keras.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Classification using Keras \n",
    "Note: Keras uses TensorFlow by default. You can switch backend to Theano (see https://keras.io/backend/ for details)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1034,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.0.4'"
      ]
     },
     "execution_count": 1034,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import keras \n",
    "keras.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1035,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3 classes\n",
      "4 dims\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd \n",
    "import numpy as np\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import accuracy_score, precision_score, log_loss, classification_report\n",
    "from sklearn.metrics import confusion_matrix , mean_squared_error, r2_score\n",
    "from sklearn.datasets import load_iris, load_boston\n",
    "\n",
    "# Import `Sequential` from `keras.models`\n",
    "from keras.models import Sequential\n",
    "# Import `Dense` from `keras.layers`\n",
    "from keras.layers import Dense, Activation, Dropout\n",
    "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from keras.optimizers import SGD, RMSprop, Adam\n",
    "from keras.regularizers import l1, l2\n",
    "\n",
    "np.random.seed = 42   # for reproduction\n",
    "iris = load_iris()\n",
    "X = iris.data\n",
    "y = iris.target \n",
    "X = X.astype('float32')\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, stratify = y, train_size = 0.7, random_state=707)\n",
    "\n",
    "nb_classes = len(set(y_train))\n",
    "print(nb_classes, 'classes')\n",
    "\n",
    "dims = X_train.shape[1]\n",
    "print(dims, 'dims')#\n",
    "\n",
    "y_test_c = y_test\n",
    "# Convert labels to categorical one-hot encoding (there is also a sklear wrapper)\n",
    "y_train = keras.utils.to_categorical(y_train, num_classes=3)\n",
    "y_test = keras.utils.to_categorical(y_test, num_classes=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1036,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_497 (Dense)            (None, 10)                50        \n",
      "_________________________________________________________________\n",
      "dense_498 (Dense)            (None, 3)                 33        \n",
      "=================================================================\n",
      "Total params: 83\n",
      "Trainable params: 83\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "#### https://keras.io/getting-started/sequential-model-guide/\n",
    "sgdi = SGD(lr=0.01, decay=1e-6, momentum=0.9, nesterov=True)\n",
    "model = Sequential()\n",
    "model.add(Dense(10, activation='relu', input_dim=dims))\n",
    "model.add(Dense(nb_classes, activation='softmax'))\n",
    "model.compile(optimizer=sgdi, loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1037,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 84 samples, validate on 21 samples\n",
      "Epoch 1/50\n",
      "84/84 [==============================] - 2s - loss: 1.3085 - acc: 0.3333 - val_loss: 1.3342 - val_acc: 0.2381\n",
      "Epoch 2/50\n",
      "84/84 [==============================] - 0s - loss: 0.9324 - acc: 0.6548 - val_loss: 0.8086 - val_acc: 0.7619\n",
      "Epoch 3/50\n",
      "84/84 [==============================] - 0s - loss: 0.7091 - acc: 0.8214 - val_loss: 0.8038 - val_acc: 0.5238\n",
      "Epoch 4/50\n",
      "84/84 [==============================] - 0s - loss: 0.5683 - acc: 0.7024 - val_loss: 0.6324 - val_acc: 0.5714\n",
      "Epoch 5/50\n",
      "84/84 [==============================] - 0s - loss: 0.4720 - acc: 0.7262 - val_loss: 0.5430 - val_acc: 0.7619\n",
      "Epoch 6/50\n",
      "84/84 [==============================] - 0s - loss: 0.4136 - acc: 0.8690 - val_loss: 0.4834 - val_acc: 0.8095\n",
      "Epoch 7/50\n",
