Chapter02/1 Perceptron.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-19T11:38:47.898934Z",
     "start_time": "2019-01-19T11:38:47.374844Z"
    }
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns; sns.set()\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "# initiating random number\n",
    "np.random.seed(11)\n",
    "\n",
    "#### Creating the dataset\n",
    "\n",
    "# mean and standard deviation for the x belonging to the first class\n",
    "mu_x1, sigma_x1 = 0, 0.1 \n",
    "\n",
    "# constat to make the second distribution different from the first\n",
    "x2_mu_diff = 0.35\n",
    "\n",
    "# creating the first distribution\n",
    "d1 = pd.DataFrame({'x1': np.random.normal(mu_x1, sigma_x1 , 1000),\n",
    "                   'x2': np.random.normal(mu_x1, sigma_x1 , 1000),\n",
    "                   'type': 0})\n",
    "\n",
    "# creating the second distribution\n",
    "d2 = pd.DataFrame({'x1': np.random.normal(mu_x1, sigma_x1 , 1000) + x2_mu_diff,\n",
    "                   'x2': np.random.normal(mu_x1, sigma_x1 , 1000) + x2_mu_diff,\n",
    "                   'type': 1})\n",
    "\n",
    "data = pd.concat([d1, d2], ignore_index=True)\n",
    "\n",
    "\n",
    "ax = sns.scatterplot(x=\"x1\", y=\"x2\", hue=\"type\",\n",
    "                      data=data)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-19T11:38:53.295248Z",
     "start_time": "2019-01-19T11:38:53.278308Z"
    }
   },
   "outputs": [],
   "source": [
    "class Perceptron(object):\n",
    "    \"\"\"\n",
    "    Simple implementation of the perceptron algorithm\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self, w0=1, w1=0.1, w2=0.1):\n",
    "\n",
    "        # weights\n",
    "        self.w0 = w0  # bias\n",
    "        self.w1 = w1\n",
    "        self.w2 = w2\n",
    "\n",
    "    def step_function(self, z):\n",
    "        if z >= 0:\n",
    "            return 1\n",
    "        else:\n",
    "            return 0\n",
    "\n",
    "    def weighted_sum_inputs(self, x1, x2):\n",
    "        return sum([1 * self.w0, x1 * self.w1, x2 * self.w2])\n",
    "\n",
    "    def predict(self, x1, x2):\n",
    "        \"\"\"\n",
    "        Uses the step function to determine the output\n",
    "        \"\"\"\n",
    "        z = self.weighted_sum_inputs(x1, x2)\n",
    "\n",
    "        return self.step_function(z)\n",
    "\n",
    "    def predict_boundary(self, x):\n",
    "        \"\"\"\n",
    "        Used to predict the boundaries of our classifier\n",
    "        \"\"\"\n",
    "        return -(self.w1 * x + self.w0) / self.w2\n",
    "\n",
    "    def fit(self, X, y, epochs=1, step=0.1, verbose=True):\n",
    "        \"\"\"\n",
    "        Train the model given the dataset\n",
    "        \"\"\"\n",
    "        errors = []\n",
    "\n",
    "        for epoch in range(epochs):\n",
    "            error = 0\n",
    "            for i in range(0, len(X.index)):\n",
    "                x1, x2, target = X.values[i][0], X.values[i][1], y.values[i]\n",
    "                # The update is proportional to the step size and the error\n",
    "                update = step * (target - self.predict(x1, x2))\n",
    "                self.w1 += update * x1\n",
    "                self.w2 += update * x2\n",
    "                self.w0 += update\n",
    "                error += int(update != 0.0)\n",
    "            errors.append(error)\n",
    "            if verbose:\n",
    "                print('Epochs: {} - Error: {} - Errors from all epochs: {}'\\\n",
    "                      .format(epoch, error, errors))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-19T11:38:53.896208Z",
     "start_time": "2019-01-19T11:38:53.882750Z"
    }
   },
   "outputs": [],
   "source": [
    "# Splitting the dataset in training and test set\n",
    "msk = np.random.rand(len(data)) < 0.8\n",
    "\n",
    "# Roughly 80% of data will go in the training set\n",
    "train_x, train_y = data[['x1','x2']][msk], data.type[msk]\n",
    "# Everything else will go into the validation set\n",
