numpy_matplotlib_sklearn.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Numpy, Matplotlib and Sklearn Tutorial"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We often use numpy to handle high dimensional arrays.\n",
    "\n",
    "Let's try the basic operation of numpy:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2 (2, 3) 6 int64 <class 'numpy.ndarray'>\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "a = np.array([[1,2,3], [2,3,4]])\n",
    "print(a.ndim, a.shape, a.size, a.dtype, type(a))\n",
    "\n",
    "b = np.zeros((3,4))\n",
    "c = np.ones((3,4))\n",
    "d = np.random.randn(2,3)\n",
    "e = np.array([[1,2], [2,3], [3,4]])\n",
    "f = b*2 - c*3\n",
    "g = 2*c*f\n",
    "h = np.dot(a,e)\n",
    "i = d.mean()\n",
    "j = d.max(axis=1)\n",
    "k = a[-1][:2]\n",
    "\n",
    "# You can print a to k for details"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "matplotlib.pyplot provides very useful apis for drawing graphs.\n",
    "\n",
    "Let's try the basic operation of matplotlib.pyplot:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "x = np.arange(2, 10, 0.2)\n",
    "\n",
    "plt.plot(x, x**1.5*.5, 'r-', x, np.log(x)*5, 'g--', x, x, 'b.')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If you want to print them in different graphs, try this:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def f(x):\n",
    "    return np.sin(np.pi*x)\n",
    "\n",
    "x1 = np.arange(0, 5, 0.1)\n",
    "x2 = np.arange(0, 5, 0.01)\n",
    "\n",
    "plt.subplot(211)\n",
    "plt.plot(x1, f(x1), 'go', x2, f(x2-1))\n",
    "\n",
    "plt.subplot(212)\n",
    "plt.plot(x2, f(x2), 'r--')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "How about printing images?\n",
    "\n",
    "Let's try to print a image whose pixels gradually change:\n",
    "\n",
    "Different pixel values represent different gray levels."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "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": [
    "img = np.arange(0, 1, 1/32/32) # define an 1D array with 32x32 elements gradually increasing\n",
    "img = img.reshape(32, 32) # reshape it into 32x32 array, the array represents a 32x32 image,\n",
    "                          # each element represents the corresponding pixel of the image\n",
    "plt.imshow(img, cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "Based on numpy, Scikit-learn (sklearn) provides a lot of tools for machine learning.It is a very powerful machine learning library.\n",
    "\n",
    "Then, let's use it for mnist classification:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/macbook/anaconda2/envs/python37/lib/python3.7/site-packages/sklearn/utils/deprecation.py:77: DeprecationWarning: Function fetch_mldata is deprecated; fetch_mldata was deprecated in version 0.20 and will be removed in version 0.22\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n",
      "/Users/macbook/anaconda2/envs/python37/lib/python3.7/site-packages/sklearn/utils/deprecation.py:77: DeprecationWarning: Function mldata_filename is deprecated; mldata_filename was deprecated in version 0.20 and will be removed in version 0.22\n",
      "  warnings.warn(msg, category=DeprecationWarning)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "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": [
    "from sklearn.datasets import fetch_mldata\n",
    "\n",
    "# download and read mnist\n",
    "mnist = fetch_mldata('MNIST original', data_home='./')\n",
    "\n",
    "# 'mnist.data' is 70k x 784 array, each row represents the pixels from a 28x28=784 image\n",
    "# 'mnist.target' is 70k x 1 array, each row represents the target class of the corresponding image\n",
    "images = mnist.data\n",
    "targets = mnist.target\n",
    "\n",
    "# make the value of pixels from [0, 255] to [0, 1] for further process\n",
    "X = mnist.data / 255.\n",
    "Y = mnist.target\n",
    "\n",
    "# print the first image of the dataset\n",
    "img1 = X[0].reshape(28, 28)\n",
    "plt.imshow(img1, cmap='gray')\n",
    "plt.show()\n",
    "\n",
    "# print the images after simple transformation\n",
    "img2 = 1 - img1\n",
    "plt.imshow(img2, cmap='gray')\n",
    "plt.show()\n",
    "\n",
    "img3 = img1.transpose()\n",
    "plt.imshow(img3, cmap='gray')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# split data to train and test (for faster calculation, just use 1/10 data)\n",
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X[::10], Y[::10], test_size=1000)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "#### Q1:\n",
