Delta_rule.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Delta rule.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "Shibu Mohapatra\n",
        "\n",
        "Artificial Intelligence"
      ],
      "metadata": {
        "id": "IDGVUX9hlWk6"
      }
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Jver4jDT1u1N"
      },
      "source": [
        "**Write a program to implement Delta rule.**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 267
        },
        "id": "WLwvhuJU3RsT",
        "outputId": "3aa97226-4f20-4045-c3fd-b9689c2188ac"
      },
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# the number of data points per class\n",
        "num_points = 20\n",
        "\n",
        "# data points for class -1\n",
        "data1 = np.random.multivariate_normal([1,1], [[0.8, 0.6],[0.6, 0.7]], num_points)\n",
        "# data points for class 1\n",
        "data2 = np.random.multivariate_normal([1,-1], [[0.8, 0.6],[0.6, 0.7]], num_points)\n",
        "\n",
        "# plot both arrays\n",
        "# note that arrays are indexed [row, col]\n",
        "plt.scatter(data1[:,0], data1[:,1], c='green')\n",
        "plt.scatter(data2[:,0], data2[:,1], c='blue')\n",
        "plt.show()"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CfRlNMuU3YK0",
        "outputId": "e223cc0c-2230-484e-9e87-092499ccf05f"
      },
      "source": [
        "# we create one dataset with the class attribute\n",
        "# 1) add one column with the class label\n",
        "# create the class labels for both classes\n",
        "c0 = - np.ones([num_points, 1])\n",
        "c1 = np.ones([num_points, 1])\n",
        "# append the class labels to both datasets\n",
        "data1 = np.append(data1, c0, axis=1)\n",
        "data2 = np.append(data2, c1, axis=1)\n",
        "# 2) combine the two classes in one data array\n",
        "data = np.append(data1, data2, axis=0)\n",
        "\n",
        "# we shuffle the data by rows because we want positive and negative examples to be interleaved; the algorithm below will update the weights after processing one data example\n",
        "np.random.shuffle(data)\n",
        "\n",
        "# print the first 5 rows to see what the data looks like\n",
        "print(data[:5])"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[[ 1.25417684 -1.39316656  1.        ]\n",
            " [ 1.15265869  1.04247086 -1.        ]\n",
            " [ 0.90804521 -1.55141543  1.        ]\n",
            " [ 1.01883423 -1.4530016   1.        ]\n",