      "84/84 [==============================] - 0s - loss: 0.3940 - acc: 0.8690 - val_loss: 0.4533 - val_acc: 0.8095\n",
      "Epoch 8/50\n",
      "84/84 [==============================] - 0s - loss: 0.3469 - acc: 0.9048 - val_loss: 0.3967 - val_acc: 0.9048\n",
      "Epoch 9/50\n",
      "84/84 [==============================] - 0s - loss: 0.3355 - acc: 0.8690 - val_loss: 0.3460 - val_acc: 0.9048\n",
      "Epoch 10/50\n",
      "84/84 [==============================] - 0s - loss: 0.3031 - acc: 0.9286 - val_loss: 0.5044 - val_acc: 0.7143\n",
      "Epoch 11/50\n",
      "84/84 [==============================] - 0s - loss: 0.3104 - acc: 0.8571 - val_loss: 0.3188 - val_acc: 0.8571\n",
      "Epoch 12/50\n",
      "84/84 [==============================] - 0s - loss: 0.3036 - acc: 0.8810 - val_loss: 0.3812 - val_acc: 0.8571\n",
      "Epoch 13/50\n",
      "84/84 [==============================] - 0s - loss: 0.2697 - acc: 0.9167 - val_loss: 0.3213 - val_acc: 0.9048\n",
      "Epoch 14/50\n",
      "84/84 [==============================] - 0s - loss: 0.2445 - acc: 0.9286 - val_loss: 0.2655 - val_acc: 0.9524\n",
      "Epoch 15/50\n",
      "84/84 [==============================] - 0s - loss: 0.2309 - acc: 0.9524 - val_loss: 0.2610 - val_acc: 0.9524\n",
      "Epoch 16/50\n",
      "84/84 [==============================] - 0s - loss: 0.2146 - acc: 0.9762 - val_loss: 0.2375 - val_acc: 0.9524\n",
      "Epoch 17/50\n",
      "84/84 [==============================] - 0s - loss: 0.2070 - acc: 0.9643 - val_loss: 0.2394 - val_acc: 0.9524\n",
      "Epoch 18/50\n",
      "84/84 [==============================] - 0s - loss: 0.2042 - acc: 0.9405 - val_loss: 0.2788 - val_acc: 0.9048\n",
      "Epoch 19/50\n",
      "84/84 [==============================] - 0s - loss: 0.1982 - acc: 0.9643 - val_loss: 0.2087 - val_acc: 0.9048\n",
      "Epoch 20/50\n",
      "84/84 [==============================] - 0s - loss: 0.1863 - acc: 0.9286 - val_loss: 0.2645 - val_acc: 0.9048\n",
      "Epoch 21/50\n",
      "84/84 [==============================] - 0s - loss: 0.1910 - acc: 0.9286 - val_loss: 0.2437 - val_acc: 0.9048\n",
      "Epoch 22/50\n",
      "84/84 [==============================] - 0s - loss: 0.1811 - acc: 0.9405 - val_loss: 0.1911 - val_acc: 0.9048\n",
      "Epoch 23/50\n",
      "84/84 [==============================] - 0s - loss: 0.1754 - acc: 0.9524 - val_loss: 0.1815 - val_acc: 0.9048\n",
      "Epoch 24/50\n",
      "84/84 [==============================] - 0s - loss: 0.1664 - acc: 0.9643 - val_loss: 0.2192 - val_acc: 0.9524\n",
      "Epoch 25/50\n",
      "84/84 [==============================] - 0s - loss: 0.1702 - acc: 0.9762 - val_loss: 0.1714 - val_acc: 0.9524\n",
      "Epoch 26/50\n",
      "84/84 [==============================] - 0s - loss: 0.1569 - acc: 0.9524 - val_loss: 0.2015 - val_acc: 0.9524\n",
      "Epoch 27/50\n",
      "84/84 [==============================] - 0s - loss: 0.1556 - acc: 0.9643 - val_loss: 0.1783 - val_acc: 0.9524\n",
      "Epoch 28/50\n",
      "84/84 [==============================] - 0s - loss: 0.1464 - acc: 0.9643 - val_loss: 0.1614 - val_acc: 0.9048\n",
      "Epoch 29/50\n",
      "84/84 [==============================] - 0s - loss: 0.1558 - acc: 0.9524 - val_loss: 0.1579 - val_acc: 0.9524\n",
      "Epoch 30/50\n",
      "84/84 [==============================] - 0s - loss: 0.1573 - acc: 0.9524 - val_loss: 0.1957 - val_acc: 0.9524\n",
      "Epoch 31/50\n",
      "84/84 [==============================] - 0s - loss: 0.1477 - acc: 0.9405 - val_loss: 0.1675 - val_acc: 0.9524\n",
      "Epoch 32/50\n",
      "84/84 [==============================] - 0s - loss: 0.1347 - acc: 0.9643 - val_loss: 0.1554 - val_acc: 0.9048\n",
      "Epoch 33/50\n",
      "84/84 [==============================] - 0s - loss: 0.1575 - acc: 0.9405 - val_loss: 0.1532 - val_acc: 0.9048\n",
      "Epoch 34/50\n",