    "test_x, test_y = data[['x1','x2']][~msk], data.type[~msk]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-13T23:58:00.899918Z",
     "start_time": "2019-01-13T23:58:00.795720Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epochs: 0 - Error: 33 - Errors from all epochs: [33]\n",
      "        Predicted 0  Predicted 1\n",
      "True 0          201            5\n",
      "True 1            0          176\n"
     ]
    }
   ],
   "source": [
    "my_perceptron = Perceptron(0.1,0.1)\n",
    "\n",
    "my_perceptron.fit(train_x, train_y, epochs=1, step=0.005)\n",
    "\n",
    "pred_y = test_x.apply(lambda x: my_perceptron.predict(x.x1, x.x2), axis=1)\n",
    "\n",
    "cm = confusion_matrix(test_y, pred_y, labels=[0, 1])\n",
    "\n",
    "print(pd.DataFrame(cm,\n",
    "                   index=['True 0', 'True 1'], \n",
    "                   columns=['Predicted 0', 'Predicted 1']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-13T23:58:21.189603Z",
     "start_time": "2019-01-13T23:58:21.180054Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(-0.025000000000000015, 0.09623421555805478, 0.09472342868828437)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "my_perceptron.w0, my_perceptron.w1, my_perceptron.w2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fb1ac093438>]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Adds decision boundary line to the scatterplot\n",
    "\n",
    "ax = sns.scatterplot(x=\"x1\", y=\"x2\", hue=\"type\",\n",
    "                      data=data[~msk])\n",
    "ax.autoscale(False)\n",
    "x_vals = np.array(ax.get_xlim())\n",
    "y_vals = my_perceptron.predict_boundary(x_vals)\n",
    "ax.plot(x_vals, y_vals, '--', c=\"red\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-20T04:12:44.747113Z",
     "start_time": "2019-01-20T04:12:41.193681Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /home/leo/anaconda3/lib/python3.7/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Colocations handled automatically by placer.\n",
      "WARNING:tensorflow:From /home/leo/anaconda3/lib/python3.7/site-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "Epoch 1/1\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/leo/anaconda3/lib/python3.7/site-packages/ipykernel_launcher.py:22: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1618/1618 [==============================] - 2s 1ms/step - loss: 0.1573\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7fb180d130f0>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "from keras.models import Sequential\n",
    "from keras.optimizers import SGD\n",
    "from keras.layers import Dense\n",
    "import keras.backend as K\n",
    "import tensorflow as tf\n",
    "\n",
    "    \n",
    "network = Sequential()\n",
    "network.add(Dense(1, input_dim = 2, activation = 'sigmoid'))\n",
    "network.compile(loss = \"mse\", optimizer = SGD(lr = 0.01))\n",
    "network.fit(train_x, train_y , nb_epoch = 1, batch_size = 1, shuffle = False)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-14T00:09:54.486094Z",
     "start_time": "2019-01-14T00:09:54.465707Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9988415710503089"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "network.get_weights()\n",
    "\n",
    "pred_y = network.predict(test_x)\n",
    "\n",
    "\n",
    "from sklearn.metrics import roc_auc_score\n",
    "\n",
    "roc_auc_score(test_y, pred_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-20T04:15:52.001970Z",
     "start_time": "2019-01-20T04:14:36.004198Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/30\n",
      "1618/1618 [==============================] - 1s 562us/step - loss: 0.2227\n",
      "Epoch 2/30\n",
      "1618/1618 [==============================] - 1s 561us/step - loss: 0.1793\n",
      "Epoch 3/30\n",
      "1618/1618 [==============================] - 1s 570us/step - loss: 0.1480\n",
      "Epoch 4/30\n",
      "1618/1618 [==============================] - 1s 599us/step - loss: 0.1254\n",
      "Epoch 5/30\n",
