Neural Network.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Neural Network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Neural Network For Continuous Input\n",
    "\n",
    "\n",
    "* Load CSV Iris dataset\n",
    "* Construct Neural Network Classifier\n",
    "* Fit the model with the trainning data\n",
    "* Evaluate the accuracy of the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the iris CSV data to TensorFlow\n",
    "\n",
    "Iris data set contains 150 rows of data, comprising 50 samples from each of three related Iris species: Iris setosa, Iris virginica, and Iris versicolor."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from six.moves import urllib\n",
    "import os\n",
    "\n",
    "def download_file_from_url(source_url, dest_path):\n",
    "    # Create directory if not already existed\n",
    "    path_folder = '/'.join(dest_path.split('/')[:-1])\n",
    "    if not os.path.exists(path_folder):\n",
    "        os.mkdir(path_folder)\n",
    "        print (\"Making path_folder: %s\" % path_folder)\n",
    "    urllib.request.urlretrieve(source_url, dest_path)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Making path_folder: /tmp/workspace/continuous-dnn\n"
     ]
    }
   ],
   "source": [
    "iris_data_path = \"/tmp/workspace/continuous-dnn/iris.data\"\n",
    "iris_data_source_url = \"https://archive.ics.uci.edu/ml/machine-learning-databases/iris/iris.data\"\n",
    "\n",
    "download_file_from_url(iris_data_source_url, iris_data_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "IRIS_COLUMNS = [\"sepal_length\", \"sepal_width\", \"petal_length\", \"petal_width\", \"species\"]\n",
    "\n",
    "IRIS_CONTINUOUS_COLUMNS = [\"sepal_length\", \"sepal_width\", \"petal_length\", \"petal_width\"]\n",
    "\n",
    "IRIS_LABEL_COLUMN = 'species'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data exploration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "\n",
    "def read_csv_with_column(csv_file_path, column_names):\n",
    "    return pd.read_csv(csv_file_path, names=column_names, skipinitialspace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 328,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df_iris_data = read_csv_with_column(train_data_path, IRIS_COLUMNS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convert labels to different int values\n",
    "\n",
    "Since classifier can only work with int32/int64, we need to convert the training and testing lable data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 327,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Convert categorical data to real values with pandas.Categorical\n",
    "def convert_categorical_nomial(df, column_name):\n",
    "    return pd.to_numeric(pd.Categorical.from_array(df[column_name]).codes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 329,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Convert the caregorical column to int values\n",
    "df_iris_data[IRIS_LABEL_COLUMN] = convert_categorical_nomial(df_iris_data, IRIS_LABEL_COLUMN)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 334,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# the above computation equals the following\n",
    "# df_iris_data[IRIS_LABEL_COLUMN] = df_data[\"species\"].apply(lambda x : 0 if 'Iris-setosa' in x else (1 if 'Iris-virginica' in x else 2))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 335,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW0AAAELCAYAAAASrNdlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x12354cfd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# enable pylab mode in ipython notebook as pylab has been deprecated\n",
    "%matplotlib inline\n",
    "def group_n_plot_by_label(df, label_column):\n",