            " [ 0.39017284 -2.14597111  1.        ]]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 267
        },
        "id": "GMS0EKJI3c-O",
        "outputId": "109e6cb4-4adf-46ca-f005-5ba2a176bed7"
      },
      "source": [
        "plt.scatter(data[:,0], data[:,1], c=data[:,2] )\n",
        "plt.show()"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4sYCUYrh3jR6",
        "outputId": "3aaf116c-2a29-4cdc-cb79-de61580893f0"
      },
      "source": [
        "# A coroutine that receives data and prints it\n",
        "def printer():\n",
        "    i = 0\n",
        "    line=\"\"\n",
        "    while True:\n",
        "        vec = yield\n",
        "        if vec.size == 1:\n",
        "            if i > 0:\n",
        "                # print anything that is left in the buffer\n",
        "                print(line)\n",
        "                i = 0\n",
        "                line = \"\"\n",
        "        else:\n",
        "            i = i + 1\n",
        "            line = line + \" \" + str(np.round(vec,4))\n",
        "            if i % 4 == 0:\n",
        "                print(line)\n",
        "                line = \"\"\n",
        "                i = 0\n",
        "\n",
        "printer_handler = printer()\n",
        "# start the printer\n",
        "printer_handler.send(None)\n",
        "\n",
        "# we need 3 parameters: one for x, one for y, and one for the bias term\n",
        "w = np.random.normal(0, 0.5, 3)\n",
        "# learning rate\n",
        "epsilon = 0.05\n",
        "delta = 100\n",
        "iteration = 0\n",
        "precision=0.01\n",
        "\n",
        "while delta > precision and iteration < 10:\n",
        "    iteration = iteration + 1\n",
        "    # to be consistent with the textbook, we will process one data point at a time\n",
        "    for s in data:\n",
        "        # compute the linear activation (predicted output)\n",
        "        s_with_bias = np.copy(s)\n",
        "        s_with_bias[2] = 1 # the last element is used as the bias activation\n",
        "        o_k =  w.dot(s_with_bias) # see Eq. 5.6 in O'reilly & Munakata (2000)\n",
        "        t_k = s[2]\n",
        "        # the delta rule equation in the vector form\n",
        "        delta_w = epsilon * (t_k - o_k) * s_with_bias # see Eq. 5.3\n",
        "        w = w + delta_w\n",
        "        # if the max change is tiny, we will stop learning in the while loop\n",
        "        delta=np.max(np.absolute(delta_w))\n",
        "        # print the current weights\n",