      "84/84 [==============================] - 0s - loss: 0.1413 - acc: 0.9524 - val_loss: 0.1997 - val_acc: 0.9048\n",
      "Epoch 35/50\n",
      "84/84 [==============================] - 0s - loss: 0.1536 - acc: 0.9643 - val_loss: 0.1798 - val_acc: 0.9524\n",
      "Epoch 36/50\n",
      "84/84 [==============================] - 0s - loss: 0.1378 - acc: 0.9762 - val_loss: 0.1430 - val_acc: 0.9048\n",
      "Epoch 37/50\n",
      "84/84 [==============================] - 0s - loss: 0.1277 - acc: 0.9643 - val_loss: 0.2061 - val_acc: 0.9048\n",
      "Epoch 38/50\n",
      "84/84 [==============================] - 0s - loss: 0.1692 - acc: 0.9048 - val_loss: 0.1719 - val_acc: 0.9524\n",
      "Epoch 39/50\n",
      "84/84 [==============================] - 0s - loss: 0.1325 - acc: 0.9524 - val_loss: 0.1789 - val_acc: 0.9524\n",
      "Epoch 40/50\n",
      "84/84 [==============================] - 0s - loss: 0.1377 - acc: 0.9643 - val_loss: 0.1438 - val_acc: 0.9524\n",
      "Epoch 41/50\n",
      "84/84 [==============================] - 0s - loss: 0.1280 - acc: 0.9643 - val_loss: 0.2449 - val_acc: 0.9048\n",
      "Epoch 42/50\n",
      "84/84 [==============================] - 0s - loss: 0.1421 - acc: 0.9167 - val_loss: 0.1430 - val_acc: 0.9524\n",
      "Epoch 43/50\n",
      "84/84 [==============================] - 0s - loss: 0.1207 - acc: 0.9524 - val_loss: 0.1494 - val_acc: 0.9524\n",
      "Epoch 44/50\n",
      "84/84 [==============================] - 0s - loss: 0.1314 - acc: 0.9643 - val_loss: 0.1339 - val_acc: 0.9524\n",
      "Epoch 45/50\n",
      "84/84 [==============================] - 0s - loss: 0.1259 - acc: 0.9643 - val_loss: 0.1345 - val_acc: 0.9048\n",
      "Epoch 46/50\n",
      "84/84 [==============================] - 0s - loss: 0.1698 - acc: 0.9167 - val_loss: 0.1294 - val_acc: 0.9524\n",
      "Epoch 47/50\n",
      "84/84 [==============================] - 0s - loss: 0.1147 - acc: 0.9643 - val_loss: 0.1289 - val_acc: 0.9524\n",
      "Epoch 48/50\n",
      "84/84 [==============================] - 0s - loss: 0.1168 - acc: 0.9762 - val_loss: 0.1511 - val_acc: 0.9524\n",
      "Epoch 49/50\n",
      "84/84 [==============================] - 0s - loss: 0.1106 - acc: 0.9762 - val_loss: 0.1515 - val_acc: 0.9048\n",
      "Epoch 50/50\n",
      "84/84 [==============================] - 0s - loss: 0.1279 - acc: 0.9762 - val_loss: 0.1269 - val_acc: 0.9524\n"
     ]
    }
   ],
   "source": [
    "early_stop = EarlyStopping(monitor='val_loss', patience=5, verbose=1) \n",
    "Result = model.fit(X_train, y_train, validation_split =0.2, epochs=50, # validation_data = (X_test, y_test)\n",
    "          batch_size=15, verbose=True, callbacks=[early_stop]) #callbacks=[best_model, early_stop]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1038,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1c7adf9b080>"
      ]
     },
     "execution_count": 1038,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFJRSSjX6/2UwkFJD01onAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1c7aaed14e0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBatkAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1c7ae04dfd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "plt.figure()\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.plot(Result.history['loss'])\n",
    "plt.plot(Result.history['val_loss'])\n",
    "plt.legend(['Training', 'Validation'])\n",
    "\n",
    "plt.figure()\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.plot(Result.history['acc'])\n",
    "plt.plot(Result.history['val_acc'])\n",
    "plt.legend(['Training', 'Validation'], loc='lower right')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1039,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "32/45 [====================>.........] - ETA: 0s\n",
      " score:  [0.051197536951965758, 0.97777777777777775]\n"
     ]
    }
   ],
   "source": [