    "    df.groupby(label_column).size().plot(kind=\"bar\")\n",
    "    \n",
    "group_n_plot_by_label(df_iris_data, IRIS_LABEL_COLUMN)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 336,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Divid data into testing data and \n",
    "def split_data(df, sample_rate, random_state=15):\n",
    "    df_train = df.sample(frac=0.8, random_state=random_state)\n",
    "    df_test = df.drop(df_train.index)\n",
    "    return df_train, df_test\n",
    "\n",
    "# Verify that the train data and test data are evenly sampled\n",
    "def plot_train_test_data(df_train, df_test, label_column):\n",
    "    group_n_plot_by_label(df_train, label_column)\n",
    "    group_n_plot_by_label(df_test, label_column)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 337,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW0AAAELCAYAAAASrNdlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1259ce090>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_iris_train, df_iris_test = split_data(df_iris_data, 0.8)\n",
    "\n",
    "plot_train_test_data(df_iris_train, df_iris_test, IRIS_LABEL_COLUMN)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convert data to tensor\n",
    "Since all input are continuous input, use tf.constant is a good choice"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 338,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def input_fn(df, continuous_columns, label_column):\n",
    "    feature_cols = {k : tf.constant(df[k].values) for k in continuous_columns}                        \n",
    "    label = tf.constant(df[label_column].values)\n",
    "    return feature_cols, label"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 339,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def iris_train_input_fn():\n",
    "    return input_fn(df_iris_train, IRIS_CONTINUOUS_COLUMNS, IRIS_LABEL_COLUMN)\n",
    "\n",
    "def iris_test_input_fn():\n",
    "    return input_fn(df_iris_test, IRIS_CONTINUOUS_COLUMNS, IRIS_LABEL_COLUMN)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create feature columns\n",
    "We replaced the 'class' column with 'label' column, so all of them a real valued features: \n",
    "* real_valued_column"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 322,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# \"sepal_length\", \"sepal_width\", \"petal_length\", \"petal_width\"\n",
    "def create_feature_columns(feature_columns_names):\n",
    "    feature_columns = []\n",
    "    for feature_name in feature_columns_names:\n",
    "        feature = tf.contrib.layers.real_valued_column(feature_name)\n",
    "        feature_columns.append(feature)\n",
    "    return feature_columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 323,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "iris_feature_column = create_feature_columns(IRIS_CONTINUOUS_COLUMNS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create DNN Model\n",
    "* DNNClassifier\n",
    "* feature_columns = feature_columns\n",
    "* hidden_units = [10, 20, 10], three hidden layers, containing 10, 20, and 10 neurons.\n",
    "* n_classes=3, Three target classes, representing the three iris species\n",
    "* model_model Tensorflow use it to save check point data during model trainning."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 340,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Using default config.\n",
      "INFO:tensorflow:Using config: {'_save_checkpoints_secs': 600, '_num_ps_replicas': 0, '_keep_checkpoint_max': 5, '_tf_random_seed': None, '_task_type': None, '_environment': 'local', '_is_chief': True, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x125b91050>, '_tf_config': gpu_options {\n",
      "  per_process_gpu_memory_fraction: 1\n",
