simple median

neural median.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from tqdm import tqdm_notebook\n",
    "import sklearn as sk\n",
    "import sklearn.metrics\n",
    "\n",
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "def mean_absolute_percentage_error(y_true, y_pred): \n",
    "    y_true, y_pred = np.array(y_true), np.array(y_pred)\n",
    "    return np.mean(np.abs((y_true - y_pred) / y_true)) * 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "xs_train = np.random.uniform(-1, 1, (1000, 5))\n",
    "ys_train = np.median(xs_train, axis=1)\n",
    "\n",
    "xs_test = np.random.uniform(-1, 1, (1000, 5))\n",
    "ys_test = np.median(xs_test, axis=1)\n",
    "\n",
    "xs_test2 = np.random.normal(0, 1, (1000, 5))\n",
    "ys_test2 = np.median(xs_test, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_model(layers):\n",
    "    model = Sequential()\n",
    "    model.add(Dense(layers[0], input_dim=5, activation='relu'))\n",
    "    for l in layers[1:]:\n",
    "        model.add(Dense(l, input_dim=5, activation='relu'))\n",
    "    model.add(Dense(1))\n",
    "    model.compile(loss='mean_squared_error', optimizer='adam')\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[5]\n",
      "0.030256047280279485 21.129024587973653\n",
      "0.031121650823280233 20.429932248712433\n",
      "0.39859154903909766 26.796106251682545\n",
      "\n",
      "[10]\n",
      "0.019739526289567665 21.59204996700235\n",
      "0.02095126037665723 20.815109864833577\n",
      "0.42457059391257224 27.499531821913436\n",
      "\n",
      "[100]\n",
      "0.0010199594643338689 22.463057269252957\n",
      "0.004249229495881521 21.669214427122103\n",
      "0.4113864029702579 27.05308372155763\n",
      "\n",
      "[1000]\n",
      "0.0004520376869600072 22.444454632977852\n",
      "0.003649622967328898 21.669094535414356\n",
      "0.40105973103053405 26.673923790970306\n",
      "\n",
      "[10, 10]\n",
      "0.005699108450494615 22.17993472187734\n",
      "0.010154578177297265 21.604647398146643\n",
      "0.43389573135131215 27.958277738307064\n",
      "\n",
      "[50, 50]\n",
      "0.00020974870240867818 22.31495489879563\n",
      "0.003244904106442189 21.61388607850857\n",
      "0.3824939181705547 26.157605809455546\n",
      "\n",
      "[100, 100]\n",
      "1.4699913802082056e-05 22.42388236348745\n",
      "0.002165714208292056 21.72246746460389\n",
      "0.38703285454082936 26.46199116474197\n",
      "\n",
      "[1000, 1000]\n",
      "7.010253801702123e-05 22.489723406664233\n",
      "0.0011533135005442645 21.7244513095109\n",
      "0.3937379193300744 26.73518432667773\n",
      "\n",
      "[10, 10, 10]\n",
      "0.008072452595890461 22.120908518254954\n",
      "0.020859176172920692 21.493421902845057\n",
      "0.43488613612458044 28.599210984527723\n",
      "\n",
      "[100, 100, 100]\n",
      "2.2458645510210152e-05 22.40489708352883\n",
      "0.00233151248430418 21.660590168667024\n",
      "0.38288814536745314 26.311009261936274\n",
      "\n",
      "[10, 10, 10, 10]\n",
      "0.004796239732775853 22.181825668667766\n",
      "0.011370873609471229 21.566316132684758\n",
      "0.39455097127413946 26.70313833759559\n",
      "\n",
      "[100, 100, 100, 100]\n",
      "5.834447390057562e-06 22.45040559059972\n",
      "0.0022323232883944954 21.658926231938512\n",
      "0.3864176872546934 26.44920075218356\n",
      "\n"
     ]
    }
   ],
   "source": [
    "for arch in [\n",
    "    [5],\n",
    "    [10],\n",
    "    [100],\n",
    "    [1000],\n",
    "    [10, 10],\n",
    "    [50, 50],\n",
    "    [100, 100],\n",
    "    [1000, 1000],\n",
    "    [10, 10, 10],\n",
    "    [100, 100, 100],\n",
    "    [10, 10, 10, 10],\n",
    "    [100, 100, 100, 100],\n",
    "]:\n",
    "    model = get_model(arch)\n",
    "    model.fit(xs_train, ys_train, batch_size=100, epochs=1000, verbose=0)\n",
    "    print(arch)\n",
    "    for xs, ys in [(xs_train, ys_train), (xs_test, ys_test), (xs_test2, ys_test2)]:\n",
    "        print(\n",
    "            sk.metrics.mean_squared_error(ys, model.predict(xs)),\n",
    "            mean_absolute_percentage_error(ys + 2, model.predict(xs) + 2),\n",
    "        )\n",
    "    print()"
   ]
  },
  {
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}