    "y_pred = model.predict(X_test)\n",
    "score = model.evaluate(X_test, y_test,verbose=1)\n",
    "print(\"\\n score: \",score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1040,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\r",
      "32/45 [====================>.........] - ETA: 0s"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[15,  0,  0],\n",
       "       [ 0, 14,  1],\n",
       "       [ 0,  0, 15]])"
      ]
     },
     "execution_count": 1040,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_c = model.predict_classes(X_test)\n",
    "# Confusion matrix\n",
    "confusion_matrix(y_test_c, y_pred_c)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1041,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.051197537285689682"
      ]
     },
     "execution_count": 1041,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Precision \n",
    "#precision_score(y_test, y_pred)\n",
    "log_loss(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1042,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([2, 1, 2, 0, 0, 1, 2, 0, 1, 2], dtype=int64)"
      ]
     },
     "execution_count": 1042,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred_c[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1043,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  4.29935562e-06,   1.22276517e-02,   9.87767994e-01],\n",
       "       [  5.33890584e-03,   9.85195398e-01,   9.46568418e-03],\n",
       "       [  4.50710431e-05,   1.54280931e-01,   8.45673978e-01],\n",
       "       [  9.96310174e-01,   3.68983857e-03,   3.73962528e-10],\n",
       "       [  9.97425973e-01,   2.57403450e-03,   4.58050924e-11],\n",
       "       [  2.96956971e-02,   9.63809311e-01,   6.49506226e-03],\n",
       "       [  2.19391595e-07,   4.52001439e-03,   9.95479703e-01],\n",
       "       [  9.94876564e-01,   5.12344716e-03,   2.60549582e-10],\n",
       "       [  5.38422726e-03,   9.78536487e-01,   1.60792395e-02],\n",
       "       [  1.41054825e-05,   1.25773221e-01,   8.74212623e-01]], dtype=float32)"
      ]
     },
     "execution_count": 1043,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_pred[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Regression example in Keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1044,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_boston \n",
    "from sklearn.preprocessing import StandardScaler, MinMaxScaler\n",
    "from sklearn.utils import shuffle\n",
    "from sklearn.metrics import r2_score, mean_squared_error\n",
    "boston = load_boston()\n",
    "Xb, yb = shuffle(boston.data, boston.target, random_state=13)\n",
    "Xb_train, Xb_test, yb_train, yb_test = train_test_split(Xb, yb, train_size=0.751, random_state=717)\n",
    "scalerS = StandardScaler()\n",
    "scalerM = MinMaxScaler()\n",
    "scalerS.fit(Xb_train)\n",
    "Xb_train = scalerS.transform(Xb_train)\n",
    "Xb_test = scalerS.transform(Xb_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1045,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_499 (Dense)            (None, 52)                728       \n",
      "_________________________________________________________________\n",
      "dense_500 (Dense)            (None, 26)                1378      \n",
      "_________________________________________________________________\n",
      "dense_501 (Dense)            (None, 1)                 27        \n",
      "=================================================================\n",
      "Total params: 2,133\n",
      "Trainable params: 2,133\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "## https://keras.io/initializers/\n",
    "modelb = Sequential()\n",
    "modelb.add(Dense(52, input_dim=13, activation='relu', kernel_regularizer=l2(0.01),\n",
    "                activity_regularizer=l1(0.01),\n",
    "        kernel_initializer='random_normal', bias_initializer='random_normal')) # try w/out kernel & bias\n",