      "}\n",
      ", '_task_id': 0, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_evaluation_master': '', '_keep_checkpoint_every_n_hours': 10000, '_master': ''}\n"
     ]
    }
   ],
   "source": [
    "model_dir = \"/tmp/workspace/continuous-dnn/\"\n",
    "iris_classifier = tf.contrib.learn.DNNClassifier(feature_columns=iris_feature_column,\n",
    "                                            hidden_units=[10, 20, 10],\n",
    "                                            n_classes=3,\n",
    "                                            model_dir=model_dir)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fit the DNNClassifier to the iris training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 341,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:From /Users/xozhang/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/tensorflow/contrib/learn/python/learn/estimators/head.py:1362: scalar_summary (from tensorflow.python.ops.logging_ops) is deprecated and will be removed after 2016-11-30.\n",
      "Instructions for updating:\n",
      "Please switch to tf.summary.scalar. Note that tf.summary.scalar uses the node name instead of the tag. This means that TensorFlow will automatically de-duplicate summary names based on the scope they are created in. Also, passing a tensor or list of tags to a scalar summary op is no longer supported.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Saving checkpoints for 6001 into /tmp/workspace/continuous-dnn/model.ckpt.\n",
      "INFO:tensorflow:loss = 7.81006, step = 6001\n",
      "INFO:tensorflow:global_step/sec: 800.614\n",
      "INFO:tensorflow:loss = 0.0777194, step = 6101\n",
      "INFO:tensorflow:global_step/sec: 727.728\n",
      "INFO:tensorflow:loss = 0.0605825, step = 6201\n",
      "INFO:tensorflow:global_step/sec: 690.327\n",
      "INFO:tensorflow:loss = 0.0542215, step = 6301\n",
      "INFO:tensorflow:global_step/sec: 696.563\n",
      "INFO:tensorflow:loss = 0.0505909, step = 6401\n",
      "INFO:tensorflow:global_step/sec: 832.896\n",
      "INFO:tensorflow:loss = 0.0481317, step = 6501\n",
      "INFO:tensorflow:global_step/sec: 737.883\n",
      "INFO:tensorflow:loss = 0.0462774, step = 6601\n",
      "INFO:tensorflow:global_step/sec: 545.589\n",
      "INFO:tensorflow:loss = 0.0449129, step = 6701\n",
      "INFO:tensorflow:global_step/sec: 752.582\n",
      "INFO:tensorflow:loss = 0.0434973, step = 6801\n",
      "INFO:tensorflow:global_step/sec: 778.992\n",
      "INFO:tensorflow:loss = 0.0423911, step = 6901\n",
      "INFO:tensorflow:global_step/sec: 742.009\n",
      "INFO:tensorflow:loss = 0.0414055, step = 7001\n",
      "INFO:tensorflow:global_step/sec: 705.095\n",
      "INFO:tensorflow:loss = 0.04061, step = 7101\n",
      "INFO:tensorflow:global_step/sec: 752.265\n",
      "INFO:tensorflow:loss = 0.0397081, step = 7201\n",
      "INFO:tensorflow:global_step/sec: 729.145\n",
      "INFO:tensorflow:loss = 0.039451, step = 7301\n",
      "INFO:tensorflow:global_step/sec: 661.928\n",
      "INFO:tensorflow:loss = 0.0382674, step = 7401\n",
      "INFO:tensorflow:global_step/sec: 698.016\n",
      "INFO:tensorflow:loss = 0.0380904, step = 7501\n",
      "INFO:tensorflow:global_step/sec: 593.229\n",
      "INFO:tensorflow:loss = 0.0370452, step = 7601\n",
      "INFO:tensorflow:global_step/sec: 650.115\n",
      "INFO:tensorflow:loss = 0.0366884, step = 7701\n",
      "INFO:tensorflow:global_step/sec: 569.856\n",
      "INFO:tensorflow:loss = 0.0359754, step = 7801\n",
      "INFO:tensorflow:global_step/sec: 726.76\n",
      "INFO:tensorflow:loss = 0.0354981, step = 7901\n",
