2_genre_classification_classical_and_rock.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Two genre classification: classical vs rock music\n",
    "\n",
    "In this notebook, we attempt to train a model to classify if a given song is classical or rock. The 2 song attributes used to perform this classification are danceability and energy. We use the gradient descent optimization algorithm with multiclass perceptron for this task. Of the 100 songs, or points of testing data, we get an output of 0 misclassifications."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import autograd.numpy as np\n",
    "from autograd import grad \n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.gridspec as gridspec\n",
    "from common import gradient_descent, plot_cost_histories"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Read data\n",
    "DATA_BASE_URL = \"https://raw.githubusercontent.com/sql-injection/spotify_data/master/\"\n",
    "datasets = {\n",
    "    \"train\": DATA_BASE_URL + \"train.csv\",\n",
    "    \"test\": DATA_BASE_URL + \"test.csv\",\n",
    "    \"all\": DATA_BASE_URL + \"spotify.csv\"\n",
    "}\n",
    "\n",
    "total_df = pd.read_csv(datasets[\"all\"])\n",
    "attribute_names = list(total_df)[:-1]\n",
    "x = total_df[attribute_names].values\n",
    "y = total_df[\"Class\"].values\n",
    "\n",
    "original_x = np.copy(x)\n",
    "original_y = np.copy(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "genre_histograms = {\n",
    "    \"edm\":       np.asarray([1, 0, 0, 0, 0, 0, 0]),\n",
    "    \"hiphop\":    np.asarray([0, 1, 0, 0, 0, 0, 0]),\n",
    "    \"classical\": np.asarray([0, 0, 1, 0, 0, 0, 0]),\n",
    "    \"jazz\":      np.asarray([0, 0, 0, 1, 0, 0, 0]),\n",
    "    \"pop\":       np.asarray([0, 0, 0, 0, 1, 0, 0]),\n",
    "    \"country\":   np.asarray([0, 0, 0, 0, 0, 1, 0]),\n",
    "    \"rock\":      np.asarray([0, 0, 0, 0, 0, 0, 1])\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Genre encoded into one-hot vectors\n",
    "y_encoded = np.array([genre_histograms[genre] for genre in y])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "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": [
    "feature_a_indices =  np.where(y == \"classical\")[0]\n",
    "feature_b_indices = np.where(y == \"rock\")[0]\n",
    "feature_a_color = \"cornflowerblue\";\n",
    "feature_b_color = \"salmon\";\n",
    "\n",
    "attribute_names = np.array(attribute_names)\n",
    "energy_attribute_index = np.where(attribute_names == \"energy\")[0][0]\n",
    "danceability_attribute_index = np.where(attribute_names == \"danceability\")[0][0]\n",
    "\n",
    "f = plt.figure()\n",
    "ax = plt.subplot()\n",
    "ax.set_xlabel(\"Energy\")\n",
    "ax.set_ylabel(\"Danceability\")\n",
    "ax.set_title(\"Classical Music (Blue) vs. Rock (Red) in terms of Energy and Danceability\")\n",
    "for i in feature_a_indices:\n",
    "    ax.scatter(original_x[i, energy_attribute_index],\n",
    "               original_x[i, danceability_attribute_index],\n",
    "               color=feature_a_color, edgecolor='k')\n",
    "    \n",
    "for i in feature_b_indices:\n",
    "    ax.scatter(original_x[i, energy_attribute_index],\n",
    "               original_x[i, danceability_attribute_index],\n",
    "               color=feature_b_color, edgecolor='k')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = original_x[:, :2]\n",
    "included_indices = np.concatenate((feature_a_indices, feature_b_indices))\n",
    "X = np.take(X, included_indices, axis=0)\n",
    "Y = np.take(original_y, included_indices, axis=0)\n",
    "Y[Y == \"classical\"] = 0\n",
    "Y[Y == \"rock\"] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = X\n",
    "y = Y[:, np.newaxis]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model(x, w):\n",
    "    a = w[0] + np.dot(x.T, w[1:])\n",
    "    return a.T\n",
    "\n",
    "lam = 10**-5  # our regularization paramter \n",
    "def multiclass_perceptron(w):        \n",
    "    # pre-compute predictions on all points\n",
    "    all_evals = model(x.T, w)\n",
    "    \n",
    "    # compute maximum across data points\n",
    "    a = np.max(all_evals,axis = 0)    \n",
    "\n",
    "    # compute cost in compact form using numpy broadcasting\n",
    "    b = all_evals[y.astype(int).flatten(),np.arange(np.size(y))]\n",
    "    cost = np.sum(a - b)\n",
    "    \n",
    "    # add regularizer\n",
    "    cost = cost + lam*np.linalg.norm(w[1:,:],'fro')**2\n",
    "    \n",
    "    # return average\n",
    "    return cost/float(np.size(y))\n",
    "                              "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "w = 0.1 * np.random.randn(3, 70)\n",
    "learning_rate = 0.1\n",
    "g = multiclass_perceptron\n",
    "max_its = 200"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "weight_history, cost_history = gradient_descent(g, learning_rate, max_its, w)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "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": [
    "plot_cost_histories(cost_histories=[cost_history], labels=[\"Cost history 1\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "best_idx = np.argmin(cost_history)\n",
    "best_weight = weight_history[best_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "def fusion_rule(x, w):\n",
    "    return np.argmax(model(x, w))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "def num_misclassifications(best_weight):\n",
    "    points_count = len(x)\n",
    "\n",
    "    miscounts = 0\n",
    "    for p in range(points_count):\n",
    "        x_val = np.array([x[p, 0], x[p, 1]])\n",
    "        if fusion_rule(x_val, best_weight) != y[p]:\n",
    "            miscounts = miscounts + 1\n",
    "    \n",
    "    return miscounts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Out of 100 points, the number of misclassifications is 0\n"
     ]
    }
   ],
   "source": [
    "print(\"Out of {0} points, the number of misclassifications is {1}\".format(len(x), num_misclassifications(best_weight)))"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}