    "modelb.add(Dense(26, activation='relu'))\n",
    "modelb.add(Dense(1, activation='linear'))\n",
    "# Compile model\n",
    "#adam = Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)\n",
    "sgd = SGD(lr=0.00125, decay=1e-6, momentum=0.9, nesterov=True)\n",
    "modelb.compile(loss='mean_squared_error', optimizer=sgd)  # rmsprop\n",
    "modelb.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1046,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 380 samples, validate on 126 samples\n",
      "Epoch 1/200\n",
      "380/380 [==============================] - 2s - loss: 565.2475 - val_loss: 577.1082\n",
      "Epoch 2/200\n",
      "380/380 [==============================] - 0s - loss: 475.4054 - val_loss: 321.2440\n",
      "Epoch 3/200\n",
      "380/380 [==============================] - 0s - loss: 200.2006 - val_loss: 100.9111\n",
      "Epoch 4/200\n",
      "380/380 [==============================] - 0s - loss: 77.3871 - val_loss: 45.5275\n",
      "Epoch 5/200\n",
      "380/380 [==============================] - 0s - loss: 48.5056 - val_loss: 46.5390\n",
      "Epoch 6/200\n",
      "380/380 [==============================] - 0s - loss: 44.1652 - val_loss: 41.4886\n",
      "Epoch 7/200\n",
      "380/380 [==============================] - 0s - loss: 41.9866 - val_loss: 37.6308\n",
      "Epoch 8/200\n",
      "380/380 [==============================] - 0s - loss: 38.7698 - val_loss: 32.7485\n",
      "Epoch 9/200\n",
      "380/380 [==============================] - 0s - loss: 35.1914 - val_loss: 30.9177\n",
      "Epoch 10/200\n",
      "380/380 [==============================] - 0s - loss: 33.3120 - val_loss: 29.5645\n",
      "Epoch 11/200\n",
      "380/380 [==============================] - 0s - loss: 31.7160 - val_loss: 28.5022\n",
      "Epoch 12/200\n",
      "380/380 [==============================] - 0s - loss: 30.1266 - val_loss: 26.8476\n",
      "Epoch 13/200\n",
      "380/380 [==============================] - 0s - loss: 28.8810 - val_loss: 25.8951\n",
      "Epoch 14/200\n",
      "380/380 [==============================] - 0s - loss: 27.4150 - val_loss: 24.4825\n",
      "Epoch 15/200\n",
      "380/380 [==============================] - 0s - loss: 26.1043 - val_loss: 23.1066\n",
      "Epoch 16/200\n",
      "380/380 [==============================] - 0s - loss: 24.8590 - val_loss: 22.0929\n",
      "Epoch 17/200\n",
      "380/380 [==============================] - 0s - loss: 23.7576 - val_loss: 20.8466\n",
      "Epoch 18/200\n",
      "380/380 [==============================] - 0s - loss: 22.6302 - val_loss: 20.5567\n",
      "Epoch 19/200\n",
      "380/380 [==============================] - 0s - loss: 21.6532 - val_loss: 19.9833\n",
      "Epoch 20/200\n",
      "380/380 [==============================] - 0s - loss: 21.0329 - val_loss: 19.6740\n",
      "Epoch 21/200\n",
      "380/380 [==============================] - 0s - loss: 20.3414 - val_loss: 19.0599\n",
      "Epoch 22/200\n",
      "380/380 [==============================] - 0s - loss: 19.3470 - val_loss: 18.8129\n",
      "Epoch 23/200\n",
      "380/380 [==============================] - 0s - loss: 18.7477 - val_loss: 18.7240\n",
      "Epoch 24/200\n",
      "380/380 [==============================] - 0s - loss: 18.1434 - val_loss: 18.5766\n",
      "Epoch 25/200\n",
      "380/380 [==============================] - 0s - loss: 17.7896 - val_loss: 18.2205\n",
      "Epoch 26/200\n",
      "380/380 [==============================] - 0s - loss: 17.2921 - val_loss: 17.8079\n",
      "Epoch 27/200\n",
      "380/380 [==============================] - 0s - loss: 16.9559 - val_loss: 17.4841\n",
      "Epoch 28/200\n",
      "380/380 [==============================] - 0s - loss: 16.4008 - val_loss: 17.3898\n",
      "Epoch 29/200\n",