      "INFO:tensorflow:Saving checkpoints for 8000 into /tmp/workspace/continuous-dnn/model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 0.0349211.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "DNNClassifier(params={'head': <tensorflow.contrib.learn.python.learn.estimators.head._MultiClassHead object at 0x125b91110>, 'hidden_units': [10, 20, 10], 'feature_columns': (_RealValuedColumn(column_name='sepal_length', dimension=1, default_value=None, dtype=tf.float32, normalizer=None), _RealValuedColumn(column_name='sepal_width', dimension=1, default_value=None, dtype=tf.float32, normalizer=None), _RealValuedColumn(column_name='petal_length', dimension=1, default_value=None, dtype=tf.float32, normalizer=None), _RealValuedColumn(column_name='petal_width', dimension=1, default_value=None, dtype=tf.float32, normalizer=None)), 'embedding_lr_multipliers': None, 'optimizer': None, 'dropout': None, 'gradient_clip_norm': None, 'activation_fn': <function relu at 0x10b9199b0>, 'input_layer_min_slice_size': None})"
      ]
     },
     "execution_count": 341,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "iris_classifier.fit(input_fn=iris_train_input_fn, steps = 2000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluate the model accuracy\n",
    "\n",
    "Evaluate the accuracy of the function with test input\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:From /Users/xozhang/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/tensorflow/contrib/learn/python/learn/estimators/head.py:1362: scalar_summary (from tensorflow.python.ops.logging_ops) is deprecated and will be removed after 2016-11-30.\n",
      "Instructions for updating:\n",
      "Please switch to tf.summary.scalar. Note that tf.summary.scalar uses the node name instead of the tag. This means that TensorFlow will automatically de-duplicate summary names based on the scope they are created in. Also, passing a tensor or list of tags to a scalar summary op is no longer supported.\n",
      "INFO:tensorflow:Starting evaluation at 2017-03-28-07:05:44\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Finished evaluation at 2017-03-28-07:05:44\n",
      "INFO:tensorflow:Saving dict for global step 4200: accuracy = 0.966667, auc = 0.998333, global_step = 4200, loss = 0.113523\n",
      "WARNING:tensorflow:Skipping summary for global_step, must be a float or np.float32.\n"
     ]
    }
   ],
   "source": [
    "iris_results = classifier.evaluate(input_fn=test_input_fn, steps=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 251,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'accuracy': 0.96666664,\n",
       " 'auc': 0.99833333,\n",
       " 'global_step': 4200,\n",
       " 'loss': 0.1135229}"
      ]
     },
     "execution_count": 251,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "iris_results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Neural Netowrk With Categorical Input\n",
    "\n",
    "* Load Car Evaluation Data set: http://archive.ics.uci.edu/ml/datasets/Car+Evaluation\n",
    "* Load CSV Data\n",
    "* Constructy Neural Network Classifier\n",
    "* Fit the model with the trainning data\n",
    "* Evaluate the accuracy of the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 304,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "car_data_path = \"/tmp/workspace/categorical-dnn/car.data\"\n",
    "car_data_source_url = \"http://archive.ics.uci.edu/ml/machine-learning-databases/car/car.data\"\n",
    "\n",
    "download_file_from_url(car_data_source_url, car_data_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load car CSV data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 313,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "CAR_COLUMNS = ['buying', 'maint', 'doors', 'persons', 'lug_boot', 'safety', 'classes']\n",
    "\n",
    "CAR_CATEGORICAL_COLUMN = ['buying', 'maint', 'doors', 'persons', 'lug_boot', 'safety']\n",
    "\n",
