MachineLearningCap13.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Net input: 2.50\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "#pip install Theano\n",
    "\n",
    "import theano\n",
    "from theano import tensor as T\n",
    "\n",
    "#initialize\n",
    "x1 = T.scalar()\n",
    "w1 = T.scalar()\n",
    "w0 = T.scalar()\n",
    "z1 = w1 * x1 + w0\n",
    "\n",
    "#compile\n",
    "net_input = theano.function(inputs=[w1,x1,w0],outputs=z1)\n",
    "\n",
    "#execute\n",
    "print('Net input: %.2f' %net_input(2.0,1.0,0.5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "float32\n"
     ]
    }
   ],
   "source": [
    "print(theano.config.floatX)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "theano.config.floatX = 'float32'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "float32\n"
     ]
    }
   ],
   "source": [
    "print(theano.config.floatX)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Otras opciones:\n",
    "# export THEANO_FLAGS=floatX=float32    (en bash)\n",
    "#THEANO_FLAGS=floatX=float32 python the_script.py    (al ejecutar el script)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cpu\n"
     ]
    }
   ],
   "source": [
    "print(theano.config.device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Por ejemplo, se puede:\n",
    "#THEANO_FLAGS=device=gpu,floatX=float32 python the_script.py\n",
    "#echo -e \"\\n[global]\\nfloatX=float32\\ndevice=gpu\\n\" >> ~/.theanorc  (para hacerlo permanente)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Column sum: [ 2.  4.  6.]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "#initialize\n",
    "#Para 64 bit, habría que usar \"dmatrix\" en vez de \"fmatrix\"\n",
    "x = T.fmatrix(name='x')\n",
    "x_sum = T.sum(x,axis=0)\n",
    "\n",
    "#compile\n",
    "calc_sum = theano.function(inputs=[x],outputs=x_sum)\n",
    "\n",
    "#execute (Python list)\n",
    "ary = [[1,2,3],[1,2,3]]\n",
    "print('Column sum:',calc_sum(ary))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Column sum: [ 2.  4.  6.]\n"
     ]
    }
   ],
   "source": [
    "#execute (numpy array)\n",
    "ary = np.array([[1,2,3],[1,2,3]], dtype=theano.config.floatX)\n",
    "print('Column sum:',calc_sum(ary))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x\n"
     ]
    }
   ],
   "source": [
    "print(x) #Muestra el nombre, si se le dio alguno. Si no, muestra el tipo de tensor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<TensorType(float32, matrix)>\n"
     ]
    }
   ],
   "source": [
    "print(x.type())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "z0:  [[ 0.]]\n",
      "z1:  [[ 6.]]\n",
      "z2:  [[ 12.]]\n",
      "z3:  [[ 18.]]\n",
      "z4:  [[ 24.]]\n"
     ]
    }
   ],
   "source": [
    "#initialize\n",
    "x = T.fmatrix('x')\n",
    "w = theano.shared(np.asarray([[0.0, 0.0, 0.0]],dtype=theano.config.floatX))\n",
    "z = x.dot(w.T)\n",
    "update = [[w,w+1.0]]\n",
    "\n",
    "#compile\n",
    "net_input = theano.function(inputs=[x],updates=update,outputs=z)\n",
    "\n",
    "#execute\n",
    "data = np.array([[1,2,3]],dtype=theano.config.floatX)\n",
    "for i in range(5):\n",
    "    print('z%d: '% i,net_input(data))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "z:  [[ 0.]]\n",
      "z:  [[ 6.]]\n",
      "z:  [[ 12.]]\n",
      "z:  [[ 18.]]\n",
      "z:  [[ 24.]]\n"
     ]
    }
   ],
   "source": [
    "#initialize\n",
    "data = np.array([[1,2,3]],dtype=theano.config.floatX)\n",
    "x = T.fmatrix('x')\n",
    "w = theano.shared(np.asarray([[0.0,0.0,0.0]],dtype=theano.config.floatX))\n",
    "z = x.dot(w.T)\n",
    "update = [[w,w+1.0]]\n",
    "\n",
    "#compile\n",
    "net_input = theano.function(inputs=[], updates=update, givens={x:data}, outputs=z)\n",
    "\n",
    "#execute\n",
    "for i in range(5):\n",
    "    print('z: ',net_input())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Linear Regression Example\n",
    "\n",
    "X_train = np.asarray([[0.0],[1.0],[2.0],[3.0],[4.0],[5.0],[6.0],[7.0],\n",
    "                     [8.0],[9.0]],dtype=theano.config.floatX)\n",
    "y_train = np.asarray([1.0, 1.3, 3.1, 2.0, 5.0, 6.3, 6.6, 7.4, 8.0, 9.0],dtype=theano.config.floatX)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import theano\n",
    "from theano import tensor as T\n",
    "import numpy as np\n",
    "\n",
    "def train_linreg(X_train,y_train,eta,epochs):\n",
    "    costs = []\n",
    "    #initialize arrays\n",
    "    eta0 = T.scalar('eta0')\n",
    "    y = T.fvector('y')\n",
    "    X = T.fmatrix('X')\n",
    "    w = theano.shared(np.zeros(shape=(X_train.shape[1]+1),dtype=theano.config.floatX),name='w')\n",
    "    \n",
    "    #calculate cost\n",
    "    net_input = T.dot(X,w[1:]) + w[0]\n",
