1NN_estimation_rate.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import ot\n",
    "import numpy as np\n",
    "import time\n",
    "\n",
    "import pickle\n",
    "import matplotlib.pyplot as plt\n",
    "import statsmodels.api as sm\n",
    "\n",
    "from data_gen import *\n",
    "from NN_map import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def nn_estimation_rate(sampling_function,mapfnt,Ns,Ds,data,numTrials=10,N_sampling=10000,title=None):  \n",
    "    L2_NN_Error = np.zeros( (len(Ds), len(Ns), numTrials))\n",
    "    tstart = time.time()\n",
    "    for k, d in enumerate(Ds):\n",
    "        for i, n in enumerate(Ns):\n",
    "            for t in range(numTrials):\n",
    "                \n",
    "                ### generate ``training\" data \n",
    "                source = sampling_function(n,d)\n",
    "                source_ = sampling_function(n,d)\n",
    "                target = mapfnt(source_)\n",
    "                \n",
    "                ### generate \"L^2(P) estimation data\"\n",
    "                source_mc  = sampling_function(N_sampling,d)\n",
    "                ot_t = mapfnt(source_mc)\n",
    "\n",
    "                G0 = OT_0(source,target)\n",
    "                tnn = NNEstimator(source_mc,source,target,G0)\n",
    "\n",
    "                L2_NN_Error[k,i,t] = (np.linalg.norm( (tnn - ot_t),axis=1)**2).mean()\n",
    "\n",
    "                if (t+1)%numTrials == 0:\n",
    "                    print('d=%f, n=%d, trials: %d/%d, time=%d'%(d, n,t+1, numTrials, time.time()-tstart))\n",
    "        print('==== Done %f d level ===='%(d))\n",
    "\n",
    "    dict_data = {'Ds': Ds,\n",
    "                 'N_sampling': N_sampling,\n",
    "                 'Ns': Ns,\n",
    "                 'numTrials': numTrials,\n",
    "                 'L2_NN_Error': L2_NN_Error,\n",
    "                 'data': data}\n",
    "\n",
    "    pkl_title = 'CPU-{}_error.pkl'.format(data)\n",
    "    output = open(pkl_title,'wb')\n",
    "    pickle.dump(dict_data,output)\n",
    "    output.close()\n",
    "    \n",
    "def CreatePlot_CPU_Errors(file,save=False):\n",
    "    with open(file, 'rb') as f:\n",
    "        file1 = pickle.load(f)\n",
    "\n",
    "    l2_nn = file1['L2_NN_Error']\n",
    "\n",
    "    Ds = file1['Ds']\n",
    "    Ns = file1['Ns']\n",
    "    data = file1['data']\n",
    "    \n",
    "    plt.figure(figsize=([6,5]))\n",
    "    cz = ['r', 'b', 'g']\n",
    "    print('Estimation rates for 1NN estimator')\n",
    "    print('Note: theoretical rate is (-2/d)')\n",
    "    for i,d in enumerate(Ds):\n",
    "\n",
    "        c = cz[i]\n",
    "        l2nn_d = l2_nn[i]\n",
    "\n",
    "        ynn_means = np.mean(l2nn_d,axis=-1)\n",
    "        ynn_std = np.std(l2nn_d,axis=-1)\n",
    "\n",
    "        x_ = sm.add_constant(np.log(np.array(Ns)))\n",
    "        modelnn = sm.OLS(np.log(ynn_means), x_).fit()\n",
    "        print('d={}, rate={}'.format(d,modelnn.params[1]))\n",
    "\n",
    "        plt.loglog(Ns,ynn_means,label='d={}'.format(d),c=c)\n",
    "        plt.errorbar(Ns,ynn_means,yerr=ynn_std,c=c)\n",
    "\n",
    "    plt.legend()\n",
    "    plt.xlabel('$n$ samples')\n",
    "    plt.ylabel('Mean Squared Error')\n",
    "    if save == True:\n",
    "        plt.savefig('cpu_error_comp_{}.pdf'.format(data))\n",
    "    else:\n",
    "        plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "d=2.000000, n=100, trials: 10/10, time=0\n",
      "d=2.000000, n=500, trials: 10/10, time=2\n",
      "d=2.000000, n=1000, trials: 10/10, time=8\n",
      "d=2.000000, n=1500, trials: 10/10, time=22\n",
      "d=2.000000, n=2000, trials: 10/10, time=44\n",
      "d=2.000000, n=2500, trials: 10/10, time=77\n",
      "==== Done 2.000000 d level ====\n",
      "d=5.000000, n=100, trials: 10/10, time=77\n",
      "d=5.000000, n=500, trials: 10/10, time=79\n",
      "d=5.000000, n=1000, trials: 10/10, time=85\n",
      "d=5.000000, n=1500, trials: 10/10, time=98\n",
      "d=5.000000, n=2000, trials: 10/10, time=123\n",
      "d=5.000000, n=2500, trials: 10/10, time=160\n",
      "==== Done 5.000000 d level ====\n",
      "d=10.000000, n=100, trials: 10/10, time=161\n",
      "d=10.000000, n=500, trials: 10/10, time=163\n",
      "d=10.000000, n=1000, trials: 10/10, time=170\n",
      "d=10.000000, n=1500, trials: 10/10, time=185\n",
      "d=10.000000, n=2000, trials: 10/10, time=211\n",
      "d=10.000000, n=2500, trials: 10/10, time=250\n",
      "==== Done 10.000000 d level ====\n"
     ]
    }
   ],
   "source": [
    "### Example 1\n",
    "### P = Unif(-1,1)^d\n",
    "### T_0(x) = exp(x) coordinate-wise\n",
    "\n",
    "Ds = [2,5,10]\n",
    "Ns = [100, 500, 1000, 1500, 2000, 2500]\n",
    "data = 'unif_exp'\n",
    "\n",
    "nn_estimation_rate(sample_uniform,OT_exp,Ns,Ds,data=data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Estimation rates for 1NN estimator\n",
      "Note: theoretical rate is (-2/d)\n",
      "d=2, rate=-0.8588697913372276\n",
      "d=5, rate=-0.4190826947987557\n",
      "d=10, rate=-0.2104679751584282\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "CreatePlot_CPU_Errors('CPU-unif_exp_error.pkl')"
   ]
  },
  {
   "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": "3.8.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}