Divorce.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 528,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import random\n",
    "import matplotlib.pyplot as plt\n",
    "from MyGradUpdated.MyGrad.mygrad.tensor_base import Tensor\n",
    "from MyGradUpdated.MyGrad.mygrad.nnet.activations import sigmoid, relu\n",
    "from MyGradUpdated.MyGrad.mygrad.nnet.losses import SoftmaxCrossEntropy\n",
    "from MyGradUpdated.MyGrad.mygrad.math import log\n",
    "from MyGradUpdated.MyGrad.mygrad.nnet.layers import dense\n",
    "import pandas as pd\n",
    "from itertools import product "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 536,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 3.  4.  3. ...,  4.  4.  4.]\n",
      " [ 3.  3.  4. ...,  4.  4.  4.]\n",
      " [ 1.  1.  0. ...,  3.  0.  0.]\n",
      " ..., \n",
      " [ 4.  3.  2. ...,  4.  3.  4.]\n",
      " [ 0.  0.  0. ...,  4.  2.  2.]\n",
      " [ 4.  4.  3. ...,  3.  3.  4.]]\n",
      "[ 1.  1.  0.  0.  1.  1.  0.  0.  1.  0.  1.  1.  1.  1.  0.  0.  0.  1.\n",
      "  0.  1.  0.  0.  0.  1.  0.  0.  0.  0.  0.  0.  1.  0.  0.  0.  1.  0.\n",
      "  0.  1.  1.  0.  1.  0.  0.  0.  1.  1.  1.  1.  0.  0.  1.  1.  1.  1.\n",
      "  0.  0.  1.  1.  1.  1.  1.  0.  1.  0.  0.  1.  0.  1.  1.  0.  1.  0.\n",
      "  1.  1.  1.  1.  0.  1.  0.  1.  1.  1.  1.  0.  1.  1.  1.  0.  1.  1.\n",
      "  0.  0.  0.  1.  0.  1.  0.  0.  0.  1.  1.  0.  0.  0.  0.  0.  0.  1.\n",
      "  1.  1.  1.  0.  1.  1.  1.  0.  0.  1.  1.  1.  0.  1.  0.  1.  0.  0.\n",
      "  1.  1.  0.  0.  0.  1.  1.  1.  0.  1.]\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Data Preprocessing: dividing training data from testing data\"\"\"\n",
    "fileName = \"divorce.xlsx\" #you can change this string file name\n",
    "data_frame_convert = pd.read_excel(fileName)\n",
    "parseArray = data_frame_convert.values\n",
    "\n",
    "train_input = parseArray[:136, :-2]\n",
    "train_output = parseArray[:136, -2]\n",
    "\n",
    "test_input = parseArray[136: ,:-2]\n",
    "test_output = parseArray[136:,-2]\n",
    "\n",
    "print(train_input)\n",
    "print(train_output)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 358,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\"\"\"Neural Net Functions\"\"\"\n",
    "\n",
    "def he_normal(shape):\n",
    "    N = shape[0] #gets how many \"images\" there are or \"items\"\n",
    "    scale = 1 / np.sqrt(2*N)\n",
    "    return np.random.randn(*shape)*scale\n",
    "\n",
    "def fwd_pass(x, weights_1, bias_1, weights_2, bias_2):\n",
    "    o1 = relu(dense(x, weights_1.data) + bias_1.data)\n",
    "    o2 = sigmoid(dense(o1, weights_2.data) + bias_2.data)\n",
    "    return o2\n",
    "    \n",
    "def cross_entropy(p_pred, p_true):\n",
    "    N = p_pred.shape[0]\n",
    "    p_logq = (p_true) * log(p_pred)\n",
    "    return (-1/ N) * p_logq.sum()  \n",
    "    \n",
    "def gradient_descent(param, rate):\n",
    "    print(param.grad)\n",
    "    \n",
    "def compute_accuracy(pred, actual):\n",
    "    #inputs are assumed to be a list\n",
    "    total = 0\n",
    "    for i in range(len(pred)):\n",
    "        if pred[i] == actual[i]:\n",
    "            total += 1\n",
    "    return total / len(pred)\n",
    "\n",
    "def mse(result, actual):\n",
    "    errors = [(actual - calculated) ** 2 for\n",
    "              calculated, actual in list(zip(result, actual))]   \n",
    "    return sum(errors) / len(result)\n",
    "\n",
    "def rounding(x):\n",
    "    if x >= 0.9:\n",
    "        return 1\n",
    "    return 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 532,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1e-07 10000.0\n",
      "1e-07 1000.0\n",
      "1e-07 100.0\n",
      "1e-05 10000.0\n",
      "1e-05 1000.0\n",
      "1e-05 100.0\n",
      "1e-06 10000.0\n",
      "1e-06 1000.0\n",
      "1e-06 100.0\n",
      "1e-08 10000.0\n",