        "        printer_handler.send(w)\n",
        "\n",
        "# print the remaining values if there are any\n",
        "printer_handler.send(np.zeros([1,1]))"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            " [0.4921 0.9596 0.7947] [0.2983 0.7844 0.6266] [0.3582 0.682  0.6926] [0.4058 0.6142 0.7392]\n",
            " [0.4335 0.4618 0.8103] [0.3824 0.342  0.6881] [0.2286 0.2481 0.5663] [0.1863 0.0683 0.6303]\n",
            " [ 0.0188 -0.0743  0.5287] [-0.0444 -0.1943  0.4577] [-0.0602 -0.2284  0.3902] [-0.0025 -0.2274  0.4259]\n",
            " [ 0.0257 -0.2454  0.4426] [-0.0196 -0.2845  0.3769] [-0.0101 -0.3013  0.3889] [-0.0954 -0.3549  0.3353]\n",
            " [ 0.0952 -0.2819  0.4017] [ 0.0618 -0.2783  0.3288] [ 0.0672 -0.2945  0.3392] [ 0.0844 -0.3026  0.3655]\n",
            " [ 0.0396 -0.3703  0.3131] [-0.0186 -0.4256  0.2932] [-0.0008 -0.4386  0.3037] [-0.0201 -0.474   0.2541]\n",
            " [-0.0021 -0.4893  0.2718] [ 0.0137 -0.5024  0.2862] [ 0.0775 -0.5118  0.3103] [ 0.0774 -0.5111  0.3098]\n",
            " [ 0.1728 -0.4733  0.357 ] [ 0.0868 -0.5127  0.292 ] [ 0.037  -0.5612  0.2558] [-0.0089 -0.5987  0.22  ]\n",
            " [-0.0235 -0.6239  0.2025] [-0.002  -0.6423  0.1535] [ 0.0117 -0.6562  0.1745] [-0.0053 -0.6778  0.158 ]\n",
            " [-0.031  -0.6956  0.1441] [ 0.109  -0.5763  0.1875] [ 0.1155 -0.584   0.1953] [ 0.1148 -0.5835  0.1949]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 266
        },
        "id": "Eo7xCyF03r4P",
        "outputId": "9849a11d-7302-42c9-a146-d080c567b8ff"
      },
      "source": [
        "# plot the final line\n",
        "\n",
        "# construct the line using 100 points\n",
        "x = np.linspace(np.min(data[:,0]),np.max(data[:,0]),100)\n",
        "# this equation corresponds to our decision boundary on the xy-plane when the threshold is zero\n",
        "y = -w[0]/w[1]*x-w[2]/w[1]\n",
        "ymin = np.min(data[:,1])\n",
        "ymax = np.max(data[:,1])\n",
        "\n",
        "# plot the data and the line\n",
        "plt.ylim(ymin-0.3, ymax+0.3)\n",
        "# plot two classes separately\n",
        "plt.scatter(data1[:,0], data1[:,1], c='green', label='Class -1' )\n",
        "plt.scatter(data2[:,0], data2[:,1], c='blue', label='Class 1' )\n",