      "380/380 [==============================] - 0s - loss: 16.1781 - val_loss: 17.4202\n",
      "Epoch 30/200\n",
      "380/380 [==============================] - 0s - loss: 15.6026 - val_loss: 17.1789\n",
      "Epoch 31/200\n",
      "380/380 [==============================] - 0s - loss: 15.3038 - val_loss: 16.8558\n",
      "Epoch 32/200\n",
      "380/380 [==============================] - 0s - loss: 15.0080 - val_loss: 17.0201\n",
      "Epoch 33/200\n",
      "380/380 [==============================] - 0s - loss: 14.7538 - val_loss: 16.4229\n",
      "Epoch 34/200\n",
      "380/380 [==============================] - 0s - loss: 14.5047 - val_loss: 16.3564\n",
      "Epoch 35/200\n",
      "380/380 [==============================] - 0s - loss: 14.0920 - val_loss: 16.0785\n",
      "Epoch 36/200\n",
      "380/380 [==============================] - 0s - loss: 13.8175 - val_loss: 16.0320\n",
      "Epoch 37/200\n",
      "380/380 [==============================] - 0s - loss: 13.6735 - val_loss: 15.8433\n",
      "Epoch 38/200\n",
      "380/380 [==============================] - 0s - loss: 13.4361 - val_loss: 15.4476\n",
      "Epoch 39/200\n",
      "380/380 [==============================] - 0s - loss: 13.1141 - val_loss: 15.7032\n",
      "Epoch 40/200\n",
      "380/380 [==============================] - 0s - loss: 13.1892 - val_loss: 15.2028\n",
      "Epoch 41/200\n",
      "380/380 [==============================] - 0s - loss: 13.2132 - val_loss: 14.8403\n",
      "Epoch 42/200\n",
      "380/380 [==============================] - 0s - loss: 12.6814 - val_loss: 15.0054\n",
      "Epoch 43/200\n",
      "380/380 [==============================] - 0s - loss: 12.3161 - val_loss: 14.9436\n",
      "Epoch 44/200\n",
      "380/380 [==============================] - 0s - loss: 12.2721 - val_loss: 15.3113\n",
      "Epoch 45/200\n",
      "380/380 [==============================] - 0s - loss: 12.0416 - val_loss: 14.9246\n",
      "Epoch 46/200\n",
      "380/380 [==============================] - 0s - loss: 11.9851 - val_loss: 14.4857\n",
      "Epoch 47/200\n",
      "380/380 [==============================] - 0s - loss: 11.7462 - val_loss: 14.1279\n",
      "Epoch 48/200\n",
      "380/380 [==============================] - 0s - loss: 11.6145 - val_loss: 13.9770\n",
      "Epoch 49/200\n",
      "380/380 [==============================] - 0s - loss: 11.4112 - val_loss: 14.5147\n",
      "Epoch 50/200\n",
      "380/380 [==============================] - 0s - loss: 11.3802 - val_loss: 14.2342\n",
      "Epoch 51/200\n",
      "380/380 [==============================] - 0s - loss: 11.1146 - val_loss: 14.0029\n",
      "Epoch 52/200\n",
      "380/380 [==============================] - 0s - loss: 10.9092 - val_loss: 13.6294\n",
      "Epoch 53/200\n",
      "380/380 [==============================] - 0s - loss: 10.8678 - val_loss: 13.4445\n",
      "Epoch 54/200\n",
      "380/380 [==============================] - 0s - loss: 10.6353 - val_loss: 13.1401\n",
      "Epoch 55/200\n",
      "380/380 [==============================] - 0s - loss: 10.4404 - val_loss: 13.3196\n",
      "Epoch 56/200\n",
      "380/380 [==============================] - 0s - loss: 10.3312 - val_loss: 13.0466\n",
      "Epoch 57/200\n",
      "380/380 [==============================] - 0s - loss: 10.1780 - val_loss: 12.7262\n",
      "Epoch 58/200\n",
      "380/380 [==============================] - 0s - loss: 10.3548 - val_loss: 12.7003\n",
      "Epoch 59/200\n",
      "380/380 [==============================] - 0s - loss: 10.0449 - val_loss: 12.8189\n",
      "Epoch 60/200\n",
      "380/380 [==============================] - 0s - loss: 9.9904 - val_loss: 12.9814\n",
      "Epoch 61/200\n",
      "380/380 [==============================] - 0s - loss: 9.8595 - val_loss: 13.0647\n",
      "Epoch 62/200\n",
      "380/380 [==============================] - 0s - loss: 9.6200 - val_loss: 13.0649\n",