      "1e-08 1000.0\n",
      "1e-08 100.0\n",
      "{(1e-07, 10000.0): (0.23986727309079423, 0.25040670833584394), (1e-07, 1000.0): (0.1916134594008835, 0.2643617323926168), (1e-07, 100.0): (0.17678666615387778, 0.2861755754686406), (1e-05, 10000.0): (0.25956102900025146, 0.30588488056972524), (1e-05, 1000.0): (0.23716397985383222, 0.31982278906076783), (1e-05, 100.0): (0.265699793490664, 0.4030574720806593), (1e-06, 10000.0): (0.23760916939112786, 0.25465756338199386), (1e-06, 1000.0): (0.2095100305806184, 0.2619138823533212), (1e-06, 100.0): (0.22540693972645412, 0.37169028892418193), (1e-08, 10000.0): (0.255503917306982, 0.25036846209983227), (1e-08, 1000.0): (0.25800522750948335, 0.25959770863438747), (1e-08, 100.0): (0.2019899699728047, 0.2709409002485619)}\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Cross Validation to determine best step-size and regularization factor\"\"\"\n",
    "iterations = 8000\n",
    "\n",
    "results = {}\n",
    "best_mse = float('inf')   # The highest validation accuracy that we have seen so far.\n",
    "best_loss = []  # list of loss values measured during best training session\n",
    "best_MSEs = []  # list of validation accuracies measured during best training session\n",
    "step_size = [1e-7,1e-5,1e-6, 1e-8]\n",
    "reg_strengths = [1e4,1e3,1e2]\n",
    "your_various_rate_reg_pairs = product(step_size, reg_strengths)\n",
    "flag = True\n",
    "for rate, reg in your_various_rate_reg_pairs:\n",
    "    #Weights init\n",
    "    layer_1_weights = Tensor(he_normal((54, 170)))\n",
    "    #print(\"WEIGHTS\", layer_1_weights)\n",
    "    bias_1 = Tensor(np.zeros((170,), dtype=layer_1_weights.dtype))\n",
    "    layer_2_weights = Tensor(he_normal((170, 1)))\n",
    "    bias_2 = Tensor(np.zeros((1,), dtype=layer_2_weights.dtype))\n",
    "    params = [layer_1_weights, layer_2_weights, bias_1, bias_2]\n",
    "    print(rate, reg)\n",
    "    \n",
    "    losses = []  # will store loss for each training iteration\n",
    "    liveMSE = []  # will store training accuracy for each training iteration\n",
    "    \n",
    "    for it in range(iterations):\n",
    "        mask = np.random.choice(len(train_input), size=34, replace=False)\n",
    "        inputs = train_input[mask]\n",
    "        truths = train_output[mask]\n",
    "        \n",
    "        inputs = Tensor(inputs)\n",
    "        truths = Tensor(truths)\n",
    "        \n",
    "        o1 = relu(dense(inputs, layer_1_weights) + bias_1)\n",
    "        pred = sigmoid(dense(o1, layer_2_weights) + bias_2)\n",
    "        reg_term = reg * (layer_2_weights**2).sum() #this is regularlization value, Loss big = bad, loss small = good. \n",
    "        #reg is bad when all the parameters are really bad  Try to keep elements in W small so it does not overfit\n",
    "        loss = cross_entropy(pred, truths) + reg_term \n",
    "        #how good are our predictions and are we keeping the weights moderate. \n",
    "        loss.backward()\n",
    "        #gives the gradient/values for W.grad and b.grad\n",
    "         \n",
    "        for param in params:\n",
    "            param.data -= rate*param.grad\n",
    "        \n",
    "        # !!! Update model parameters using gradient descent with!!!\n",
    "        ### with the correct learning rate\n",
    "        loss.null_gradients()  # this is super important!!\n",
    "        \n",
    "        predicted_list = [float(ten.data) for ten in list(pred)]\n",
    "        #val_list = list(map(rounding, list(pred)))       \n",
    "        #train_accuracy = compute_accuracy(val_list, truths.data.tolist())\n",
    "        # the accuracy for this training batch\n",
    "        current_mse = mse(predicted_list, truths.data.tolist())\n",
    "        liveMSE.append(current_mse)\n",
    "        losses.append(loss.data.item())\n",
    "       \n",
    "    # Training is complete for this hyperparameter configuration\n",
    "    \n",
