figure6.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "72a27a16",
   "metadata": {},
   "source": [
    "# Import libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "f031f752",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from dirrac.data.synthetic_data import DataSynthesizer\n",
    "from dirrac.classifier.logistic import logistic_classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "209ea109",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Matplotlib config\n",
    "SMALL_SIZE = 8\n",
    "MEDIUM_SIZE = 10\n",
    "BIGGER_SIZE = 13\n",
    "\n",
    "plt.rc('font', size=BIGGER_SIZE)          # controls default text sizes\n",
    "plt.rc('axes', titlesize=BIGGER_SIZE)     # fontsize of the axes title\n",
    "plt.rc('axes', labelsize=BIGGER_SIZE)    # fontsize of the x and y labels\n",
    "plt.rc('xtick', labelsize=BIGGER_SIZE)    # fontsize of the tick labels\n",
    "plt.rc('ytick', labelsize=BIGGER_SIZE)    # fontsize of the tick labels\n",
    "plt.rc('legend', fontsize=MEDIUM_SIZE)    # legend fontsize"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cb6a6b67",
   "metadata": {},
   "source": [
    "# Synthesize data and train classifiers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "b8c0ebac",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Original data parameters\n",
    "mean_0 = np.ones(2) * (-3)\n",
    "mean_1 = np.ones(2) * 3\n",
    "cov_0 = cov_1 = np.identity(2)\n",
    "n = 1000 # 1000 samples, 500 each class\n",
    "\n",
    "\"\"\" Here we synthesize original data, \n",
    "and 100 data distribution shifts \n",
    "with 33 mean shifts, 33 covariance shifts and 34 mean and covariance shift\n",
    "Shifted parameters \\alpha=[1.5, 0] and \\beta=3 \"\"\"\n",
    "sd = DataSynthesizer(mean_0, cov_0, mean_1, cov_1, n)\n",
    "features, labels = sd.synthesize_modes_data(100, [0.33, 0.33, 0.34], [0.1, 0.1], same=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "c5b9e79a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Train 101 classifiers, the first is on the original dataset and the others are on shifted data\n",
    "all_coef = np.zeros((101, 2))\n",
    "for i in range(101):\n",
    "    coef = logistic_classifier(features[i], labels[i])[1].T\n",
    "    all_coef[i] = np.squeeze(coef)\n",
    "\n",
    "# Get theta with 3 types of distribution shifts\n",
    "mean_shift = all_coef[1:34]\n",
    "cov_shift = all_coef[34:67]\n",
    "both_shift = all_coef[67:101]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "4a983835",
   "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\n",
    "fig, ax = plt.subplots()\n",
    "ax.scatter(all_coef[0][0], all_coef[0][1], s=100, marker='*', label = 'Original')\n",
    "ax.scatter(mean_shift[:, 0], mean_shift[:, 1], marker='o', label = 'Mean shift')\n",
    "ax.scatter(cov_shift[:, 0], cov_shift[:, 1], marker='o', label = 'Cov shift')\n",
    "ax.scatter(both_shift[:, 0], both_shift[:, 1], marker='o', label = 'Both shift')\n",
    "\n",
    "ax.set(xlabel='$\\\\theta_{0}$', ylabel='$\\\\theta_{1}$')\n",
    "ax.legend(loc='upper right', frameon=False)\n",
    "\n",
    "ax.set_xlim(0.5, 2)\n",
    "ax.set_ylim(1, 2.75)\n",
    "plt.tight_layout()\n",
    "\n",
    "if not os.path.exists('result/figure6/'):\n",
    "    os.makedirs('result/figure6/')\n",
    "plt.savefig('result/figure6/classifiers.pdf', dpi=500, transparent=True)\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}