blobs circles moons strawberry fields scipy-1.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "469da9a5",
   "metadata": {},
   "source": [
    "# Blobs Circles Moons strawberry fields"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "17216257",
   "metadata": {},
   "outputs": [],
   "source": [
    "from qutip import destroy\n",
    "import strawberryfields as sf\n",
    "from strawberryfields.ops import *\n",
    "import numpy as np\n",
    "import scipy.optimize\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib as mpl\n",
    "from matplotlib import colors, colorbar, cm\n",
    "import time\n",
    "from tqdm.notebook import tqdm"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f431aafe",
   "metadata": {},
   "source": [
    "1-moodeののQNNを構成する"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "7260f9ce",
   "metadata": {},
   "outputs": [],
   "source": [
    "def one_mode_variational_quantum_circuit_cubic(params, depth, input_state=None):\n",
    "    # Gate layer: R-S-R-D-K\n",
    "    def layer(i, q):\n",
    "        Rgate(params[i*7+0])                | q\n",
    "        Sgate(params[i*7+1], params[i*7+2]) | q\n",
    "        Rgate(params[i*7+3])                | q\n",
    "        Dgate(params[i*7+4], params[i*7+5]) | q\n",
    "        Vgate(params[i*7+6])                | q\n",
    "        return q\n",
    "    \n",
    "    # Start SF program\n",
    "    prog = sf.Program(1)\n",
    "\n",
    "    # construct the circuit\n",
    "    with prog.context as q:\n",
    "        if input_state is not None:\n",
    "            Ket(input_state) | q\n",
    "        else:\n",
    "            Vacuum() | q\n",
    "        for k in range(depth):\n",
    "            q = layer(k, q)\n",
    "    \n",
    "    eng = sf.Engine(\"fock\", backend_options={\"cutoff_dim\": dim})\n",
    "    state = eng.run(prog).state\n",
    "    ket = state.ket()\n",
    "    return ket"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f4424004",
   "metadata": {},
   "outputs": [],
   "source": [
    "def one_mode_U(x1_0, x2_0):\n",
    "    x = np.array([x1_0, x2_0])\n",
    "    params = (x @ u + v) / 1\n",
    "    phi = one_mode_variational_quantum_circuit_cubic(params, depth_U)\n",
    "    return phi\n",
    "\n",
    "def one_mode_expect_Fock(psi):\n",
    "    n = np.zeros(dim)\n",
    "    prob = np.real(np.conjugate(psi) * psi)\n",
    "    fock_basis = np.arange(dim)\n",
    "    '''\n",
    "    prob_posi = 0\n",
    "    prob_nega = 0\n",
    "    for i in range(dim):\n",
    "        if i % 2 == 0:\n",
    "            prob_posi += prob[i]\n",
    "        else:\n",
    "            prob_nega += prob[i]\n",
    "    '''\n",
    "    return np.sum(fock_basis * prob) #, prob_posi, prob_nega\n",
    "\n",
    "def one_mode_expect_X(psi):\n",
    "    # 測定用の演算子\n",
    "    x = 1/np.sqrt(2) * (adag + a)\n",
    "    \n",
    "    # それぞれのモードをx = 1/√2 (a† + a)で測定\n",
    "    expect_value = np.conjugate(psi.T) @ x @ psi\n",
    "    return expect_value"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e2c6bb9f",
   "metadata": {},
   "source": [
    "Cost関数を定義。ここではbinary cross entropyとする。\n",
    "\n",
    "$C(\\theta) = -1/N \\sum_{n=1}^{N} [t_n \\log{y_n} + (1 - t_n) \\log{(1 - y_n)}]$\n",
    "\n",
    "$t_n$は正解の0,1、$y_n$は分類器によって出力された0,1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "2fbfca48",
   "metadata": {},
   "outputs": [],
   "source": [
    "def discriminator(x1_0, x2_0, params):\n",
    "    phi = one_mode_U(x1_0, x2_0)\n",
    "    state = one_mode_variational_quantum_circuit_cubic(params, depth, phi)\n",
    "    expect_value = one_mode_expect_Fock(state)\n",
    "\n",
    "    # シグモイド関数により0 - 1にする\n",
    "    result = 1 / (1 + np.exp(10*(params[7*depth] - expect_value)))\n",
    "    return result\n",
    "    \n",
    "def get_cost(y, t):\n",
    "    delta = 1e-12 # logの中身が0にならないようにするため\n",
    "    return -np.sum(t * np.log(y+delta) + (1-t) * np.log(1-y+delta)) / len(y)\n",
    "\n",
    "def cost(params):\n",
    "    y = np.zeros(len(t))\n",
    "    for i in range(len(t)):\n",
    "        y[i] = discriminator(x1[i], x2[i], params)\n",
    "    c = get_cost(y, t)\n",
    "    return c\n",
    "\n",
    "# callback用の関数\n",
    "def callback(params):\n",