    "    # Use your trained model parameters to compute the predicted scores\n",
    "    # on the training data. Use this to compute the training-accuracy.\n",
    "    # save this to `train_acc`\n",
    "    \n",
    "    train_predict = fwd_pass(train_input, layer_1_weights, bias_1, layer_2_weights, bias_2)\n",
    "    predicted_list = [float(ten.data) for ten in list(train_predict)]\n",
    "    #what if you gradiented instead\n",
    "    \"\"\"for i, tensor in enumerate(predicted_list):\n",
    "        value = 0\n",
    "        if float(tensor.data) >= 0.95:\n",
    "            value = 1\n",
    "        predicted_list[i] = value\n",
    "    \"\"\"    \n",
    "    #train_acc = compute_accuracy(predicted_list, train_output.tolist())\n",
    "    train_mse = mse(predicted_list, train_output.tolist())\n",
    "\n",
    "    # Use your trained model parameters to compute the predicted scores\n",
    "    # on the validation data. Use this to compute the training-accuracy.\n",
    "    # save this to `val_acc`\n",
    "    \n",
    "    val_predict = fwd_pass(test_input, layer_1_weights, bias_1, layer_2_weights, bias_2)\n",
    "    val_list = [float(ten.data) for ten in list(val_predict)]\n",
    "    \"\"\"for i, tensor in enumerate(val_list):\n",
    "        value = 0\n",
    "        if float(tensor.data) >= 0.95:\n",
    "            value = 1\n",
    "        val_list[i] = value\n",
    "        \"\"\"\n",
    "        \n",
    "    #val_acc = compute_accuracy(val_list, test_output.tolist())\n",
    "    val_mse = mse(val_list, train_output.tolist())\n",
    "        \n",
    "    results[(rate, reg)] = (train_mse, val_mse)\n",
    "    if best_mse > val_mse:\n",
    "        best_loss = losses\n",
    "        best_MSEs = liveMSE\n",
    "        best_mse = val_mse\n",
    "        best_train = train_acc\n",
    "print(results)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 546,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAEWCAYAAABsY4yMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJ3jnnKtQnuCdc65C/X+Sn/R5ZkRPBgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2b619685f60>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Visualize the cross-validation results\n",
    "import matplotlib.cm as cm\n",
    "import math\n",
    "x_scatter = [math.log10(x[0]) for x in results]\n",
    "y_scatter = [math.log10(x[1]) for x in results]\n",
    "\n",
    "# plot training accuracy\n",
    "marker_size = 100\n",
    "colors = [results[x][0] for x in results]\n",
    "#plt.subplot(2, 1, 1)\n",
    "plt.scatter(x_scatter, y_scatter, marker_size, c=colors, cmap=cm.viridis)\n",
    "plt.colorbar()\n",
    "plt.xlabel('log learning rate')\n",
    "plt.ylabel('log regularization strength')\n",
    "plt.title('Divorce training MSE')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 547,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAEWCAYAAABsY4yMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZ0uwsR9nHAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2b60f954588>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot validation accuracy\n",
    "colors = [results[x][1] for x in results] # default size of markers is 20\n",
    "#plt.subplot(2, 1, 2)\n",
    "plt.scatter(x_scatter, y_scatter, marker_size, c=colors,  cmap=cm.viridis)\n",
    "plt.colorbar()\n",
    "plt.xlabel('log learning rate')\n",
    "plt.ylabel('log regularization strength')\n",
    "plt.title(\"Divorce testing MSE\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 538,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "1000\n",
      "2000\n",
      "3000\n",
      "4000\n",
      "5000\n",
      "6000\n",
      "7000\n",
      "8000\n",
      "9000\n",
      "10000\n",
      "11000\n",
      "12000\n",
      "13000\n",
      "14000\n",
      "15000\n",
      "16000\n",
      "17000\n",
      "18000\n",
      "19000\n",
      "test_input [ 2.  0.  2.  4.  2.  2.  4.  3.  4.  3.  2.  3.  3.  0.  4.  4.  3.  3.\n",
