{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "# Notebook 8: Bagging a simple binary classifier" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "## Learning Goals\n", "The goal of this notebook is to understand Bootstrap Aggregation or Bagging using a simple classifier. We will write code and try to gain intuition for why ensemble methods turn out to be so powerful (especially when we know features we care about).\n", "\n", "## Overview\n", "In this notebook, we introduce the perceptron learning algorithm (PLA) that is often used for binary classification. Then we treat PLA as the base algorithm and demonstrate how to combine it with bootstrapping (i.e. bootstrap aggregation, bagging). " ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Perceptron Learning algorithm (PLA): ### \n", "\n", "Suppose that we're given a set of $N$ observations each bearing $p$ features, $\\textbf{x}_n=(x_1^{(n)},\\cdots, x_p^{(n)})\\in\\mathbb{R}^p$, $n=1,\\cdots, N$. The goal of binary classification is to relate these observations to their corresponding binary label $y_n \\in\\{+1,-1\\}$. Concretely, this amounts to finding a function $h: \\mathbb{R}^p\\rightarrow \\{+1,-1\\}$ such that $h(\\textbf{x}_n)$ is ideally the same as $y_n$. A perceptron accomplishes this feat by utilizing a set of weights $\\textbf{w}=(w_0,w_1,\\cdots, w_d)\\in\\mathbb{R}^{p+1}$ to construct $h$ so that labeling is done through\n", "\n", "$$\n", "h(\\textbf{x}_n)=\\text{sign }\\left(w_0+\\sum_{i=1}^p w_ix_i^{(n)}\\right) =\\text{sign }(\\textbf{w}^T\\tilde{\\textbf{x}}_n),\n", "$$\n", "where $\\tilde{\\textbf{x}}_n=(1,x_1^{(n)},\\cdots, x_p^{(n)}) = (1,\\textbf{x}_n)$. The perceptron can be viewed as the zero-temperature limit of the logistic regression where the sigmoid (Fermi-function) becomes a step function.\n", "\n", "PLA begins with randomized weights. It then selects a point from the training set at random. If this point, say, $\\textbf{x}_n$, is misclassified, namely, $y_n\\neq \\text{sign }(\\textbf{w}^T\\tilde{\\textbf{x}}_n)$, weights are updated according to \n", "$$\n", "\\textbf{w}\\leftarrow \\textbf{w}+ y_n\\tilde{\\textbf{x}}_n\n", "$$\n", "Otherwise, $\\textbf{w}$ is preserved and PLA moves on to select another point. This procedure continues until a specified threshold is met, after which PLA outputs $h$. It is clear that PLA is an online algorithm since it does not treat all available data at the same time. Instead, it learns the weights as it progress along data points in the training set one-by-one. The update rule is built on the intuition that whenever a mistake is encountered, it corrects its weights by moving towards the right direction. \n", "\n", "The following implementation of perceptron class is adapted from [this blog](https://datasciencelab.wordpress.com/2014/01/10/machine-learning-classics-the-perceptron/). It considers 2-feature observations in $[-1,1]\\times[-1,1]$. This means that we can write down the weight vector as $\\textbf{w}=(w_0,w_1,w_2)$. Prediction for any point $\\textbf{x}_n=(x_1^{(n)}, x_2^{(n)})$ in this domain is therefore \n", "$$\n", "h(\\textbf{x}_n)=\\text{sign} (w_0+w_1x_1^{(n)}+w_2x_1^{(n)})\n", "$$" ] }, { "cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Automatically created module for IPython interactive environment\n" ] } ], "source": [ "import numpy as np\n", "import