      "Epoch 63/200\n",
      "380/380 [==============================] - 0s - loss: 9.5952 - val_loss: 12.7397\n",
      "Epoch 64/200\n",
      "380/380 [==============================] - 0s - loss: 9.5468 - val_loss: 12.3544\n",
      "Epoch 65/200\n",
      "380/380 [==============================] - 0s - loss: 9.3882 - val_loss: 12.0328\n",
      "Epoch 66/200\n",
      "380/380 [==============================] - 0s - loss: 9.4106 - val_loss: 12.2240\n",
      "Epoch 67/200\n",
      "380/380 [==============================] - 0s - loss: 9.1371 - val_loss: 12.0000\n",
      "Epoch 68/200\n",
      "380/380 [==============================] - 0s - loss: 9.1963 - val_loss: 12.9668\n",
      "Epoch 69/200\n",
      "380/380 [==============================] - 0s - loss: 9.1957 - val_loss: 11.9417\n",
      "Epoch 70/200\n",
      "380/380 [==============================] - 0s - loss: 9.0435 - val_loss: 12.0548\n",
      "Epoch 71/200\n",
      "380/380 [==============================] - ETA: 0s - loss: 8.938 - 0s - loss: 8.9136 - val_loss: 11.6622\n",
      "Epoch 72/200\n",
      "380/380 [==============================] - 0s - loss: 9.0454 - val_loss: 11.7711\n",
      "Epoch 73/200\n",
      "380/380 [==============================] - 0s - loss: 8.9491 - val_loss: 11.7554\n",
      "Epoch 74/200\n",
      "380/380 [==============================] - 0s - loss: 8.8328 - val_loss: 12.3775\n",
      "Epoch 75/200\n",
      "380/380 [==============================] - 0s - loss: 8.7829 - val_loss: 11.6114\n",
      "Epoch 76/200\n",
      "380/380 [==============================] - 0s - loss: 8.7702 - val_loss: 11.3503\n",
      "Epoch 77/200\n",
      "380/380 [==============================] - 0s - loss: 8.4967 - val_loss: 10.9734\n",
      "Epoch 78/200\n",
      "380/380 [==============================] - 0s - loss: 8.6216 - val_loss: 10.9293\n",
      "Epoch 79/200\n",
      "380/380 [==============================] - 0s - loss: 8.6634 - val_loss: 11.2114\n",
      "Epoch 80/200\n",
      "380/380 [==============================] - 0s - loss: 8.3452 - val_loss: 10.9365\n",
      "Epoch 81/200\n",
      "380/380 [==============================] - 0s - loss: 8.3287 - val_loss: 11.6567\n",
      "Epoch 82/200\n",
      "380/380 [==============================] - 0s - loss: 8.3446 - val_loss: 11.0912\n",
      "Epoch 83/200\n",
      "380/380 [==============================] - 0s - loss: 8.2224 - val_loss: 10.6377\n",
      "Epoch 84/200\n",
      "380/380 [==============================] - 0s - loss: 8.3771 - val_loss: 10.4636\n",
      "Epoch 85/200\n",
      "380/380 [==============================] - 0s - loss: 8.0033 - val_loss: 10.7075\n",
      "Epoch 86/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "380/380 [==============================] - 0s - loss: 8.0164 - val_loss: 10.3223\n",
      "Epoch 87/200\n",
      "380/380 [==============================] - 0s - loss: 8.1769 - val_loss: 10.3573\n",
      "Epoch 88/200\n",
      "380/380 [==============================] - 0s - loss: 8.1123 - val_loss: 10.0470\n",
      "Epoch 89/200\n",
      "380/380 [==============================] - 0s - loss: 7.9204 - val_loss: 10.1309\n",
      "Epoch 90/200\n",
      "380/380 [==============================] - 0s - loss: 8.0432 - val_loss: 10.5751\n",
      "Epoch 91/200\n",
      "380/380 [==============================] - 0s - loss: 7.9096 - val_loss: 9.8698\n",
      "Epoch 92/200\n",
      "380/380 [==============================] - 0s - loss: 8.2153 - val_loss: 9.6688\n",
      "Epoch 93/200\n",
      "380/380 [==============================] - 0s - loss: 8.3476 - val_loss: 10.9251\n",
      "Epoch 94/200\n",
      "380/380 [==============================] - 0s - loss: 7.9757 - val_loss: 10.2974\n",
      "Epoch 95/200\n",
      "380/380 [==============================] - 0s - loss: 7.7295 - val_loss: 10.4619\n",
