 iteration : 0.257468\n",
      "cost after 9220 iteration : 0.257413\n",
      "cost after 9230 iteration : 0.257357\n",
      "cost after 9240 iteration : 0.257302\n",
      "cost after 9250 iteration : 0.257247\n",
      "cost after 9260 iteration : 0.257191\n",
      "cost after 9270 iteration : 0.257136\n",
      "cost after 9280 iteration : 0.257081\n",
      "cost after 9290 iteration : 0.257026\n",
      "cost after 9300 iteration : 0.256971\n",
      "cost after 9310 iteration : 0.256917\n",
      "cost after 9320 iteration : 0.256862\n",
      "cost after 9330 iteration : 0.256807\n",
      "cost after 9340 iteration : 0.256753\n",
      "cost after 9350 iteration : 0.256699\n",
      "cost after 9360 iteration : 0.256644\n",
      "cost after 9370 iteration : 0.256590\n",
      "cost after 9380 iteration : 0.256536\n",
      "cost after 9390 iteration : 0.256482\n",
      "cost after 9400 iteration : 0.256428\n",
      "cost after 9410 iteration : 0.256374\n",
      "cost after 9420 iteration : 0.256321\n",
      "cost after 9430 iteration : 0.256267\n",
      "cost after 9440 iteration : 0.256213\n",
      "cost after 9450 iteration : 0.256160\n",
      "cost after 9460 iteration : 0.256107\n",
      "cost after 9470 iteration : 0.256053\n",
      "cost after 9480 iteration : 0.256000\n",
      "cost after 9490 iteration : 0.255947\n",
      "cost after 9500 iteration : 0.255894\n",
      "cost after 9510 iteration : 0.255841\n",
      "cost after 9520 iteration : 0.255788\n",
      "cost after 9530 iteration : 0.255735\n",
      "cost after 9540 iteration : 0.255683\n",
      "cost after 9550 iteration : 0.255630\n",
      "cost after 9560 iteration : 0.255578\n",
      "cost after 9570 iteration : 0.255525\n",
      "cost after 9580 iteration : 0.255473\n",
      "cost after 9590 iteration : 0.255421\n",
      "cost after 9600 iteration : 0.255369\n",
      "cost after 9610 iteration : 0.255317\n",
      "cost after 9620 iteration : 0.255265\n",
      "cost after 9630 iteration : 0.255213\n",
      "cost after 9640 iteration : 0.255161\n",
      "cost after 9650 iteration : 0.255109\n",
      "cost after 9660 iteration : 0.255058\n",
      "cost after 9670 iteration : 0.255006\n",
      "cost after 9680 iteration : 0.254955\n",
      "cost after 9690 iteration : 0.254903\n",
      "cost after 9700 iteration : 0.254852\n",
      "cost after 9710 iteration : 0.254801\n",
      "cost after 9720 iteration : 0.254750\n",
      "cost after 9730 iteration : 0.254699\n",
      "cost after 9740 iteration : 0.254648\n",
      "cost after 9750 iteration : 0.254597\n",
      "cost after 9760 iteration : 0.254546\n",
      "cost after 9770 iteration : 0.254495\n",
      "cost after 9780 iteration : 0.254445\n",
      "cost after 9790 iteration : 0.254394\n",
      "cost after 9800 iteration : 0.254344\n",
      "cost after 9810 iteration : 0.254293\n",
      "cost after 9820 iteration : 0.254243\n",
      "cost after 9830 iteration : 0.254193\n",
      "cost after 9840 iteration : 0.254143\n",
      "cost after 9850 iteration : 0.254093\n",
      "cost after 9860 iteration : 0.254043\n",
      "cost after 9870 iteration : 0.253993\n",
      "cost after 9880 iteration : 0.253943\n",
      "cost after 9890 iteration : 0.253893\n",
      "cost after 9900 iteration : 0.253844\n",
      "cost after 9910 iteration : 0.253794\n",
      "cost after 9920 iteration : 0.253745\n",
      "cost after 9930 iteration : 0.253695\n",
      "cost after 9940 iteration : 0.253646\n",
      "cost after 9950 iteration : 0.253597\n",
      "cost after 9960 iteration : 0.253548\n",
      "cost after 9970 iteration : 0.253499\n",
      "cost after 9980 iteration : 0.253450\n",
      "cost after 9990 iteration : 0.253401\n",
      "\n",
      "Train accuracy: 91.95979899497488 %\n",
      "\n",
      "Test accuracy: 91.81286549707602 %\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "init()"
   ]
  },
  {
   "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}