    "    errors = y - net_input\n",
    "    cost = T.sum(T.pow(errors,2))\n",
    "    \n",
    "    #perform gradient update\n",
    "    gradient = T.grad(cost,wrt=w)\n",
    "    update = [(w, w - eta0 * gradient)]\n",
    "    \n",
    "    #compile model\n",
    "    train = theano.function(inputs=[eta0],outputs=cost,updates=update,givens={X: X_train, y:y_train})\n",
    "    \n",
    "    for _ in range(epochs):\n",
    "        costs.append(train(eta))\n",
    "    \n",
    "    return costs,w\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbcAAAEpCAYAAAAQ+2zYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nStTUQwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x111b7d5c0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "costs,w = train_linreg(X_train, y_train, eta=0.001, epochs=10)\n",
    "plt.plot(range(1,len(costs)+1),costs)\n",
    "plt.tight_layout()\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('cost')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def predict_linreg(X,w):\n",
    "    Xt = T.matrix(name='X')\n",
    "    net_input = T.dot(Xt,w[1:]) + w[0]\n",
    "    predict = theano.function(inputs=[Xt],givens={w: w},outputs=net_input)\n",
    "    \n",
    "    return predict(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEPCAYAAACtCNj2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x111e42f60>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(X_train,y_train,marker='s',s=50)\n",
    "#print('Xtrain.shape: ',X_train.shape,'\\n Predict_linreg: ',predict_linreg(X_train,w))\n",
    "plt.plot(range(X_train.shape[0]),predict_linreg(X_train,w),\n",
    "        color='gray',\n",
    "        marker='o',\n",
    "        markersize=4,\n",
    "        linewidth=3)\n",
    "plt.xlabel('x')\n",
    "plt.ylabel('y')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "P(y=1|x) = 0.707\n"
     ]
    }
   ],
   "source": [
    "X = np.array([[1.0, 1.4, 1.5]])\n",
    "w = np.array([0.0,0.2,0.4])\n",
    "\n",
    "def net_input(X,w):\n",
    "    z = X.dot(w)\n",
    "    return z\n",
    "\n",
    "def logistic(z):\n",
    "    return 1.0 / (1.0 + np.exp(-z))\n",
    "\n",
    "def logistic_activation(X,w):\n",
    "    z = net_input(X,w)\n",
    "    return logistic(z)\n",
    "\n",
    "print('P(y=1|x) = %.3f' % logistic_activation(X,w)[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Probabilities:\n",
      " [[ 0.87653295]\n",
      " [ 0.57688526]\n",
      " [ 0.90114393]]\n"
     ]
    }
   ],
   "source": [
    "#Con múltiples clases\n",
    "W = np.array([[1.1,1.2,1.3,0.5],\n",
    "             [0.1,0.2,0.4,0.1],\n",
    "            [0.2,0.5,2.1,1.9]])\n",
    "A = np.array([[1.0],\n",
    "             [0.1],\n",
    "             [0.3],\n",
    "             [0.7]])\n",
    "Z = W.dot(A)\n",
    "y_probas = logistic(Z)\n",
    "\n",
    "print('Probabilities:\\n',y_probas)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicted class label: 2\n"
     ]
    }
   ],
   "source": [
    "#Se ve que las \"probabilidades\" no suman 1, con la función logística\n",
    "y_class = np.argmax(Z,axis=0)\n",
    "print('predicted class label: %d'%y_class[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Probabilities:\n",
      " [[ 0.40386493]\n",
      " [ 0.07756222]\n",
      " [ 0.51857284]]\n"
     ]
    }
   ],
   "source": [
    "def softmax(z):\n",
    "    return np.exp(z) / np.sum(np.exp(z))\n",
    "\n",
    "def softmax_activation(X,w):\n",
    "    z = net_input(X,w)\n",
    "    return softmax(z)\n",
    "\n",
    "y_probas = softmax(Z)\n",
    "print('Probabilities:\\n',y_probas)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_probas.sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "predicted class label: 2\n"
     ]
    }
   ],
   "source": [
    "y_class = np.argmax(Z,axis=0)\n",
    "print('predicted class label: %d'%y_class[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaoAAAEbCAYAAACLGcAmAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x111ebc3c8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def tanh(z):\n",
    "    e_p = np.exp(z) \n",
    "    e_m = np.exp(-z)\n",
    "    return (e_p - e_m) / (e_p + e_m)  \n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "z = np.arange(-5, 5, 0.005)\n",
    "log_act = logistic(z)\n",
    "tanh_act = tanh(z)\n",
    "\n",
    "# alternatives:\n",
    "# from scipy.special import expit\n",
    "# log_act = expit(z)\n",
    "# tanh_act = np.tanh(z)\n",
    "\n",
    "plt.ylim([-1.5, 1.5])\n",
    "plt.xlabel('net input $z$')\n",
    "plt.ylabel('activation $\\phi(z)$')\n",
    "plt.axhline(1, color='black', linestyle='--')\n",
    "plt.axhline(0.5, color='black', linestyle='--')\n",
    "plt.axhline(0, color='black', linestyle='--')\n",
    "plt.axhline(-1, color='black', linestyle='--')\n",
