04. LinearRegressor and DNNClassifier (Estimator API).ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.datasets import make_regression\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Number_pregnant</th>\n",
       "      <th>Glucose_concentration</th>\n",
       "      <th>Blood_pressure</th>\n",
       "      <th>Triceps</th>\n",
       "      <th>Insulin</th>\n",
       "      <th>BMI</th>\n",
       "      <th>Pedigree</th>\n",
       "      <th>Age</th>\n",
       "      <th>Class</th>\n",
       "      <th>Group</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6</td>\n",
       "      <td>0.743719</td>\n",
       "      <td>0.590164</td>\n",
       "      <td>0.353535</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.500745</td>\n",
       "      <td>0.234415</td>\n",
       "      <td>50</td>\n",
       "      <td>1</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>0.427136</td>\n",
       "      <td>0.540984</td>\n",
       "      <td>0.292929</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.396423</td>\n",
       "      <td>0.116567</td>\n",
       "      <td>31</td>\n",
       "      <td>0</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>8</td>\n",
       "      <td>0.919598</td>\n",
       "      <td>0.524590</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.347243</td>\n",
       "      <td>0.253629</td>\n",
       "      <td>32</td>\n",
       "      <td>1</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>0.447236</td>\n",
       "      <td>0.540984</td>\n",
       "      <td>0.232323</td>\n",
       "      <td>0.111111</td>\n",
       "      <td>0.418778</td>\n",
       "      <td>0.038002</td>\n",
       "      <td>21</td>\n",
       "      <td>0</td>\n",
       "      <td>B</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0.688442</td>\n",
       "      <td>0.327869</td>\n",
       "      <td>0.353535</td>\n",
       "      <td>0.198582</td>\n",
       "      <td>0.642325</td>\n",
       "      <td>0.943638</td>\n",
       "      <td>33</td>\n",
       "      <td>1</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Number_pregnant  Glucose_concentration  Blood_pressure   Triceps   Insulin  \\\n",
       "0                6               0.743719        0.590164  0.353535  0.000000   \n",
       "1                1               0.427136        0.540984  0.292929  0.000000   \n",
       "2                8               0.919598        0.524590  0.000000  0.000000   \n",
       "3                1               0.447236        0.540984  0.232323  0.111111   \n",
       "4                0               0.688442        0.327869  0.353535  0.198582   \n",
       "\n",
       "        BMI  Pedigree  Age  Class Group  \n",
       "0  0.500745  0.234415   50      1     B  \n",
       "1  0.396423  0.116567   31      0     C  \n",
       "2  0.347243  0.253629   32      1     B  \n",
       "3  0.418778  0.038002   21      0     B  \n",
       "4  0.642325  0.943638   33      1     C  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "diabates = pd.read_csv(\"datasets/pima-indians-diabetes.csv\")\n",
    "diabates.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Normalize numerical columns\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Number_pregnant', 'Glucose_concentration', 'Blood_pressure', 'Triceps',\n",
       "       'Insulin', 'BMI', 'Pedigree', 'Age', 'Class', 'Group'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "diabates.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "numerical_cols = ['Number_pregnant', 'Glucose_concentration', 'Blood_pressure', 'Triceps',\n",
    "                  'Insulin', 'BMI', 'Pedigree']\n",
    "diabates[numerical_cols] = diabates[numerical_cols].apply(lambda x: (x - x.min()) / (x.max() - x.min()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Number_pregnant</th>\n",
       "      <th>Glucose_concentration</th>\n",
       "      <th>Blood_pressure</th>\n",
       "      <th>Triceps</th>\n",
       "      <th>Insulin</th>\n",
       "      <th>BMI</th>\n",
       "      <th>Pedigree</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.352941</td>\n",
       "      <td>0.743719</td>\n",
       "      <td>0.590164</td>\n",
       "      <td>0.353535</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.500745</td>\n",
       "      <td>0.234415</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.058824</td>\n",