      "  3.  2.  2.  1.  2.  0.  0.  0.  0.  0.  2.  2.  3.  3.  3.  3.  3.  3.\n",
      "  3.  3.  1.  3.  2.  3.  3.  4.  4.  2.  3.  2.  2.  3.  3.  4.  2.  2.]\n",
      "test_output Tensor([ 0.77198076])\n",
      "should be greater than 0.75 and divorced\n",
      "model's MSE on test data 0.2530412012610141\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Hyperparameters: Stepsize = 1e-6.  Regularization = 1000\"\"\"\n",
    "\"Actual model training\"\n",
    "layer_1_weights = Tensor(he_normal((54, 170)))\n",
    "bias_1 = Tensor(np.zeros((170,), dtype=layer_1_weights.dtype))\n",
    "layer_2_weights = Tensor(he_normal((170, 1)))\n",
    "bias_2 = Tensor(np.zeros((1,), dtype=layer_2_weights.dtype))\n",
    "params = [layer_1_weights, layer_2_weights, bias_1, bias_2]\n",
    "\n",
    "rate = 1e-7\n",
    "reg = 1000\n",
    "\n",
    "iterations = 20000\n",
    "\n",
    "liveMSE = []\n",
    "losses = []\n",
    "\n",
    "for it in range(iterations):\n",
    "    if (it % 1000 == 0):\n",
    "        print(it)\n",
    "        \n",
    "    inputs = Tensor(train_input)\n",
    "    truths = Tensor(train_output)\n",
    "\n",
    "    o1 = relu(dense(inputs, layer_1_weights) + bias_1)\n",
    "    pred = sigmoid(dense(o1, layer_2_weights) + bias_2)\n",
    "    reg_term = reg * (layer_2_weights**2).sum() #this is regularlization value, Loss big = bad, loss small = good. \n",
    "    #reg is bad when all the parameters are really bad  Try to keep elements in W small so it does not overfit\n",
    "    loss = cross_entropy(pred, truths) + reg_term \n",
    "    #how good are our predictions and are we keeping the weights moderate. \n",
    "    loss.backward()\n",
    "    #gives the gradient/values for W.grad and b.grad\n",
    "\n",
    "    for param in params:\n",
    "        param.data -= rate*param.grad\n",
    "\n",
    "    # !!! Update model parameters using gradient descent with!!!\n",
    "    ### with the correct learning rate\n",
    "    loss.null_gradients()  # this is super important!!\n",
    "\n",
    "    predicted_list = [float(ten.data) for ten in list(pred)]\n",
    "    #val_list = list(map(rounding, list(pred)))       \n",
    "    #train_accuracy = compute_accuracy(val_list, truths.data.tolist())\n",
    "    # the accuracy for this training batch\n",
    "    current_mse = mse(predicted_list, truths.data.tolist())\n",
    "    liveMSE.append(current_mse)\n",
    "    losses.append(loss.data.item())\n",
    "\n",
    "print(\"test_input\", test_input[10])\n",
    "test_result = fwd_pass(Tensor(test_input), layer_1_weights, bias_1, layer_2_weights, bias_2)\n",
    "print(\"test_output\", test_result[10])\n",
    "print(\"should be greater than 0.75 and divorced\")\n",
    "test_list = [float(ten.data) for ten in list(test_result)]\n",
    "final_mse = mse(test_list, test_output.tolist())\n",
    "\n",
    "print(\"model's MSE on test data\", final_mse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 540,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test_input [ 1.  1.  0.  0.  0.  0.  0.  0.  0.  1.  0.  0.  1.  0.  0.  0.  0.  0.\n",
      "  0.  0.  1.  0.  0.  0.  1.  0.  1.  1.  1.  1.  0.  0.  0.  0.  0.  0.\n",
      "  0.  0.  0.  0.  0.  1.  2.  0.  2.  1.  0.  0.  1.  1.  1.  1.  1.  1.]\n",
      "test_output Tensor([ 0.58261788])\n",
      "safe\n",
      "model's MSE on test data 0.2530412012610141\n"
     ]
    }
   ],
   "source": [
    "\"Testing the model and its MSE and analyzing individual results\"\n",
    "print(\"test_input\", test_input[31])\n",
    "test_result = fwd_pass(Tensor(test_input), layer_1_weights, bias_1, layer_2_weights, bias_2)\n",
    "print(\"test_output\", test_result[31])\n",
    "if ((float(test_result[31].data))>0.75):\n",
    "    print(\"divorce\")\n",
    "else:\n",
    "    print(\"safe\")\n",
    "test_list = [float(ten.data) for ten in list(test_result)]\n",