    "\n",
    "plt.plot(z, tanh_act, \n",
    "         linewidth=2, \n",
    "         color='black', \n",
    "         label='tanh')\n",
    "plt.plot(z, log_act, \n",
    "         linewidth=2, \n",
    "         color='lightgreen', \n",
    "         label='logistic')\n",
    "\n",
    "plt.legend(loc='lower right')\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Keras:  >pip install Keras\n",
    "#http://yann.lecun.com/exdb/mnist\n",
    "\n",
    "import os\n",
    "import struct\n",
    "import numpy as np\n",
    "\n",
    "def load_mnist(path, kind='train'):\n",
    "    \"\"\"Load MNIST data from 'path'\"\"\"\n",
    "    labels_path = os.path.join(path,'%s-labels.idx1-ubyte'%kind)\n",
    "    images_path = os.path.join(path,'%s-images.idx3-ubyte'%kind)\n",
    "    with open(labels_path, 'rb') as lbpath:\n",
    "        magic,n = struct.unpack('>II',lbpath.read(8))\n",
    "        labels = np.fromfile(lbpath,dtype=np.uint8)\n",
    "    with open(images_path,'rb') as imgpath:\n",
    "        masgic,num,rows,cols = struct.unpack(\">IIII\",imgpath.read(16))\n",
    "        images = np.fromfile(imgpath,dtype=np.uint8).reshape(len(labels),784)\n",
    "\n",
    "    return images, labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Rows: 60000, columns: 784\n"
     ]
    }
   ],
   "source": [
    "X_train,y_train = load_mnist('/Users/antelInvestigacion/IA/mnist',kind='train')\n",
    "print('Rows: %d, columns: %d'%(X_train.shape[0], X_train.shape[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Rows: 10000, columns: 784\n"
     ]
    }
   ],
   "source": [
    "X_test,y_test = load_mnist('/Users/antelInvestigacion/IA/mnist',kind='t10k')\n",
    "print('Rows: %d, columns: %d'%(X_test.shape[0], X_test.shape[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import theano\n",
    "theano.config.floatX = 'float32'\n",
    "X_train = X_train.astype(theano.config.floatX)\n",
    "y_train = y_train.astype(theano.config.floatX)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "First 3 labels:  [ 5.  0.  4.]\n"
     ]
    }
   ],
   "source": [
    "from keras.utils import np_utils\n",
    "print('First 3 labels: ',y_train[:3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "First 3 labels (one-hot):  [[ 0.  0.  0.  0.  0.  1.  0.  0.  0.  0.]\n",
      " [ 1.  0.  0.  0.  0.  0.  0.  0.  0.  0.]\n",
      " [ 0.  0.  0.  0.  1.  0.  0.  0.  0.  0.]]\n"
     ]
    }
   ],
   "source": [
    "y_train_ohe = np_utils.to_categorical(y_train)\n",
    "print('First 3 labels (one-hot): ',y_train_ohe[:3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense\n",
    "from keras.optimizers import SGD\n",
    "\n",
    "np.random.seed(1)\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(input_dim=X_train.shape[1],\n",
    "               output_dim=50,\n",
    "               init='uniform',\n",
    "               activation='tanh'))\n",
    "\n",
    "model.add(Dense(input_dim=50,\n",
    "               output_dim=50,\n",
    "               init='uniform',\n",
    "               activation='tanh'))\n",
    "\n",
    "model.add(Dense(input_dim=50,\n",
    "               output_dim=y_train_ohe.shape[1],\n",
    "               init='uniform',\n",
    "               activation='softmax'))\n",
    "\n",
    "sgd = SGD(lr=0.001,decay=1e-7,momentum=.9)\n",
    "model.compile(loss='categorical_crossentropy',optimizer=sgd,metrics=[\"accuracy\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 54000 samples, validate on 6000 samples\n",
      "Epoch 1/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1213 - val_loss: 0.1400\n",
      "Epoch 2/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1249 - val_loss: 0.1388\n",
      "Epoch 3/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1231 - val_loss: 0.1402\n",
      "Epoch 4/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1261 - val_loss: 0.1344\n",
      "Epoch 5/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1257 - val_loss: 0.1402\n",
      "Epoch 6/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1213 - val_loss: 0.1296\n",
      "Epoch 7/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1221 - val_loss: 0.1270\n",
      "Epoch 8/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1214 - val_loss: 0.1402\n",
      "Epoch 9/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1218 - val_loss: 0.1424\n",
      "Epoch 10/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1232 - val_loss: 0.1328\n",
      "Epoch 11/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1177 - val_loss: 0.1309\n",
      "Epoch 12/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1186 - val_loss: 0.1364\n",
      "Epoch 13/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1164 - val_loss: 0.1328\n",