       "      <td>0.427136</td>\n",
       "      <td>0.540984</td>\n",
       "      <td>0.292929</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.396423</td>\n",
       "      <td>0.116567</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.470588</td>\n",
       "      <td>0.919598</td>\n",
       "      <td>0.524590</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.347243</td>\n",
       "      <td>0.253629</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.058824</td>\n",
       "      <td>0.447236</td>\n",
       "      <td>0.540984</td>\n",
       "      <td>0.232323</td>\n",
       "      <td>0.111111</td>\n",
       "      <td>0.418778</td>\n",
       "      <td>0.038002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.688442</td>\n",
       "      <td>0.327869</td>\n",
       "      <td>0.353535</td>\n",
       "      <td>0.198582</td>\n",
       "      <td>0.642325</td>\n",
       "      <td>0.943638</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Number_pregnant  Glucose_concentration  Blood_pressure   Triceps   Insulin  \\\n",
       "0         0.352941               0.743719        0.590164  0.353535  0.000000   \n",
       "1         0.058824               0.427136        0.540984  0.292929  0.000000   \n",
       "2         0.470588               0.919598        0.524590  0.000000  0.000000   \n",
       "3         0.058824               0.447236        0.540984  0.232323  0.111111   \n",
       "4         0.000000               0.688442        0.327869  0.353535  0.198582   \n",
       "\n",
       "        BMI  Pedigree  \n",
       "0  0.500745  0.234415  \n",
       "1  0.396423  0.116567  \n",
       "2  0.347243  0.253629  \n",
       "3  0.418778  0.038002  \n",
       "4  0.642325  0.943638  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# observe all columns are normalized\n",
    "diabates[numerical_cols].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Define list of feature columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_preg = tf.feature_column.numeric_column('Number_pregnant')\n",
    "gluc_conc = tf.feature_column.numeric_column('Glucose_concentration')\n",
    "bp = tf.feature_column.numeric_column('Blood_pressure')\n",
    "triceps = tf.feature_column.numeric_column('Triceps')\n",
    "insulin = tf.feature_column.numeric_column('Insulin')\n",
    "bmi = tf.feature_column.numeric_column('BMI')\n",
    "pedigree = tf.feature_column.numeric_column('Pedigree')\n",
    "age = tf.feature_column.numeric_column('Age')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dealing with categorical variables\n",
    "There are 2 methods used to deal with categorical variables.\n",
    "1. Using Hash Bucket: This is similar to using scikit learn's FeatureHasher. Use this when you don't know the exact values of the features that will occur. Input is just the number of buckets to hash on. This can be lesser than the number of features. \n",
    "\n",
    "2. Using Vocabulary List: This is like using scikit learn's DictVectorizer. Use this when you know the features that are going to be included like \"A\", \"B\", \"C\", \"D\"."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Using vocabulary list. \n",
    "assigned_group = tf.feature_column.categorical_column_with_vocabulary_list(\"Group\", [\"A\", \"B\", \"C\", \"D\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f4c53e5fe48>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f4c53e5f780>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Histogram of ages\n",
    "diabates[\"Age\"].hist(bins=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "age_buckets = tf.feature_column.bucketized_column(age, boundaries=[20, 30, 40, 50, 60, 70, 80])\n",
    "feat_cols = [num_preg, gluc_conc, bp, triceps, insulin, bmi, pedigree, age, age_buckets]\n",
    "y = diabates[\"Class\"]\n",
    "X = diabates.drop(\"Class\", axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prepare dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, stratify=y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_function = tf.estimator.inputs.pandas_input_fn(X, y, batch_size=8, shuffle=False, num_epochs=1000)\n",
    "\n",
    "test_function = tf.estimator.inputs.pandas_input_fn(X_test, y_test, batch_size=8, shuffle=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Prepare Estimator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Using default config.\n",