    "CAR_LABEL_COLUMN = 'classes'\n",
    "\n",
    "df_car_data = read_csv_with_column(car_data_path, CAR_COLUMNS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Covnert categorical data to int values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 342,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def convert_categorical_columns_nomial(df, categorical_columns):\n",
    "    for column_name in categorical_columns:\n",
    "        df[column_name] = convert_categorical_nomial(df, column_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 343,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "convert_categorical_columns_nomial(df_car_data, CAR_COLUMNS)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Split car data into trainning and testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 345,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAELCAYAAADX3k30AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAEOdJREFUeJzt3X+s3XV9x/HnCysgE1nnRq9rgZahrBhRmasuZtmJbiD7\nQVkyCeoURM0SdLIt22h1W6uLP3BxG9vExKhYF5QUN0O3OegQj8ZNgSEIoxWaORAaW7INUaYD6t77\n43wL19L2lnPuveeez30+kpN8z+f7+X6/n/Pt977Op5/vj5OqQpLUrsPG3QBJ0twy6CWpcQa9JDXO\noJekxhn0ktQ4g16SGjdj0Cf5SJLdSW6bVrY0ydYkdya5Nskx0+atT7IjyfYkp08rPy3JbUnuSvLn\ns/9RJEn7cyg9+suBM/YpWwdcV1UnA9cD6wGSnAKcA6wGzgQuS5JumQ8Cb6iq5wDPSbLvOiVJc2DG\noK+qLwIP7FO8FtjUTW8Czu6mzwKurKo9VXU3sANYk2QKOLqqburqfXzaMpKkOTTsGP2xVbUboKp2\nAcd25cuBe6fV29mVLQfum1Z+X1cmSZpjS2ZpPbP6HIUkPpdBkoZQVdm3bNge/e4kywC6YZn7u/Kd\nwHHT6q3oyg5UfrDGLvjXhg0bxt6GVl7uS/fnQn5Nyv48kEMN+nSvvbYA53fT5wFXTys/N8nhSVYB\nJwE31mB458Eka7qTs6+btowkaQ7NOHST5BNAD3hmkm8AG4D3AlcluQC4h8GVNlTVtiSbgW3Ao8CF\n9fjXzJuBjwFHAp+pqmtm96NIkvZnxqCvqlcfYNbPH6D+e4D37Kf8ZuB5T6p1C1yv1xt3E5rhvpxd\n7s/ZNen7Mwcb1xmXJLUQ2yVJC1kSahZPxkqSJoRBL0mNM+glqXEGvSQ1zqCXxmxqaiVJFvxramrl\nuHeVhuRVN9KYDe4hnITjPQe9+1Lj51U3krRIGfSS1DiDXpIaZ9BLUuMMeklqnEEvSY0z6CWpcQa9\nJDXOoJekxhn0ktQ4g16SGmfQS1LjDHpJapxBL0mNM+glqXEGvSQ1zqCXpMYZ9JLUOINekhpn0EtS\n4wx6SWqcQS9JjTPoJalxBr0kNc6gl6TGGfSS1DiDXpIaZ9BLUuMMeklq3EhBn2R9kjuS3JbkiiSH\nJ1maZGuSO5Ncm+SYfervSLI9yemjN1+SNJOhgz7JCcCbgBdW1anAEuBVwDrguqo6GbgeWN/VPwU4\nB1gNnAlcliSjNV+SNJNRevTfBh4BfijJEuBpwE5gLbCpq7MJOLubPgu4sqr2VNXdwA5gzQjblyQd\ngqGDvqoeAN4PfINBwD9YVdcBy6pqd1dnF3Bst8hy4N5pq9jZlUmS5tCSYRdMciLw28AJwIPAVUle\nA9Q+Vfd9f0g2btz42HSv16PX6w3VTklqVb/fp9/vz1gvVUPlMEnOAX6hqt7UvX8t8BLgZUCvqnYn\nmQI+V1Wrk6wDqqou6epfA2yoqhv2s+4atl3SpBmcqpqE4z34d7mwJaGqnnDuc5Qx+juBlyQ5sjup\n+nJgG7AFOL+rcx5wdTe9BTi3uzJnFXAScOMI25ckHYKhh26q6qtJPg7cDHwfuAX4EHA0sDnJBcA9\nDK60oaq2JdnM4MvgUeBCu+2SNPeGHrqZSw7daDFx6EazZS6GbiRJE8Cgl6TGGfSS1DiDXpIaZ9BL\nUuMMeklqnEEvSY0z6CWpcQa9JDXOoJekxhn0ktQ4g16SGmfQS1LjDHpJapxBL0mNM+glqXEGvSQ1\nzqCXpMYZ9JLUOINekhpn0EtS4wx6SWqcQS9JjTPoJalxBr0kNc6gl6TGGfSS1DiDXpIaZ9BLUuMM\neklqnEEvSY0z6CWpcQa9JDXOoJekxhn0ktS4kYI+yTFJrkqyPckdSV6cZGmSrUnuTHJtkmOm1V+f\nZEdX//TRmy9JmsmoPfpLgc9U1Wrg+cDXgHXAdVV1MnA9sB4gySnAOcBq4EzgsiQZcfuSpBkMHfRJ\nngH8bFVdDlBVe6rqQWAtsKmrtgk4u5s+C7iyq3c3sANYM+z2JUmHZpQe/SrgP5NcnuQrST6U5Chg\nWVXtBqiqXcCxXf3lwL3Tlt/ZlUmS5tAoQb8EOA34QFWdBvwPg2Gb2qfevu8lSfNoyQjL3gfcW1X/\n2r3/GwZBvzvJsqranWQKuL+bvxM4btryK7qy/dq4ceNj071ej16vN0JTJak9/X6ffr8/Y71UDd/h\nTvJ54E1VdVeSDcBR3az/rqpLklwMLK2qdd3J2CuAFzMYsvkn4Nm1nwYk2V+x1KTBNQmTcLwH/y4X\ntiRU1RMuchmlRw/wVuCKJE8Fvg68HngKsDnJBcA9DK60oaq2JdkMbAMeBS40zSVp7o3Uo58r9ui1\nmNij12w5UI/eO2MlqXEGvSQ1zqCXpMYZ9JLUOINekhpn0EtS4wx6SWqcQS9JjTPoJalxBr0kNc6g\nl6TGGfSS1DiDXpIaZ9BLUuMMeklqnEEvSY0z6CWpcQa9JDXOoJekxhn0ktQ4g16SGmfQS1LjDHpJ\napxBL0mNM+glqXEGvSQ1zqCXpMYZ9JLUOINekhpn0EtS4wx6SWqcQS9JjTPoJalxBr0kNc6gl6TG\nGfSS1LiRgz7JYUm+kmRL935pkq1J7kxybZJjptVdn2RHku1JTh9125Kkmc1Gj/4iYNu09+uA66rq\nZOB6YD1AklOAc4DVwJnAZUkyC9uXJB3ESEGfZAXwi8CHpxWvBTZ105uAs7vps4Arq2pPVd0N7ADW\njLJ9SdLMRu3R/xnwe0BNK1tWVbsBqmoXcGxXvhy4d1q9nV2ZJGkOLRl2wSS/BOyuqluT9A5StQ4y\n74A2btz42HSv16PXO9gmJGnx6ff79Pv9GeulaqgcJsm7gV8H9gBPA44GPg28COhV1e4kU8Dnqmp1\nknVAVdUl3fLXABuq6ob9rLuGbZc0aQanqibheA/+XS5sSaiqJ5z7HHropqreVlXHV9WJwLnA9VX1\nWuDvgPO7aucBV3fTW4BzkxyeZBVwEnDjsNuXJB2aoYduDuK9wOYkFwD3MLjShqralmQzgyt0HgUu\ntNsuSXNv6KGbueTQjRYTh240W2Z96EaSNBkMeklqnEEvSY0z6CWpcQa9JDXOoJekxhn0ktQ4g16S\nGmfQS1LjDHpJapxBL0mNM+glqXEGvSQ1zqCXpMYZ9JLUuEUT9FNTK0myoF9TUyvHvZskNWjR/PDI\nZPy4gz/ssBhNxrEJHp8Lnz88IkmLlEEvSY0z6CWpcQa9JDXOoJekxhn0ktQ4g16SGmfQS1LjDHpJ\napxBL0mNM+glqXFLxt0ASUcAT3g8yQJ0xLgboCEZ9NLYPcykPNRMk8mhG0lqnEEvSY0z6CWpcQa9\nJDXOoJekxhn0ktS4oYM+yYok1ye5I8ntSd7alS9NsjXJnUmuTXLMtGXWJ9mRZHuS02fjA0iSDm7o\nHwdPMgVMVdWtSZ4O3AysBV4P/FdVvS/JxcDSqlqX5BTgCuCngRXAdcCz9/cr4P44uBaTyTg2weNz\n4Zv1Hwevql1VdWs3/RCwnUGArwU2ddU2AWd302cBV1bVnqq6G9gBrBl2+5KkQzMrY/RJVgIvAL4M\nLKuq3TD4MgCO7aotB+6dttjOrkySNIdGfgRCN2zzKeCiqnooyb7/txvq/3obN258bLrX69Hr9YZt\noiQ1qd/v0+/3Z6w39Bg9QJIlwN8D/1hVl3Zl24FeVe3uxvE/V1Wrk6wDqqou6epdA2yoqhv2s17H\n6LVoTMaxCR6fC9+sj9F3Pgps2xvynS3A+d30ecDV08rPTXJ4klXAScCNI25fkjSDUa66eSnwBeB2\nBt2RAt7GILw3A8cB9wDnVNW3umXWA28AHmUw1LP1AOu2R69FYzKOTfD4XPgO1KMfaehmrhj0Wkwm\n49gEj8+Fb66GbiRJC5xBL0mNM+glqXEGvSQ1zqCXpMYZ9JLUOINekhpn0EtS4wx6SWqcQS9JjTPo\nJalxBr0kNc6gl6TGGfSS1DiDXpIaZ9BLUuMMeklqnEEvSY0z6CWpcQa9JDXOoJekxhn0ktQ4g16S\nGrdk3A2YP0cAGXcjZnDEuBsgqUGLKOgfBmrcjZjBQv8ikjSJHLqRpMYZ9JLUOINeT9rU1EqSLPjX\n1NTKce8qjYHH5xOlauGNWyep2W5XEiZhjH4h/nvsazL2Jbg/Z5v7c3bN/v5MQlU94WSfPXpJapxB\nL0mNM+glqXEGvSQ1zqCXpMYZ9JLUuHkP+iSvSPK1JHcluXi+tz+7+uNuQEP6425AY/rjbkBj+uNu\nwEjmNeiTHAb8FXAG8FzgVUl+cj7bMLv6425AQ/rjbkBj+uNuQGP6427ASOb7oWZrgB1VdQ9AkiuB\ntcDX5rkdGslcPQn0HbO8Pp8GKsH8D90sB+6d9v6+rkwTZe+TQGfztWEO1vnwXO0ALWh7OyKz+XrH\nHKxz/joiC/YxxYPbmGd9rXOwztnthc7N554LC39fgvtztk3O/pwED8/b/pzvoN8JHD/t/Yqu7Afs\n71kNkqThzPfQzU3ASUlOSHI4cC6wZZ7bIEmLyrz26Kvq+0neAmxl8CXzkaraPp9tkKTFZkE+pliS\nNHu8M1aSGmfQS1LjFuzllQtNdwfvWh6/7n8nsMVzDFoIuuNzOXBDVT00rfwVVXXN+Fo2eZK8FHig\nqrYl+TngRcCtVfXZMTdtaPboD0H3TJ4rGVzsfGP3CvDJJOvG2bbWJHn9uNswaZK8Fbga+E3g35Ks\nnTb73eNp1WRK8m7g/cDHkrwPuAQ4CtiQ5HfH2rgReDL2ECS5C3huVT26T/nhwB1V9ezxtKw9Sb5R\nVcfPXFN7Jbkd+JmqeijJSuBTwF9X1aVJbqmqF461gRMkyR3AqQxuW90FrKiqbyd5GvDlqnr+WBs4\nJIduDs3/AT8O3LNP+bO6eXoSktx2oFnAsvlsSyMO2ztcU1V3J+kBn0pyAnNzy23LHqmq7wPfTfLv\nVfVtgKr6XpKJ/Vs36A/NbwGfTbKDx5/VczxwEvCWsbVqci1j8ATTB/YpD/Av89+cibc7yQuq6laA\nrmf/y8BHgeeNt2kT55EkR1XVd4Gf2luY5BgGD1CaSA7dHKLuEctr+MGTsTd13/56EpJ8BLi8qr64\nn3mfqKpXj6FZEyvJCmBPVe3az7yXVtU/j6FZEynJEVX1hKfhJflR4FlVdfsYmjUyg16SGudVN5LU\nOINekhpn0EtS4wx6LVpJNiT5nXG3Q5prBr0kNc6g16KR5HVJvprkliSbmHZddJI3Jrmxm3dVkiO7\n8lcmub0r73dlpyS5IclXktya5Ce68tdMK/9gBg5LcnmS27ptXzSOz67FzcsrtSgkOQX4WwaPCngg\nyQ8DFwHfqao/TbK0qh7o6v4xsKuqPtDdxXtGVX0zyTO62+H/AvhSVX0yyRLgKcAq4H3Ar3Y/sPMB\n4EvANuC9VXV6t+5n7L3bUpov9ui1WLwMuGpvmFfVt/aZf2qSL3TB/mrguV35F4FNSd7I43eSfwl4\ne5LfB1Z2N9i8HDgNuCnJLd32TgS+DqxKcmmSM4DvzN1HlPbPoJcGLgcurKpTgXcCRwJU1YXA24Hj\ngJu7nv8ngV8Bvgf8Q/dsmQCbquq0qnphVa2uqnd2XyjPB/rAbwAfnufPJRn0WjSuB16Z5EcAkizd\nZ/7TgV1Jngq8Zm9hkhOr6qaq2gDcDxyXZFVV/UdV/SWDH7c/Ffgs8GtJfmzv+pMcn+SZwFOq6tPA\nHwI+SVLzzoeaaVHofkTiXcDnk+wBbgHunlbljxj8zsD9wA3A0V35nyTZ+xjq66rqtiQXJ3kt8Cjw\nTeBdVfWtJH8AbO2ei/QI8Gbgf4HLu7IC/P0CzTtPxkpS4xy6kaTGGfSS1DiDXpIaZ9BLUuMMeklq\nnEEvSY0z6CWpcf8PPEvLWoMl+oQAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x125b89fd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_car_train, df_car_test = split_data(df_car_data, 0.8, random_state=20)\n",
    "\n",
    "plot_train_test_data(df_car_train, df_car_test, CAR_LABEL_COLUMN)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### def input functions\n",
    "\n",
    "Convert data to tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 346,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def car_train_input_fn():\n",
    "    return input_fn(df_car_train, CAR_CATEGORICAL_COLUMN, CAR_LABEL_COLUMN)\n",
    "\n",
    "def car_test_input_fn():\n",