        "plt.plot(x, y, '-r', label='boundary')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXIAAAD5CAYAAAA6JL6mAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAf8klEQVR4nO3dfXRU9b3v8fc3IRDDQ0QIkWNIAlaQSkyEUG1Ry1FPa/Foay2rD8NtKUdS++AS773LZZuzqj1tas89PerqOi5teqxPZ+yTHm3tPW0vUrzQa60GioDPD4tQqA0hrREawUB+9489CZNkJpnM7Jk9e+bzWmvWzOzZ2fPNhnznN9/9ezDnHCIiEl4lQQcgIiKZUSIXEQk5JXIRkZBTIhcRCTklchGRkFMiFxEJuUmZHsDMyoEtwJTY8R5yzt001s/Mnj3b1dfXZ/rWIiJFZdu2bQedc1Ujt2ecyIGjwEXOucNmVgb8xsx+4Zx7KtkP1NfX09HR4cNbi4gUDzPrTLQ940TuvBFFh2NPy2I3jTISEckRX2rkZlZqZjuAA8BG59zvEuzTYmYdZtbR3d3tx9uKiAg+JXLn3HHnXBNQA7zHzJYk2KfdOdfsnGuuqhpV4hERkTT5USMf4px708w2A5cCuyfys/39/ezbt48jR474GVLRKS8vp6amhrKysqBDEZEc8aPXShXQH0viJwF/B/zzRI+zb98+pk+fTn19PWaWaVhFyTlHT08P+/btY/78+UGHIyI54kdpZS6w2cx2As/g1ch/PtGDHDlyhFmzZimJZ8DMmDVrlr7ViBSZjBO5c26nc+4c59zZzrklzrl/SvdYSuKZ0zkUSU90V5T62+sp+VoJ9bfXE90VDTqklGlkp4gUveiuKC2PtdDZ24nD0dnbSctjLb4m82x+UCiRx9mzZw9LlozqcJM1N998M9/+9rdz9n4ikljrplb6+vuGbevr76N1U6svx8/2B4USeYgcO3Ys6BBECtLe3r0T2j5R2f6gUCIf4dixY0QiERYvXszHPvYx+vr62LRpE+eccw4NDQ2sW7eOo0ePAt5UAwcPHgSgo6ODlStXAl5Le926daxcuZIFCxbwne98Z+j4bW1tLFy4kPPPP5+XXnppaPv3vvc9li9fTmNjI1dddRV9fd4/+tq1a7nmmms499xzueGGGzjjjDMYHFA1MDDAu971LjTASiQztZW1E9o+Udn+oPC1H7lvNmyAHTv8PWZTE9x++7i7vfTSS9x9992sWLGCdevWceutt/Ld736XTZs2sXDhQj796U9z5513smHDhjGP8+KLL7J582YOHTrEokWL+PznP8/OnTv54Q9/yI4dOzh27BhLly5l2bJlAHz0ox9l/fr1APzjP/4jd999N9deey3gdc188sknKS0tpbKykmg0yoYNG3j88cdpbGxEA6xEMtN2cRstj7UMazVXlFXQdnGbL8evrayls3f0NCl+fVCoRT7CvHnzWLFiBQBr1qxh06ZNzJ8/n4ULFwLwmc98hi1btox7nMsuu4wpU6Ywe/Zs5syZQ1dXF1u3buXKK6+koqKCGTNmcMUVVwztv3v3bi644AIaGhqIRqM899xzQ6+tXr2a0tJSANatW8f9998PwPe//30++9nP+va7ixSrSEOE9svbqauswzDqKutov7ydSEPEl+O3XdxGRVnFsG1+flDkZ4s8hZZztozsvnfyySfT09OTcN9JkyYxMDAAMKrv9pQpU4Yel5aWjlvfXrt2LY8++iiNjY3ce++9PPHEE0OvTZ06dejxvHnzqK6u5te//jVPP/000Wh4ukiJ5LNIQ8S3xJ3o2ODVyvf27qW2spa2i9t8ez+1yEfYu3cvv/3tbwF48MEHaW5uZs+ePbz66qsAPPDAA7z//e8HvBr5tm3bAHj44YfHPfaFF17Io48+yttvv82hQ4d47LHHhl47dOgQc+fOpb+/f9zkfPXVV7NmzZphLXURyW+Rhgh7Nuxh4KYB9mzY4+uHhhL5CIsWLeKOO+5g8eLF/OUvf+H666/nnnvuYfXq1TQ0NFBSUsI111wDwE033cR1111Hc3NzSgl16dKlfPzjH6exsZEPfehDLF++fOi1r3/965x77rmsWLGCM888c8zjXHHFFRw+fFhlFREBwLzpxHOrubnZjVxY4oUXXmDx4sU5jyWMOjo6uP7669m6dWvC13UuRQqTmW1zzjWP3J6fNXJJ6lvf+hZ33nmnauMiMkSllZC58cYb6ezs5Pzzzw86FBHJE0rkIiIhp0QuIhJySuQiIiGnRC4iEnJK5HH+9Kc/8YlPfILTTz+dZcuWsWrVKl5++eWcTm/b2trKvHnzmDZtWk7eT0TCT4k8xjnHlVdeycqVK3nttdfYtm0bt9xyC11dXTmN4/LLL+fpp5/O6XuKSLiFNpH7vdrG5s2bKSsrGxq1CdDY2MgFF1wwbL89e/ZwwQUXsHTpUpYuXcqTTz4JwBtvvMGFF15IU1MTS5YsYevWrRw/fpy1a9eyZMkSGhoauO2228aN47zzzmPu3LkZ/S4iUlxCOSBocLWNwSknB1fbANKev2D37t1DU8qOZc6cOWzcuJHy8nJeeeUVPvnJT9LR0cGDDz7IBz/4QVpbWzl+/Dh9fX3s2LGD/fv3s3v3bgDefPPNtGITERlLKFvk2V5tYyz9/f2sX7+ehoYGVq9ezfPPPw/A8uXLueeee7j55pvZtWsX06dPZ8GCBbz++utce+21/PKXv2TGjBlZj09Eik8oE3k2Vts466yzhmYyHMttt91GdXU1zz77LB0dHbzzzjuAN7Phli1bOO2001i7di33338/M2fO5Nlnn2XlypXcddddXH311cOOdfz4cZqammhqauKrX/1q2rGLSHELZWklG6ttXHTRRXzlK1+hvb2dlhavTLNz5056e3uZN2/e0H69vb3U1NRQUlLCfffdx/HjxwHo7OykpqaG9evXc/ToUbZv386qVauYPHkyV111FYsWLWLNmjXD3rO0tJQdfq+EJCJFJ5Qt8mystmFmPPLIIzz++OOcfvrpnHXWWXz5y1/m1FNPHbbfF77wBe677z4aGxt58cUXhxZ9eOKJJ2hsbOScc87hRz/6Eddddx379+9n5cqVNDU1sWbNGm655ZZx47jhhhuoqamhr6+Pmpoabr755rR/JxEpDqGdxja6K5q11TbCTtPYihSmgpvGNpvLMomIhEnGpRUzm2dmm83seTN7zsyu8yMwERFJjR8t8mPA/3DObTez6cA2M9vonHveh2OLiMg4Mm6RO+fecM5tjz0+BLwAnJbpcUVEJDW+9loxs3rgHOB3CV5rMbMOM+vo7u72821FRIqab4nczKYBDwMbnHNvjXzdOdfunGt2zjVXVVX59bYiIkXPl0RuZmV4STzqnPtPP44ZhCCmsd2yZQtLly5l0qRJPPTQQ1l5DxEpbH70WjHgbuAF59ytmYcUjKCmsa2treXee+/lU5/6VFbfR0QKlx8t8hXAfwMuMrMdsdsqH447pmgU6uuhpMS7j2Y2i21g09jW19dz9tlnU1ISykG2IpIHMu5+6Jz7DWA+xJKyaBRaWqAvNgFiZ6f3HCCS5hghTWMrImEVypGdra0nkvigvj5ve7qJPFX9/f186UtfYseOHZSWlvLyyy8D3jS269ato7+/n4985CM0NTUNm8b2sssu4wMf+EB2gxORohTK7/N7k8xWm2x7KoKYxlZExA+hTOS1SWarTbY9FRdddBFHjx6lvb19aNvOnTvZunXrsP16e3uZO3cuJSUlPPDAA8Omsa2urmb9+vVcffXVbN++nYMHDzIwMMBVV13FN77xDbZv355+gEXC7yX8RIpBKBN5WxtUDJ/FlooKb3u6gprG9plnnqGmpoaf/OQnfO5zn+Oss85K/5cIucEl/Dp7O3G4oSX8lMxFxhbeaWyjXk18716vJd7Wlv36eFiEdRrb+tvrEy4YUldZx54Ne3IfkEieKbxpbCNK3IUmG0v4iRSDUJZWpDAlW6ovkyX8RIpBXiXyIMo8hSbM5zAbS/iJFIO8SeTl5eX09PSEOhEFzTlHT08P5eXlQYeSlkhDhPbL26mrrMMw6irraL+8XStBiYwjby529vf3s2/fPo4cOZLzeApJeXk5NTU1lJWVBR2KiPgs7y92lpWVMX/+/KDDEBEJnbwprUjx0eAfEX8okUsgNPhHH2TiHyVyCUTrplb6+ofPfNbX30frptaAIsotfZCJn5TIJRDFPvin2D/IxF9K5BKIMA3+yUYJpNg/yMRfSuQSiESDfwAOv3M45USZixpztkogYfogk/ynRC6BGBz8M+ukWcO297zdk1KizFWNOVslED9GsepiqQxSIpfARBoiTJs8bdT2VBJlrmrM2SqBZDqKVRdLJV7eDAiS4pRuosxVjbm2sjbh1Lp+lEAiDZG0px8Y64NMUxoUH7XIJVDp1opzVWPO14m8dLFU4imRS6DSTZS5SrD5OpGXLpZKPJVWJFCDCbF1Uyt7e/dSW1lL28Vt4ybKdH8u3RiDTtwjtV3cRstjLcPKK/nwTUGCkTezH4rIxER3RXPyQVaowrhcZLLZD5XIRaToRKPQ0gJ9cdeLKyqgvT2/k3myRK4auRS0VPtaq092cWltHZ7EwXveGtIZEnypkZvZ94G/Bw4455b4cUyRTA32tR6sIw/2tQaGlSBS3U8Kx94knXuSbc93frXI7wUu9elYkiOF3gpNddBQrgYXFfr5DpPaJJ17km3Pd74kcufcFuDPfhxLcqMYRgam2tc6F32yi+F8h0lbm1cTj1dR4W0PI9XIi1QxTKOaal/rXPTJLobzHSaRiHdhs64OzLz7rF/odA4OHYJ33vH90DnrR25mLUALQG1Yv78UkGIYGZhqX+tc9MkuhvMdNpGID4l7YAD+/Gfo6vJuBw6cuB98HL/97bdh40a45BJffodBOUvkzrl2oB287oe5el9JLJtziOSLVAcN5WJwUTGc74LxzjvDE3KiJD14390Nx4+PPkZpKcyZ492qq2HhQu9+zhxYsMD3kH3rR25m9cDPU+m1on7kwRvZUwO8Vmg+DD8vRDrfAXIODh9OnIwTPX7zzcTHOekkLxkP3uIT9eD94ONTToES/yvXyfqR+9X98AfASmC2me0DbnLO3e3HsSU7cjnEXXS+fTcwAD09qSXmri44ciTxcWbOPJF8GxtHJ+T4x9NGT7mcLzSyU6SI5PWw9KNHvVJFspZzfHLu7vaS+UgjSxqJWsyDz6uqYPLk3P+eGchqi1wkzIplzpKRw9I7O73nkKVkPthLI1l9eeR9spLG1KknEvH8+XDeecOTdXxynjkzKyWNfKcWuRS1Yqpd19d7yXukujrYsyfFgxw/PrykMV5yTlbSOOWU0SWM+Pv4x3lc0sg1TZolQPG0PlNVf3t9wt4kdZV17NmwJ/cBZVFJiddIHmkKRzmyN4UWc1dX8pLGpEmJE3Gi51VVUFaW/V+4AKm0IppTJIGC7N8dX9KIazX/64wDTOntopou5nBg6P5keiFRL8ipU08k30QljfjkPHOmN7JGAqFEXkS0zuNooenfPVjSGKu1HP/46NFRh7geOMgsDjCHLqr5PefQM6ma911ZTdMHEpQ0pk7N/e8paVEiLyIF2frMULJRnavOWEX97fXZLUEdOZLaRcCuLjh4MHlJI7588e53J+46F+ul8asfl43qtdJUnJ/hBUWJvIjkovUZthp8ov7dq85YxX3P3jfxEpRz8NZbiYdoJ0rOb72V+DjxJY3TT4f3vjd5/+aTT55QScOXYemSd3Sxs4hku4dGofQAib8AWjIAs/ug+jA0MIfo+bcmL2kcOJCwpIEZzJqVfBTgyIuEKmlIEuq1IkB2W8yh6QESX9JI0Fre9NQPmPNXL3nP7ksyRWhZWfLEnKCkwSR9+ZXMqddKgZpoYs7mivCB1eCdg97e1Esahw4lPs60aVBdTaWbwiunHOX/zYMDU71b1zSw6mp+/MUn0ippiGSTEnmI5Vt3Ql9r8MeOneilMbJ8kSg5J5rjeWRJo7l57PJGbKWBl5KWiP4Vzjxz4r+LSJYpkYdYvnUnHHde77ffHntyo/jkfPBg4tErkycPT8ZLliQfITh7dlolDU1wJWGjGnmIlXytBMfofz/DGLgpQVc1vznnzY8Rl4Cf+f3PebLjUcp7eql75ySWldZQdWjA2ydZSWP69OQjAkc+rqxUSUOKlmrkBSgr3QmPHfNaw+MNOBlsQY8oaSwHlpvB7KrYhb5qWJxkFrrBxH3SSenHKyLFlcjD1sd5PCkvUdbXl9powAMHvLr0eCWNU0+Fs88eu6RRWprl315EBhVNIs+3C4MZiZU0IpObmTP3v/PI/72L0gMHedfxSlbNWMoZzzwEXXecSM6HDyc+zowZJxLwokVw4YXJSxozZqikIZKniqZGnvd9nPv7vZLGeCudDJY0+vtHH8PM67M83oCTwccqaYiEStHXyAPp49zXl7hXRqIk3dOT+BiTJ59Iyn/zN9DUNHw60PiErZKGSFEqmkTuy4VB5+Avf0k84CRRDXq8kkZ1NSxeDCtXJu+xoZKGiIyjaBJ5sguD37zwa/DHP44/+9zg42PHRh+8pMRrDQ8m4PPOG3vFk/LyHP7mIlLoCjOR//Wvo5JxpKuLZS+cz6vP/4apb/ZR83YZtUdKmNK6NvExpkxJXtIYmaRnzVJJQ0QCE65E/oc/wGuvjb9e4F//mvDHz5wxgzOrT4Oa6sQXA+MT9PTpKmmISCiEK5F/85tw110nnpeUDO+lcfrpY/fYmDIluNhFRLIkXIn8i1+E1auHlzRKEk4yKiJSNMKVyJcsCToCEZG8o+asiEjIKZGL5Fg0CvX1XlWwvt57LpIJXxK5mV1qZi+Z2atmdqMfxxQpRNEotLRAZ6c3vqyz03uuZC6ZyDiRm1kpcAfwIeDdwCfN7N2ZHldyK7orSv3t9ZR8rYT62+uJ7lJmyYbWVm/mhnh9fd52kXT5cbHzPcCrzrnXAczsh8CHged9OLbkQEHNDJnn9iaZ2ifZdpFU+FFaOQ34Q9zzfbFtw5hZi5l1mFlHd3e3D28rfhlryTjxV22SqX2SbRdJRc4udjrn2p1zzc655qqqqly9raQgkJkhi1Rb29Aaz0MqKrztIunyI5HvB+bFPa+JbZOQSDYDZEZLxklCkQi0t0NdnTcDRF2d9zyiCpZkwI9E/gxwhpnNN7PJwCeAn/lwXMmRtovbqCgb3kxMuGSc+CISgT17YGDAu1cSl0xlnMidc8eALwG/Al4Afuycey7T40ruRBoitF/eTl1lHYZRV1lH++XtutApEhJFs9SbiEjYJVvqTSM7RURCTolcRCTklMglY5o7RCRY4ZrGVvLO4Nwhg8POB+cOAfXGEMkVtcglI5o7RHJN3wBHU4tcMqK5QySX9A0wMbXIJSOaO0RySd8AE1Mil4xo7hDJJX0DTEyJXDKiuUMkl/QNMDElcsmY5g6RXNE3wMSUyEUkNPQNMDH1WhGRUIlElLhHUotcRCTklMhFREJOiVwyppF2IsFSjVwyopF2IsFTi1wyopF2IsFTIpeMaKSdSPCUyCUjGmknEjwlcsmIRtqJBE+JXDKikXYiwVMil4zleq4VdXcUGU6JXEJlsLtjZyc4d6K742AyV5KXYqR+5BIq43V3VJ92KUZqkUuojNXdUX3apVgpkUugJloKGau7o/q0S7HKKJGb2Woze87MBsys2a+gpDiMV+9OZKzujsXcp13XBopbpi3y3cBHgS0+xCJFJp1SyFjdHYu1T3s6H4hSWMw5l/lBzJ4A/qdzriOV/Zubm11HR0q7SgErKfESz0hmXlfGdESj3gfB3r1eS7