      "Epoch 14/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1183 - val_loss: 0.1335\n",
      "Epoch 15/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1160 - val_loss: 0.1301\n",
      "Epoch 16/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1191 - val_loss: 0.1330\n",
      "Epoch 17/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1151 - val_loss: 0.1319\n",
      "Epoch 18/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1105 - val_loss: 0.1324\n",
      "Epoch 19/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1140 - val_loss: 0.1285\n",
      "Epoch 20/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1153 - val_loss: 0.1313\n",
      "Epoch 21/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1165 - val_loss: 0.1270\n",
      "Epoch 22/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1115 - val_loss: 0.1299\n",
      "Epoch 23/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1155 - val_loss: 0.1312\n",
      "Epoch 24/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1120 - val_loss: 0.1270\n",
      "Epoch 25/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1109 - val_loss: 0.1280\n",
      "Epoch 26/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1171 - val_loss: 0.1399\n",
      "Epoch 27/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1241 - val_loss: 0.1358\n",
      "Epoch 28/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1168 - val_loss: 0.1400\n",
      "Epoch 29/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1170 - val_loss: 0.1294\n",
      "Epoch 30/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1126 - val_loss: 0.1354\n",
      "Epoch 31/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1166 - val_loss: 0.1343\n",
      "Epoch 32/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1154 - val_loss: 0.1346\n",
      "Epoch 33/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1107 - val_loss: 0.1298\n",
      "Epoch 34/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1075 - val_loss: 0.1276\n",
      "Epoch 35/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1096 - val_loss: 0.1343\n",
      "Epoch 36/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1122 - val_loss: 0.1377\n",
      "Epoch 37/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1118 - val_loss: 0.1391\n",
      "Epoch 38/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1135 - val_loss: 0.1396\n",
      "Epoch 39/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1098 - val_loss: 0.1235\n",
      "Epoch 40/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1059 - val_loss: 0.1270\n",
      "Epoch 41/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1082 - val_loss: 0.1341\n",
      "Epoch 42/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1104 - val_loss: 0.1298\n",
      "Epoch 43/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1110 - val_loss: 0.1325\n",
      "Epoch 44/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1108 - val_loss: 0.1383\n",
      "Epoch 45/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1125 - val_loss: 0.1275\n",
      "Epoch 46/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1143 - val_loss: 0.1363\n",
      "Epoch 47/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1112 - val_loss: 0.1310\n",
      "Epoch 48/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1121 - val_loss: 0.1402\n",
      "Epoch 49/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1164 - val_loss: 0.1308\n",
      "Epoch 50/50\n",
      "54000/54000 [==============================] - 0s - loss: 0.1112 - val_loss: 0.1310\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x11203f550>"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train,\n",
    "         y_train_ohe,\n",
    "         nb_epoch=50,\n",
    "         batch_size=300,\n",
    "         verbose=1,\n",
    "         validation_split=0.1)\n",
    "         #,show_accuracy=True) Deprecated"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "First 3 predictions:  [5 0 4]\n"
     ]
    }
   ],
   "source": [
    "y_train_pred = model.predict_classes(X_train,verbose=0)\n",
    "print('First 3 predictions: ',y_train_pred[:3])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training accuracy: 96.59%\n"
     ]
    }
   ],
   "source": [
    "train_acc = np.sum(y_train==y_train_pred,axis=0) / X_train.shape[0]\n",
    "print('Training accuracy: %.2f%%'% (train_acc*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test accuracy: 95.47%\n"
     ]
    }
   ],
   "source": [
    "y_test_pred = model.predict_classes(X_test,verbose=0)\n",
    "test_acc = np.sum(y_test==y_test_pred,axis=0) / X_test.shape[0]\n",
    "print('Test accuracy: %.2f%%'% (test_acc*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}