    "    return input_fn(df_car_test, CAR_CATEGORICAL_COLUMN, CAR_LABEL_COLUMN)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create Feature Columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 347,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "car_feature_columns = create_feature_columns(CAR_CATEGORICAL_COLUMN)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 348,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Using default config.\n",
      "INFO:tensorflow:Using config: {'_save_checkpoints_secs': 600, '_num_ps_replicas': 0, '_keep_checkpoint_max': 5, '_tf_random_seed': None, '_task_type': None, '_environment': 'local', '_is_chief': True, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x123528a10>, '_tf_config': gpu_options {\n",
      "  per_process_gpu_memory_fraction: 1\n",
      "}\n",
      ", '_task_id': 0, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_evaluation_master': '', '_keep_checkpoint_every_n_hours': 10000, '_master': ''}\n"
     ]
    }
   ],
   "source": [
    "model_dir = \"/tmp/workspace/categorical-dnn/\"\n",
    "car_classifier = tf.contrib.learn.DNNClassifier(feature_columns=car_feature_columns,\n",
    "                                            hidden_units=[10, 20, 10],\n",
    "                                            n_classes=4,\n",
    "                                            model_dir=model_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 349,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:From /Users/xozhang/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/tensorflow/contrib/learn/python/learn/estimators/head.py:1362: scalar_summary (from tensorflow.python.ops.logging_ops) is deprecated and will be removed after 2016-11-30.\n",
      "Instructions for updating:\n",
      "Please switch to tf.summary.scalar. Note that tf.summary.scalar uses the node name instead of the tag. This means that TensorFlow will automatically de-duplicate summary names based on the scope they are created in. Also, passing a tensor or list of tags to a scalar summary op is no longer supported.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Saving checkpoints for 1 into /tmp/workspace/categorical-dnn/model.ckpt.\n",
      "INFO:tensorflow:loss = 1.55783, step = 1\n",
      "INFO:tensorflow:global_step/sec: 386.275\n",
      "INFO:tensorflow:loss = 0.684059, step = 101\n",
      "INFO:tensorflow:global_step/sec: 412.086\n",
      "INFO:tensorflow:loss = 0.62062, step = 201\n",
      "INFO:tensorflow:global_step/sec: 447.627\n",
      "INFO:tensorflow:loss = 0.579463, step = 301\n",
      "INFO:tensorflow:global_step/sec: 458.88\n",
      "INFO:tensorflow:loss = 0.511961, step = 401\n",
      "INFO:tensorflow:global_step/sec: 435.893\n",
      "INFO:tensorflow:loss = 0.426289, step = 501\n",
      "INFO:tensorflow:global_step/sec: 421.75\n",
      "INFO:tensorflow:loss = 0.369106, step = 601\n",
      "INFO:tensorflow:global_step/sec: 420.774\n",
      "INFO:tensorflow:loss = 0.325434, step = 701\n",
      "INFO:tensorflow:global_step/sec: 449.659\n",
      "INFO:tensorflow:loss = 0.293356, step = 801\n",
      "INFO:tensorflow:global_step/sec: 444.729\n",
      "INFO:tensorflow:loss = 0.260331, step = 901\n",
      "INFO:tensorflow:global_step/sec: 424.468\n",
      "INFO:tensorflow:loss = 0.235651, step = 1001\n",
      "INFO:tensorflow:global_step/sec: 456.798\n",
      "INFO:tensorflow:loss = 0.204259, step = 1101\n",
      "INFO:tensorflow:global_step/sec: 441.334\n",
      "INFO:tensorflow:loss = 0.183843, step = 1201\n",
      "INFO:tensorflow:global_step/sec: 432.229\n",
      "INFO:tensorflow:loss = 0.157291, step = 1301\n",
      "INFO:tensorflow:global_step/sec: 451.249\n",
      "INFO:tensorflow:loss = 0.144161, step = 1401\n",
      "INFO:tensorflow:global_step/sec: 485.991\n",