ytrfAvdNbXe8l7pLo6rzuoFA4z2+acG1X9yFmN3MxazKzDzDq6u7tz9baSx7JRCgnr+qGZlEZ0bUDGTeRm9riZ7U5w+/BE3sg51+6ca3bONVdVVaUfsRSMYi2FjJRpaaSYrw2IZ9xE7py7xDm3JMHtp7kIUAqXhvd7Mu02qQ9E0YAgCZQW0s28NDJ4/ort2oCckGn3wyvNbB/wXuB/m9mv/AlLpHj4URoJ67UB8UdGidw594hzrsY5N8U5V+2c+6BfgYkUi0SlEYDDh9WFUFKjkZ0iARu8VjBr1vDtPT3qDy6pUSIXyQORCEybNnq75oqRVCiRFwgN0Q4/9QeXdCmRFwAN0S4M6g8u6VIiLwCavrUwqD+4pEuJvACE5St5PpZ/8ikmDZCSdGlAUAGorU08aVI+fSUfLP/k0+o9+RiTBkhJOtQiLwBh+Erud/nHj5a0SlJSKJTIC0AYvpL7Wf7x6+JuWEpSIuPxZT7yidJ85MXHzzmz/TqW5vGWsAl8PnIJF78vAk6k/DPee/vVkg5DSUokJc65nN+WLVvmJH/9x384V1HhnFe48G4VFd72TI9bV+ecmXef6HipvHdd3fDXB291ddmJSSRfAB0uQU5VaUVGCbLkkMp7j+xtAl5LOt+uC4j4TaUVSVmQFwFTee8wXNwVySX1I5dRguyXnup7q7+1yAlqkcsoQV4E1AVIkYlTIpdRgixdqGwiMnG62CkiEhK62CkiUqCUyEVEQk6JXEQk5JTIRURCTolcRCTklMhFREJOiVxEJOSUyEVEQi6jRG5m/2JmL5rZTjN7xMxO9iswERFJTaYt8o3AEufc2cDLwJczD0lERCYio0TunPs/zrljsadPATWZhyQiIhPhZ418HfCLZC+aWYuZdZhZR3d3t49vKyJS3Madj9zMHgdOTfBSq3Pup7F9WoFjQNKVHZ1z7UA7eJNmpRWtiIiMMm6L3Dl3iXNuSYLbYBJfC/w9EHFBTKUoE+L3osoiEryMVggys0uBG4D3O+f6xttfgjVyrcvOTu85aL5vkTDLtEb+b8B0YKOZ7TCzu3yISbKktXX4gsXgPW9tDSYeEfFHRi1y59y7/ApEsi/IRZVFJHs0srOIJFs8OReLKotI9iiRFxEtbCxSmJTIi4gWNhYpTBnVyCV8IhElbpFCoxa5iEjIKZGLiIScErmISMgpkYuIhJwSuYhIyCmRi4iEnBK5iEjIKZGLiIScErmISMgpkYuIhJwSuYhIyCmRi4iEnBK5iEjIKZGLiIScErmISMiFJpFHo1BfDyUl3n00GnREIiL5IRQLS0Sj0NJyYgX4zk7vOWiRBBGRULTIW1tPJPFBfX3edhGRYheKRL5378S2i4gUk1Ak8traiW0XESkmoUjkbW1QUTF8W0WFt11EpNiFIpFHItDeDnV1YObdt7frQqeICGTYa8XMvg58GBgADgBrnXN/9COwkSIRJW4RkUQybZH/i3PubOdcE/Bz4Ks+xCQiIhOQUSJ3zr0V93Qq4DILR0REJirjAUFm1gZ8GugF/naM/VqAFoBadTcREfGNOTd2I9rMHgdOTfBSq3Pup3H7fRkod87dNN6bNjc3u46OjonGKiJS1Mxsm3OueeT2cVvkzrlLUnyPKPBfwLiJXERE/JNRjdzMzoh7+mHgxczCERGRiRq3tDLmD5s9DCzC637YCVzjnNufws91x/bPldnAwRy+30QotvQotvTlc3yKbWx1zrmqkRszSuRhYWYdiepK+UCxpUexpS+f41Ns6QnFyE4REUlOiVxEJOSKJZG3Bx3AGBRbehRb+vI5PsWWhqKokYuIFLJiaZGLiBQsJXIRkZAryERuZqeY2UYzeyV2PzPJfsfNbEfs9rMsx3Spmb1kZq+a2Y0JXp9iZj+Kvf47M6vPZjwTjG2tmXXHnaurcxTX983sgJntTvK6mdl3YnHvNLOluYgrxdhWmllv3DnL2cygZjbPzDab2fNm9pyZXZdgn0DOXYqxBXnuys3saTN7Nhbf1xLsE9jfalLOuYK7Af8LuDH2+Ebgn5PsdzhH8ZQCrwELgMnAs8C7R+zzBeCu2ONPAD/Ko9jWAv8WwL/jhcBSYHeS11cBvwAMOA/4XR7FthL4ea7PWey95wJLY4+nAy8n+DcN5NylGFuQ586AabHHZcDvgPNG7BPI3+pYt4JskeNNF3Bf7PF9wEcCjAXgPcCrzrnXnXPvAD/EizFefMwPARebmeVJbIFwzm0B/jzGLh8G7neep4CTzWxunsQWGOfcG8657bHHh4AXgNNG7BbIuUsxtsDEzsfh2NOy2G1kj5Cg/laTKtREXu2ceyP2+E9AdZL9ys2sw8yeMrNsJvvTgD/EPd/H6P+8Q/s4547hTQs8K4sxTSQ2gKtiX8EfMrN5OYgrFanGHpT3xr6i/8LMzgoigNjX/nPwWpbxAj93Y8QGAZ47Mys1sx14q55tdM4lPXc5/ltNKuP5yIMy1vS68U+cc87MkvWxrHPO7TezBcCvzWyXc+41v2MtAI8BP3DOHTWzz+G1Ri4KOKZ8tx3