    "final_mse = mse(test_list, test_output.tolist())\n",
    "\n",
    "print(\"model's MSE on test data\", final_mse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 541,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.0\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Showing the accuracy of each prediction within the testing data set.  Note it is most often \n",
    "1.0 because our dataset is very small (34 entries)\"\"\"\n",
    "outputs = []\n",
    "test_result = fwd_pass(Tensor(test_input), layer_1_weights, bias_1, layer_2_weights, bias_2)\n",
    "for prediction in test_result:\n",
    "    if (float(prediction.data) > 0.75):\n",
    "        outputs.append(1)\n",
    "    else:\n",
    "        outputs.append(0)\n",
    "\n",
    "total = 0\n",
    "for i, output in enumerate(outputs):\n",
    "    if test_output.tolist()[i] == output:\n",
    "        total += 1\n",
    "print(total / len(test_output.tolist()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 542,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAEWCAYAAABollyxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRCTJKehFRJKcgl5EJMn9f+3Ch19O9hBsAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2b6196855f8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\"\"\"Plotting loss function over time \"\"\"\n",
    "fig,(ax) = plt.subplots(nrows=1)\n",
    "plt.title(\"Cross Entropy Loss over Time\")\n",
    "ax.plot(losses)\n",
    "plt.xlabel('Iterations')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 543,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX4AAAEWCAYAAABhffzLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2b616ed3630>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\"\"\"Plotting MSE over time.  Note: please be aware of the scale\"\"\"\n",
    "fig,(ax2) = plt.subplots(nrows=1)\n",
    "plt.title(\"MSE over time\")\n",
    "ax2.plot(liveMSE)\n",
    "plt.xlabel('Iterations')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 548,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[12.55774986325011, 11.58594221428723, 13.382600643063139, 12.662836145695891, 12.182936909200217, 13.40630389026803, 12.93501274268138, 13.094060374445203, 13.279946084146419, 13.552026248924191, 14.732561368543873, 12.757616731857981, 11.44783927407164, 13.210253768124595, 14.309080167964284, 12.340094624533215, 13.774704951040837, 13.587079975390177, 12.259341493992178, 14.950831346557507, 13.003612145545382, 13.077133555191898, 14.49707465114484, 12.951722586289995, 12.252057479579548, 11.477045317918572, 13.242116732749489, 14.086599898530107, 11.28568282402728, 13.707539038987637, 13.579228434043877, 12.949731899042856, 12.510008970843485, 12.683935474489552, 13.337476988013796, 12.488589621857663, 14.567530764861411, 12.730467797155065, 13.606837475229902, 12.472571638519861, 13.194194747297866, 13.157964528385438, 13.31419535425546, 12.296842421470057, 14.317861089542571, 12.996991146502873, 12.94326038632644, 13.410860883371772, 13.530916685031753, 13.024997835153235, 12.78270562849871, 13.836271283769928, 14.364582423659023, 13.110761610557647]\n",
      "54\n",
      "Important questions [11, 15, 20, 23, 28, 37, 45, 53]\n"
     ]
    }
   ],
   "source": [
    "\"\"\"Determining which questions hold more importance by analyzing the magnitudes of their weights. \n",
    "We can compare these questions to the one that the datasource indicates as important to test reliability\n",
    "and consistency of model\"\"\"\n",
    "sums = []\n",
    "for row in layer_1_weights:\n",
    "    total = 0\n",
    "    for elem in row.data.tolist():\n",
    "        total += abs(elem)\n",
    "    sums.append(total)\n",
    "\n",
    "print(sums)\n",
    "print(len(sums))\n",
    "\n",
    "indices = []\n",
    "for i, elem in enumerate(sums):\n",
    "    if elem >= 14:\n",
    "        indices.append(i+1)\n",
    "        \n",
    "print(\"Important questions\", indices)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}