      "INFO:tensorflow:loss = 0.133065, step = 1501\n",
      "INFO:tensorflow:global_step/sec: 443.009\n",
      "INFO:tensorflow:loss = 0.123993, step = 1601\n",
      "INFO:tensorflow:global_step/sec: 445.569\n",
      "INFO:tensorflow:loss = 0.115847, step = 1701\n",
      "INFO:tensorflow:global_step/sec: 343.731\n",
      "INFO:tensorflow:loss = 0.110578, step = 1801\n",
      "INFO:tensorflow:global_step/sec: 409.611\n",
      "INFO:tensorflow:loss = 0.104435, step = 1901\n",
      "INFO:tensorflow:Saving checkpoints for 2000 into /tmp/workspace/categorical-dnn/model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 0.100043.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "DNNClassifier(params={'head': <tensorflow.contrib.learn.python.learn.estimators.head._MultiClassHead object at 0x1235288d0>, 'hidden_units': [10, 20, 10], 'feature_columns': (_RealValuedColumn(column_name='buying', dimension=1, default_value=None, dtype=tf.float32, normalizer=None), _RealValuedColumn(column_name='maint', dimension=1, default_value=None, dtype=tf.float32, normalizer=None), _RealValuedColumn(column_name='doors', dimension=1, default_value=None, dtype=tf.float32, normalizer=None), _RealValuedColumn(column_name='persons', dimension=1, default_value=None, dtype=tf.float32, normalizer=None), _RealValuedColumn(column_name='lug_boot', dimension=1, default_value=None, dtype=tf.float32, normalizer=None), _RealValuedColumn(column_name='safety', dimension=1, default_value=None, dtype=tf.float32, normalizer=None)), 'embedding_lr_multipliers': None, 'optimizer': None, 'dropout': None, 'gradient_clip_norm': None, 'activation_fn': <function relu at 0x10b9199b0>, 'input_layer_min_slice_size': None})"
      ]
     },
     "execution_count": 349,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "car_classifier.fit(input_fn=car_train_input_fn, steps = 2000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluate the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 351,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:Rank of input Tensor (1) should be the same as output_rank (2) for column. Will attempt to expand dims. It is highly recommended that you resize your input, as this behavior may change.\n",
      "WARNING:tensorflow:From /Users/xozhang/Library/Enthought/Canopy_64bit/User/lib/python2.7/site-packages/tensorflow/contrib/learn/python/learn/estimators/head.py:1362: scalar_summary (from tensorflow.python.ops.logging_ops) is deprecated and will be removed after 2016-11-30.\n",
      "Instructions for updating:\n",
      "Please switch to tf.summary.scalar. Note that tf.summary.scalar uses the node name instead of the tag. This means that TensorFlow will automatically de-duplicate summary names based on the scope they are created in. Also, passing a tensor or list of tags to a scalar summary op is no longer supported.\n",
      "INFO:tensorflow:Starting evaluation at 2017-03-28-07:59:42\n",
      "INFO:tensorflow:Evaluation [1/1]\n",
      "INFO:tensorflow:Finished evaluation at 2017-03-28-07:59:43\n",
      "INFO:tensorflow:Saving dict for global step 2000: accuracy = 0.933526, auc = 0.991921, global_step = 2000, loss = 0.178961\n",
      "WARNING:tensorflow:Skipping summary for global_step, must be a float or np.float32.\n"
     ]
    }
   ],
   "source": [
    "car_results = car_classifier.evaluate(input_fn=car_test_input_fn, steps=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 353,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'accuracy': 0.93352604,\n",
       " 'auc': 0.99192089,\n",
       " 'global_step': 2000,\n",
       " 'loss': 0.17896064}"
      ]
     },
     "execution_count": 353,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "car_results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}