      "1618/1618 [==============================] - 2s 949us/step - loss: 0.1087\n",
      "Epoch 6/30\n",
      "1618/1618 [==============================] - 1s 887us/step - loss: 0.0960\n",
      "Epoch 7/30\n",
      "1618/1618 [==============================] - 2s 1ms/step - loss: 0.0861\n",
      "Epoch 8/30\n",
      "1618/1618 [==============================] - 1s 577us/step - loss: 0.0781\n",
      "Epoch 9/30\n",
      "1618/1618 [==============================] - 1s 536us/step - loss: 0.0717\n",
      "Epoch 10/30\n",
      "1618/1618 [==============================] - 1s 639us/step - loss: 0.0664\n",
      "Epoch 11/30\n",
      "1618/1618 [==============================] - 1s 550us/step - loss: 0.0619\n",
      "Epoch 12/30\n",
      "1618/1618 [==============================] - 1s 503us/step - loss: 0.0581\n",
      "Epoch 13/30\n",
      "1618/1618 [==============================] - 2s 1ms/step - loss: 0.0548\n",
      "Epoch 14/30\n",
      "1618/1618 [==============================] - 1s 704us/step - loss: 0.0519\n",
      "Epoch 15/30\n",
      "1618/1618 [==============================] - 1s 706us/step - loss: 0.0494\n",
      "Epoch 16/30\n",
      "1618/1618 [==============================] - 1s 926us/step - loss: 0.0472\n",
      "Epoch 17/30\n",
      "1618/1618 [==============================] - 2s 1ms/step - loss: 0.0452\n",
      "Epoch 18/30\n",
      "1618/1618 [==============================] - 1s 868us/step - loss: 0.0434\n",
      "Epoch 19/30\n",
      "1618/1618 [==============================] - 1s 700us/step - loss: 0.0418\n",
      "Epoch 20/30\n",
      "1618/1618 [==============================] - 1s 702us/step - loss: 0.0403\n",
      "Epoch 21/30\n",
      "1618/1618 [==============================] - 1s 698us/step - loss: 0.0390\n",
      "Epoch 22/30\n",
      "1618/1618 [==============================] - 1s 692us/step - loss: 0.0377\n",
      "Epoch 23/30\n",
      "1618/1618 [==============================] - 1s 694us/step - loss: 0.0366\n",
      "Epoch 24/30\n",
      "1618/1618 [==============================] - 1s 699us/step - loss: 0.0356\n",
      "Epoch 25/30\n",
      "1618/1618 [==============================] - 1s 697us/step - loss: 0.0346\n",
      "Epoch 26/30\n",
      "1618/1618 [==============================] - 1s 694us/step - loss: 0.0337\n",
      "Epoch 27/30\n",
      "1618/1618 [==============================] - 1s 705us/step - loss: 0.0329\n",
      "Epoch 28/30\n",
      "1618/1618 [==============================] - 2s 968us/step - loss: 0.0321\n",
      "Epoch 29/30\n",
      "1618/1618 [==============================] - 1s 817us/step - loss: 0.0314\n",
      "Epoch 30/30\n",
      "1618/1618 [==============================] - 2s 940us/step - loss: 0.0307\n",
      "MSE on the test set: 0.03451991137274204\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation\n",
    "from keras.optimizers import SGD\n",
    "from sklearn.metrics import mean_squared_error\n",
    "\n",
    "my_perceptron = Sequential()\n",
    "\n",
    "input_layer = Dense(1, input_dim=2, activation=\"sigmoid\", kernel_initializer=\"zero\")\n",
    "my_perceptron.add(input_layer)\n",
    "\n",
    "my_perceptron.compile(loss=\"mse\", optimizer=SGD(lr=0.01))\n",
    "\n",
    "my_perceptron.fit(train_x.values, train_y, epochs=30, batch_size=1, shuffle=True)\n",
    "pred_y = my_perceptron.predict(test_x)\n",
    "\n",
    "print('MSE on the test set:', mean_squared_error(pred_y, test_y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "hide_input": false,
  "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": "3.7.3"
  },
  "toc": {
   "colors": {
    "hover_highlight": "#DAA520",
    "running_highlight": "#FF0000",
    "selected_highlight": "#FFD700"
   },
   "moveMenuLeft": true,
   "nav_menu": {
    "height": "12px",
    "width": "252px"
   },
   "navigate_menu": true,
   "number_sections": true,
   "sideBar": true,
   "threshold": 4,
   "toc_cell": false,
   "toc_section_display": "block",
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}