      "WARNING:tensorflow:Using temporary folder as model directory: /tmp/tmpb20wcwlo\n",
      "INFO:tensorflow:Using config: {'_model_dir': '/tmp/tmpb20wcwlo', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': None, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_service': None, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7f4c534f1cf8>, '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}\n"
     ]
    }
   ],
   "source": [
    "estimator = tf.estimator.LinearClassifier(feature_columns=feat_cols, n_classes=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 1 into /tmp/tmpb20wcwlo/model.ckpt.\n",
      "INFO:tensorflow:loss = 5.5451775, step = 1\n",
      "INFO:tensorflow:global_step/sec: 109.615\n",
      "INFO:tensorflow:loss = 5.1932898, step = 101 (0.916 sec)\n",
      "INFO:tensorflow:global_step/sec: 146.904\n",
      "INFO:tensorflow:loss = 5.3029995, step = 201 (0.680 sec)\n",
      "INFO:tensorflow:global_step/sec: 149.742\n",
      "INFO:tensorflow:loss = 2.5617986, step = 301 (0.670 sec)\n",
      "INFO:tensorflow:global_step/sec: 148.075\n",
      "INFO:tensorflow:loss = 4.5560436, step = 401 (0.676 sec)\n",
      "INFO:tensorflow:global_step/sec: 153.391\n",
      "INFO:tensorflow:loss = 3.9424653, step = 501 (0.655 sec)\n",
      "INFO:tensorflow:global_step/sec: 145.7\n",
      "INFO:tensorflow:loss = 6.9965057, step = 601 (0.682 sec)\n",
      "INFO:tensorflow:global_step/sec: 151.474\n",
      "INFO:tensorflow:loss = 6.1180086, step = 701 (0.663 sec)\n",
      "INFO:tensorflow:global_step/sec: 152.176\n",
      "INFO:tensorflow:loss = 4.4505224, step = 801 (0.654 sec)\n",
      "INFO:tensorflow:global_step/sec: 100.547\n",
      "INFO:tensorflow:loss = 4.599818, step = 901 (0.994 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /tmp/tmpb20wcwlo/model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 4.8054733.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.estimator.canned.linear.LinearClassifier at 0x7f4c534f1f28>"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "estimator.train(input_fn=train_function, steps=1000)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2018-03-10-03:43:05\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /tmp/tmpb20wcwlo/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Finished evaluation at 2018-03-10-03:43:07\n",
      "INFO:tensorflow:Saving dict for global step 1000: accuracy = 0.72727275, accuracy_baseline = 0.64935064, auc = 0.8148972, auc_precision_recall = 0.68888855, average_loss = 0.5373519, global_step = 1000, label/mean = 0.35064936, loss = 4.280286, prediction/mean = 0.45529935\n"
     ]
    }
   ],
   "source": [
    "res = estimator.evaluate(input_fn=test_function, steps=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'accuracy': 0.72727275,\n",
       " 'accuracy_baseline': 0.64935064,\n",
       " 'auc': 0.8148972,\n",
       " 'auc_precision_recall': 0.68888855,\n",
       " 'average_loss': 0.5373519,\n",
       " 'global_step': 1000,\n",
       " 'label/mean': 0.35064936,\n",
       " 'loss': 4.280286,\n",
       " 'prediction/mean': 0.45529935}"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = estimator.predict(input_fn=test_function)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /tmp/tmpb20wcwlo/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[{'class_ids': array([1]),\n",
       "  'classes': array([b'1'], dtype=object),\n",
       "  'logistic': array([0.5464438], dtype=float32),\n",
       "  'logits': array([0.18631226], dtype=float32),\n",
       "  'probabilities': array([0.4535562, 0.5464438], dtype=float32)},\n",
       " {'class_ids': array([1]),\n",
       "  'classes': array([b'1'], dtype=object),\n",
       "  'logistic': array([0.56642777], dtype=float32),\n",
       "  'logits': array([0.26729125], dtype=float32),\n",
       "  'probabilities': array([0.4335722 , 0.56642777], dtype=float32)},\n",
       " {'class_ids': array([0]),\n",
       "  'classes': array([b'0'], dtype=object),\n",
       "  'logistic': array([0.19620767], dtype=float32),\n",
       "  'logits': array([-1.4101672], dtype=float32),\n",
       "  'probabilities': array([0.8037923 , 0.19620767], dtype=float32)},\n",
       " {'class_ids': array([0]),\n",
       "  'classes': array([b'0'], dtype=object),\n",