random\n", "import os, subprocess\n", "from random import randrange\n", "import matplotlib.pyplot as plt\n", "from operator import add\n", "print(__doc__)\n", "#import seaborn as sns\n", " \n", "class Perceptron:\n", " def __init__(self, N, boostrap_data, inputX = None, inputS = None):\n", " # Random linearly separated data\n", " xA,yA,xB,yB = [random.uniform(-1, 1) for i in range(4)]\n", " self.V = np.array([xB*yA-xA*yB, yB-yA, xA-xB])\n", "\n", " if boostrap_data is None:\n", " self.X = self.generate_points(N, inputX)\n", " else:\n", " self.X = bootstrap_data\n", " \n", " def generate_points(self, N, inputX = None, inputS = None):\n", " \n", " X = []\n", "\n", " if (inputX is None) and (inputS is None):\n", " for i in range(N):\n", " x1,x2 = [random.uniform(-1, 1) for i in range(2)]\n", " #x1 = random.uniform(-1,1)\n", " #x1 = np.random.randn()\n", " #x2 = np.sqrt(1-x1**2)+0.5*np.random.randn()\n", " x = np.array([1,x1,x2])\n", " s = int(np.sign(self.V.T.dot(x)))\n", " X.append((x, s))\n", " else:\n", " for i in range(N):\n", " x = inputX[i][0]\n", " s = int(inputS[i])\n", " X.append((x,s))\n", " return X\n", " \n", " def plot(self, mispts=None, vec=None, save=False):\n", " fig = plt.figure(figsize=(5,5))\n", " plt.xlim(-1,1)\n", " plt.ylim(-1,1)\n", " V = self.V\n", " a, b = -V[1]/V[2], -V[0]/V[2]\n", " l = np.linspace(-1,1)\n", " plt.plot(l, a*l+b, 'k-')\n", " cols = {1: 'r', -1: 'b'}\n", " for x,s in self.X:\n", " plt.plot(x[1], x[2], cols[s]+'o')\n", " if mispts:\n", " for x,s in mispts:\n", " plt.plot(x[1], x[2], cols[s]+'.')\n", " if vec != None:\n", " aa, bb = -vec[1]/vec[2], -vec[0]/vec[2]\n", " plt.plot(l, aa*l+bb, 'g-', lw=2)\n", " if save:\n", " if not mispts:\n", " plt.title('N = %s' % (str(len(self.X))))\n", " else:\n", " plt.title('N = %s with %s test points' \\\n", " % (str(len(self.X)),str(len(mispts))))\n", " plt.savefig('p_N%s' % (str(len(self.X))), \\\n", " dpi=200, bbox_inches='tight')\n", " \n", " def classification_error(self, vec, pts=None):\n", " # Error defined as fraction of misclassified points\n", " if not pts:\n", " pts = self.X\n", " M = len(pts)\n", " n_mispts = 0\n", " for x,s in pts:\n", " if int(np.sign(vec.T.dot(x))) != s:\n", " n_mispts += 1\n", " error = n_mispts / float(M)\n", " return error\n", " \n", " def choose_miscl_point(self, vec):\n", " # Choose a random point among the misclassified\n", " pts = self.X\n", " mispts = []\n", " for x,s in pts:\n", " if int(np.sign(vec.T.dot(x))) != s:\n", " mispts.append((x, s))\n", " return mispts[random.randrange(0,len(mispts))]\n", " \n", " def pla(self, save=False):\n", " \"\"\"Perceptron learning algorithm\"\"\"\n", " # Initialize the weigths to zeros\n", " w = np.zeros(3)\n", " X, N = self.X, len(self.X)\n", " it = 0\n", " # Iterate until all points are correctly classified\n", " while self.classification_error(w) != 0:\n", " it += 1\n", " # Pick random misclassified point\n", " x, s = self.choose_miscl_point(w)\n", " # Update weights\n", " w += s*x\n", " if save:\n", " self.plot(vec=w)\n", " plt.title('N = %s, Iteration %s\\n' \\\n", " % (str(N),str(it)))\n", " plt.savefig('p_N%s_it%s' % (str(N),str(it)), \\\n", " dpi=200, bbox_inches='tight')\n", " self.w = w\n", " \n", " def check_error(self, M, vec):\n", " check_pts = self.generate_points(M)\n", " return self.classification_error(vec, pts=check_pts)\n", "\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "The following function is not part of perceptron but will be useful in bagging." ] }, { "cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [], "source": [ "def subsample(dataset, ratio=1.0):\n", " sample = list()\n", " n_sample = round(len(dataset) * ratio)\n", " while len(sample) < n_sample:\n", " index = randrange(len(dataset))\n", " sample.append(dataset[index])\n", " return sample" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "To apply bagging, we first bootstrap $B$ sets from training set $\\mathcal{D}$, each containing $M$ points: $\\mathcal{D}_j$ with $|\\mathcal{D}_j|=M,\\,\\forall j=1,\\cdots, B$. Then we apply PLA to each bootstrap set $\\mathcal{D}_j$ to learn the corresponding weights $\\textbf{w}_j$. The bagging prediction is made through a majority vote:\n", "$$\n", "h(\\textbf{x}_n)=\\left\\{\n", "\\begin{array}{ll}\n", " 1 & \\text{if}~ \\sum_{i=1}^B \\text{sign }(\\textbf{w}_j^T{\\tilde{\\textbf{x}_n}}) \\geq 0 \\\\\n", " -1 & \\text{otherwise} \\\\\n", "\\end{array} \n", "\\right. \n", "$$" ] }, { "cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [], "source": [ "n_training_samples = 100 \n", "n_bootstrap_samples = 25 # i.e. B\n", "bootstrap_ratio = 0.1 # i.e. M = n_training_samples*bootstrap_ratio\n", "w_blended = np.zeros([1,3])\n", "\n", "\n", "p = Perceptron(n_training_samples, boostrap_data = None)\n", "\n", "for i in range(n_bootstrap_samples):\n", " bootstrap_data = subsample(p.X, bootstrap_ratio)\n", " pb = Perceptron(int(round(n_training_samples*bootstrap_ratio)), bootstrap_data)\n", " pb.pla()\n", " w_blended = np.concatenate((w_blended, [pb.w]), axis=0)\n", "\n", "\n", "w_blended = np.delete(w_blended, 0, 0)\n", "w_bag = np.sum(w_blended, axis = 0)/float(n_bootstrap_samples) \n", "\n", "\n", "pts = p.X\n", "sall = [0]*n_training_samples\n", "\n", "for i in range(n_bootstrap_samples):\n", " vec = w_blended[i]\n", " stmp = list()\n", " for x,s in pts:\n", " stmp.append(int(np.sign(vec.T.dot(x))))\n", " sall = map(add, sall, stmp)\n", "\n", "s_bag = np.sign(np.array(list(sall))/(float(n_bootstrap_samples)))\n", "Xbag = p.generate_points(n_training_samples, pts, s_bag)\n" ] }, { "cell_type": "code", "execution_count": 12, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "fig, ax = plt.subplots(1, 1,figsize=(6,5))\n", "cols = {1: 'r', -1: 'b'}\n", "l = np.linspace(-1.5,1.5)\n", "\n", "for i in range(n_training_samples):\n", " plt.plot(Xbag[i][0][1], Xbag[i][0][2], cols[Xbag[i][1]]+'o')\n", " plt.plot(Xbag[i][0][1], Xbag[i][0][2], cols[pts[i][1]]+'x', markersize=10)\n", "\n", "for i in range(n_bootstrap_samples):\n", " aa, bb = -w_blended[i][1]/w_blended[i][2], -w_blended[i][0]/w_blended[i][2]\n", " plt.plot(l, aa*l+bb,'--', color = '0.75', lw= 0.5)\n", "\n", "cc, dd = -w_bag[1]/w_bag[2], -w_bag[0]/w_bag[2]\n", "plt.plot(l, cc*l+dd,'k--', lw= 1.5)\n", "plt.xlim([-1.5,1.5])\n", "plt.ylim([-1.5,1.5])\n", "plt.xlabel('$x_1$', multialignment='left', fontweight='bold', fontsize=15)\n", "plt.ylabel('$x_2$', multialignment='left', fontweight='bold', fontsize=15)\n", "plt.title('Bagging: o, True label: x', multialignment='left', fontsize=20)\n", "plt.show()" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Exercise: ### \n", "\n" ] }, { "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": "3.6.8" } }, "nbformat": 4, "nbformat_minor": 2 }