v/9dhM1sFPAqcMc7P+MrMpgEPAxvc8EVgAjdObIGeO+fccaDJzE4GHjGzJc65hNdC8kVoW+TOuUucc0sS3H4KdA1+TYzdH0hyjP2x+9eBJ/BaB9mwH4hvxdbEtiXcx8wmAZVAT5bimVBszrke59zR2NN/B5blIK5UpHJeA+Gce2vwK7pz7r+AMjObnav3N7MyvEQZdc79Z4JdAjt348UW9LmLi+NNYDNw6YiXgvpbTSq0iXwcPwM+E3v8GeCnI3cws5lmNiX2eDawAng+S/E8A5xhZvPNbDLeBZKRvWTiY/4Y8GsXu5qSZePGNqJ2egVeXTMf/Az4dKwHxnlAb1xJLVBmdupg3dTM3oP3t5aTP/bY+94NvOCcuzXJboGcu1RiC/jcVcVa4pjZScDfMXp67qD+VpML+mprNm549apNwCvA48Apse3NwL/HHr8P2IXXS2MX8A9ZjmkV3hX61/BWVwL4J+CK2ONy4CfAq8DTwIIcnq/xYrsFeC52rjYDZ+Yorh8AbwD9eDXcfwCuwZsuGbweBnfE4t4FNOfwnI0X25fiztlTwPtyGNv5eBfodgI7YrdV+XDuUowtyHN3NvD7WHy7ga/GtufF32qym4boi4iEXKGWVkREioYSuYhIyCmRi4iEnBK5iEjIKZGLiIScErmISMgpkYuIhNz/B+kHoLn/ur2QAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hCFMFN8r30FU",
        "outputId": "fceebe98-d866-47bc-c468-b748aa633655"
      },
      "source": [
        "# we need to add bias activations in the last column\n",
        "data_with_bias = np.copy(data)\n",
        "data_with_bias[:,2]=1\n",
        "\n",
        "# predictions are the outputs computed by the linear model\n",
        "predictions = w.dot(data_with_bias.transpose())\n",
        "\n",
        "# compute the summed squared error with respect to the linear activation; i.e., we want to see see how well the linear function approximates the values of our class attribute\n",
        "SSE = np.sum(np.square(predictions-data[:,2]))\n",
        "\n",
        "# this is the same threshold that we used to derive the red decision boundary above; with this threshold, our classification below will be consistent with the red line above\n",
        "threshold = 0\n",
        "\n",
        "# apply threshold to make discrete decisions\n",
        "predictions[predictions<threshold] = -1\n",
        "predictions[predictions>=threshold] = 1\n",
        "# print(predictions)\n",
        "# print(data[:,2])\n",
        "\n",
        "# diff will be zero when the classification is correct\n",
        "diff = predictions - data[:,2]\n",
        "# indices of data items with incorrect predictions\n",
        "erroneous = diff.nonzero()[0]\n",
        "# count the number of incorrect predictions\n",
        "accuracy = np.size(erroneous)\n",
        "\n",
        "print(\"SSE on training data: \" + str(SSE) )\n",
        "print(\"Number of errors on training data: \" + str(accuracy) )"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "SSE on training data: 15.027952856480772\n",
            "Number of errors on training data: 3\n"
          ]
        }
      ]
    }
  ]
}