       "  'logistic': array([0.20084667], dtype=float32),\n",
       "  'logits': array([-1.381011], dtype=float32),\n",
       "  'probabilities': array([0.7991533 , 0.20084667], dtype=float32)},\n",
       " {'class_ids': array([0]),\n",
       "  'classes': array([b'0'], dtype=object),\n",
       "  'logistic': array([0.37437505], dtype=float32),\n",
       "  'logits': array([-0.51349306], dtype=float32),\n",
       "  'probabilities': array([0.62562495, 0.37437502], dtype=float32)}]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(predictions)[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Deep Neural Net Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Using default config.\n",
      "WARNING:tensorflow:Using temporary folder as model directory: /tmp/tmpgbmrum6x\n",
      "INFO:tensorflow:Using config: {'_model_dir': '/tmp/tmpgbmrum6x', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': None, '_save_checkpoints_secs': 600, '_session_config': None, '_keep_checkpoint_max': 5, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': 100, '_service': None, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7f4c53bf2c88>, '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1}\n"
     ]
    }
   ],
   "source": [
    "dnn_estimator = tf.estimator.DNNClassifier(hidden_units=[30, 30, 30, 30], feature_columns=feat_cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Create CheckpointSaverHook.\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Saving checkpoints for 1 into /tmp/tmpgbmrum6x/model.ckpt.\n",
      "INFO:tensorflow:loss = 9.709453, step = 1\n",
      "INFO:tensorflow:global_step/sec: 113.901\n",
      "INFO:tensorflow:loss = 5.387681, step = 101 (0.880 sec)\n",
      "INFO:tensorflow:global_step/sec: 154.455\n",
      "INFO:tensorflow:loss = 5.193118, step = 201 (0.650 sec)\n",
      "INFO:tensorflow:global_step/sec: 150.255\n",
      "INFO:tensorflow:loss = 3.3515172, step = 301 (0.668 sec)\n",
      "INFO:tensorflow:global_step/sec: 151.082\n",
      "INFO:tensorflow:loss = 5.0420866, step = 401 (0.657 sec)\n",
      "INFO:tensorflow:global_step/sec: 155.852\n",
      "INFO:tensorflow:loss = 3.9946187, step = 501 (0.645 sec)\n",
      "INFO:tensorflow:global_step/sec: 144.559\n",
      "INFO:tensorflow:loss = 8.275373, step = 601 (0.692 sec)\n",
      "INFO:tensorflow:global_step/sec: 114.946\n",
      "INFO:tensorflow:loss = 4.980707, step = 701 (0.866 sec)\n",
      "INFO:tensorflow:global_step/sec: 151.544\n",
      "INFO:tensorflow:loss = 4.9984884, step = 801 (0.664 sec)\n",
      "INFO:tensorflow:global_step/sec: 148.366\n",
      "INFO:tensorflow:loss = 5.372033, step = 901 (0.681 sec)\n",
      "INFO:tensorflow:Saving checkpoints for 1000 into /tmp/tmpgbmrum6x/model.ckpt.\n",
      "INFO:tensorflow:Loss for final step: 5.1404276.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.estimator.canned.dnn.DNNClassifier at 0x7f4c53bf2d68>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dnn_estimator.train(input_fn=train_function, steps=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Calling model_fn.\n",
      "INFO:tensorflow:Done calling model_fn.\n",
      "INFO:tensorflow:Starting evaluation at 2018-03-10-03:43:42\n",
      "INFO:tensorflow:Graph was finalized.\n",
      "INFO:tensorflow:Restoring parameters from /tmp/tmpgbmrum6x/model.ckpt-1000\n",
      "INFO:tensorflow:Running local_init_op.\n",
      "INFO:tensorflow:Done running local_init_op.\n",
      "INFO:tensorflow:Finished evaluation at 2018-03-10-03:43:43\n",
      "INFO:tensorflow:Saving dict for global step 1000: accuracy = 0.6839827, accuracy_baseline = 0.64935064, auc = 0.7399177, auc_precision_recall = 0.51408166, average_loss = 0.5993931, global_step = 1000, label/mean = 0.35064936, loss = 4.774476, prediction/mean = 0.43413964\n"
     ]
    }
   ],
   "source": [
    "res = dnn_estimator.evaluate(input_fn=test_function, steps=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'accuracy': 0.6839827,\n",
       " 'accuracy_baseline': 0.64935064,\n",
       " 'auc': 0.7399177,\n",
       " 'auc_precision_recall': 0.51408166,\n",
       " 'average_loss': 0.5993931,\n",
       " 'global_step': 1000,\n",
       " 'label/mean': 0.35064936,\n",
       " 'loss': 4.774476,\n",
       " 'prediction/mean': 0.43413964}"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "res"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}