    "Please use the logistic regression(default parameters) in sklearn to classify the data above, and print the training accuracy and test accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/macbook/anaconda2/envs/python37/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:460: FutureWarning: Default multi_class will be changed to 'auto' in 0.22. Specify the multi_class option to silence this warning.\n",
      "  \"this warning.\", FutureWarning)\n",
      "/Users/macbook/anaconda2/envs/python37/lib/python3.7/site-packages/sklearn/linear_model/logistic.py:460: FutureWarning: Default multi_class will be changed to 'auto' in 0.22. Specify the multi_class option to silence this warning.\n",
      "  \"this warning.\", FutureWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training accuracy: 95.63%\n",
      "Testing accuracy: 99.80%\n"
     ]
    }
   ],
   "source": [
    "# TODO:use logistic regression\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn import metrics\n",
    "# all parameters not specified are set to their defaults\n",
    "# default solver is incredibly slow thats why we change it\n",
    "logisticRegr = LogisticRegression(solver = 'lbfgs',max_iter = 10000)\n",
    "logisticRegr.fit(X_train, Y_train)\n",
    "predictions = logisticRegr.predict(X_train)\n",
    "train_accuracy= logisticRegr.score(X_train, Y_train)\n",
    "logisticRegr.fit(X_test, Y_test)\n",
    "predictions = logisticRegr.predict(X_test)\n",
    "test_accuracy= logisticRegr.score(X_test, Y_test)\n",
    "\n",
    "print('Training accuracy: %0.2f%%' % (train_accuracy*100))\n",
    "print('Testing accuracy: %0.2f%%' % (test_accuracy*100))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Q2:\n",
    "Please use the naive bayes(Bernoulli, default parameters) in sklearn to classify the data above, and print the training accuracy and test accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training accuracy: 83.68%\n",
      "Testing accuracy: 82.90%\n"
     ]
    }
   ],
   "source": [
    "# TODO:use naive bayes\n",
    "\n",
    "from sklearn.naive_bayes import BernoulliNB\n",
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "clf = BernoulliNB()\n",
    "clf.fit(X_train,Y_train)\n",
    "Y_predicted_train = clf.predict(X_train)\n",
    "Y_predicted = clf.predict(X_test)\n",
    "train_accuracy = accuracy_score(Y_train,Y_predicted_train)\n",
    "test_accuracy = accuracy_score(Y_test, Y_predicted)\n",
    "print('Training accuracy: %0.2f%%' % (train_accuracy*100))\n",
    "print('Testing accuracy: %0.2f%%' % (test_accuracy*100))\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Q3:\n",
    "Please use the support vector machine(default parameters) in sklearn to classify the data above, and print the training accuracy and test accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training accuracy: 83.68%\n",
      "Testing accuracy: 87.80%\n"
     ]
    }
   ],
   "source": [
    "# TODO:use support vector machine\n",
    "from sklearn.svm import LinearSVC\n",
    "#from sklearn.svm import SVC\n",
    "from sklearn.datasets import make_classification\n",
    "clf_svm = LinearSVC(max_iter=10000)\n",
    "clf_svm.fit(X_train, Y_train)\n",
    "Y_pred_svm = clf_svm.predict(X_test)\n",
    "Y_predicted = clf_svm.predict(X_train)\n",
    "test_accuracy = accuracy_score(Y_test, Y_pred_svm)\n",
    "train_accuarcy = accuracy_score(Y_train, Y_predicted)\n",
    "\n",
    "\n",
    "print('Training accuracy: %0.2f%%' % (train_accuracy*100))\n",
    "print('Testing accuracy: %0.2f%%' % (test_accuracy*100))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Q4:\n",
    "Please adjust the parameters of SVM to increase the testing accuracy, and print the training accuracy and test accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training accuracy: 83.68%\n",
      "Testing accuracy: 95.40%\n"
     ]
    }
   ],
   "source": [
    "# TODO:use SVM with another group of parameters\n",
    "from sklearn.svm import SVC\n",
    "\n",
    "clf_svm = SVC(C=2.9470517025518097, gamma=0.015998587196060572, cache_size= 2000)\n",
    "clf_svm.fit(X_train, Y_train)\n",
    "Y_pred_svm = clf_svm.predict(X_test)\n",
    "Y_predicted = clf_svm.predict(X_train)\n",
    "test_accuracy = accuracy_score(Y_test, Y_pred_svm)\n",
    "train_accuarcy = accuracy_score(Y_train, Y_predicted)\n",
    "\n",
    "print('Training accuracy: %0.2f%%' % (train_accuracy*100))\n",
    "print('Testing accuracy: %0.2f%%' % (test_accuracy*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "python37",
   "language": "python",
   "name": "my_env"
  },
  "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"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}