      "Epoch 96/200\n",
      "380/380 [==============================] - 0s - loss: 7.7986 - val_loss: 9.5904\n",
      "Epoch 97/200\n",
      "380/380 [==============================] - 0s - loss: 7.8265 - val_loss: 10.2776\n",
      "Epoch 98/200\n",
      "380/380 [==============================] - 0s - loss: 7.5255 - val_loss: 10.3129\n",
      "Epoch 99/200\n",
      "380/380 [==============================] - 0s - loss: 7.5126 - val_loss: 10.0059\n",
      "Epoch 100/200\n",
      "380/380 [==============================] - 0s - loss: 7.6892 - val_loss: 10.1437\n",
      "Epoch 101/200\n",
      "380/380 [==============================] - 0s - loss: 7.6607 - val_loss: 10.4887\n",
      "Epoch 102/200\n",
      "380/380 [==============================] - 0s - loss: 7.9578 - val_loss: 9.7212\n",
      "Epoch 00101: early stopping\n"
     ]
    }
   ],
   "source": [
    "early_stopb = EarlyStopping(monitor='val_loss', patience=5, verbose=1) \n",
    "Resultb = modelb.fit(Xb_train, yb_train, validation_data = (Xb_test, yb_test), epochs=200, \n",
    "          batch_size=128, verbose=True, validation_split=0.25, callbacks=[early_stopb])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1047,
   "metadata": {},
   "outputs": [],
   "source": [
    "#[item for item in Resultb.history], [item for item in Result.history]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1048,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1c7ae338518>"
      ]
     },
     "execution_count": 1048,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEKCAYAAAAIO8L1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1c7ada7fdd8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.plot(Resultb.history['loss'])\n",
    "plt.plot(Resultb.history['val_loss'])\n",
    "plt.legend(['Training', 'Validation'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1049,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "MSE: 7.578426\n",
      "RMSE: 2.752894\n",
      "R^2: 0.910057\n"
     ]
    }
   ],
   "source": [
    "yb_predicted = modelb.predict(Xb_test)\n",
    "score = mean_squared_error(yb_predicted, yb_test)\n",
    "Rsq = r2_score(yb_test, yb_predicted)\n",
    "print('MSE: {0:f}'.format(score))\n",
    "print('RMSE: {0:f}'.format(score**0.5))\n",
    "print('R^2: {0:f}'.format(Rsq))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1050,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 12.23628044],\n",
       "       [ 32.5753479 ],\n",
       "       [ 23.41984367],\n",
       "       [ 30.30661583],\n",
       "       [ 22.70264435],\n",
       "       [ 38.24320221],\n",
       "       [ 23.43625069],\n",
       "       [ 18.67465973],\n",
       "       [ 34.53895187],\n",
       "       [ 21.55364799]], dtype=float32)"
      ]
     },
     "execution_count": 1050,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "yb_predicted[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1051,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 14.9,  31.7,  22.5,  30.7,  20.3,  27. ,  23. ,  18.6,  34.9,  21. ])"
      ]
     },
     "execution_count": 1051,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "yb_test[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1052,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(12.23628, 14.9)\n",
      "(32.575348, 31.699999999999999)\n",
      "(23.419844, 22.5)\n",
      "(30.306616, 30.699999999999999)\n",
      "(22.702644, 20.300000000000001)\n",
      "(38.243202, 27.0)\n",
      "(23.436251, 23.0)\n",
      "(18.67466, 18.600000000000001)\n",
      "(34.538952, 34.899999999999999)\n",
      "(21.553648, 21.0)\n"
     ]
    }
   ],
   "source": [
    "# put next to each other ...\n",
    "for i in zip(list(yb_predicted[:10,0]),yb_test[:10]):\n",
    "    print (i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [default]",
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}