    "    params_history.append(params)\n",
    "    co = cost(params)\n",
    "    cost_history.append(co)\n",
    "    print(len(cost_history)-1,'cost', co)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "a5767a2b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "u [[ 0.43850091 -0.08240727 -0.20858149  0.11502725 -0.43303765  0.27727129\n",
      "   0.46762725]\n",
      " [-0.2869331  -0.15892388  0.20760404 -0.17663387 -0.41519595 -0.36893826\n",
      "  -0.05766338]]\n",
      "v [ 0.47610275  0.42101565 -0.39993243 -0.31259668  0.37989942  0.24049534\n",
      " -0.30944022]\n"
     ]
    }
   ],
   "source": [
    "dim = 50\n",
    "_a = destroy(dim)\n",
    "a = _a.full()\n",
    "adag = np.conjugate(a.T)\n",
    "\n",
    "depth_U = 1\n",
    "depth_V = 1\n",
    "\n",
    "# パラメータの決定\n",
    "g = 1\n",
    "#np.random.seed(1)\n",
    "u = (g*np.random.rand(2,7*depth_U) - g/2) / 1\n",
    "v = (g*np.random.rand(7*depth_U) - g/2) / 1\n",
    "'''\n",
    "u = np.array([[-0.49201056,  0.46031398,  0.11182869, -0.36183352,  0.2961122,   0.26468657,\n",
    "  -0.44251039],\n",
    " [ 0.0359517,   0.16110539 , 0.11427918,  0.18337387, -0.00461446, -0.23381862,\n",
    "  -0.10574061]])\n",
    "v = np.array([ 0.09133706, -0.36032127, -0.18590361,  0.11834946, -0.15407532,  0.11210203,\n",
    " -0.44292793])\n",
    "'''\n",
    "print('u',u)\n",
    "print('v',v)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "384d88b0",
   "metadata": {},
   "source": [
    "## Moons"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "4fa1c2e2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[-0.29029155  0.3456798   0.47916806 -1.03727167  1.23047539  0.57319032\n",
      "  0.84445679  1.03043774 -0.8683864   0.43467891 -0.25779211 -0.06380638\n",
      "  0.88465564  0.19693141 -0.92806637  0.63697653  0.60432816  0.70826732\n",
      "  0.61294293 -0.00410577 -1.00634662 -0.81896547 -0.63945885  0.89675433\n",
      " -0.19732942  0.85148424  0.47775303 -0.88271286 -0.01009895  0.5410771\n",
      " -0.49504722 -0.77188558  0.27333581  0.64996669  0.96953943 -0.07126385\n",
      " -0.70972406 -0.09525888 -0.83394716  0.90187187 -0.97813319 -1.05216207\n",
      " -0.92084284  1.04835153 -0.73226442  0.25473738 -0.85400026  0.45907629\n",
      " -0.79158713  0.63764944 -0.06302218  1.81106918  2.04225275  1.8000829\n",
      "  1.44904639  0.02948083  0.30019474  1.36427172  0.53047052  1.26207055\n",
      "  1.96326895  1.41790883  1.9996195   1.75712895  1.88551312  1.68086865\n",
      "  1.67317114  0.0853685  -0.03148348  2.12683121 -0.0459764   1.94445766\n",
      "  0.35221223  0.03906502  0.00564964  1.70237421  1.65460849  0.04762763\n",
      "  2.02780341  0.55131617  0.13522648  0.39260198  0.05014955  1.85133663\n",
      "  1.25490434  1.67693985  0.01724553  0.00714293  0.07462738  1.22130417\n",
      "  0.38311672  2.07777251  0.38127121  2.00196492  1.50589392  1.8798169\n",
      "  1.20623984  0.11461769  1.96716839  2.01452299]\n",
      "[0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.\n",
      " 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.\n",
      " 0. 0. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1.]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.datasets import make_moons\n",
    "\n",
    "n_each_data_train = 50\n",
    "n_each_data_test = 50\n",
    " \n",
    "x, y = make_moons(n_samples=2*(n_each_data_train+n_each_data_test), shuffle = True,\n",
    "                  noise=0.1, random_state=True,)\n",
    "\n",
    "x1_t0, x2_t0 = x[y==0,0], x[y==0,1]\n",
    "x1_t1, x2_t1 = x[y==1,0], x[y==1,1]\n",
    "\n",
    "x1_t0_train = x1_t0[0:n_each_data_train]\n",
    "x2_t0_train = x2_t0[0:n_each_data_train]\n",
    "x1_t1_train = x1_t1[0:n_each_data_train]\n",
    "x2_t1_train = x2_t1[0:n_each_data_train]\n",
    "\n",
    "x1_t0_test = x1_t0[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x2_t0_test = x2_t0[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x1_t1_test = x1_t1[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x2_t1_test = x2_t1[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "\n",
    "t0_train, t1_train = np.zeros(n_each_data_train), np.ones(n_each_data_train)\n",
    "t0_test, t1_test = np.zeros(n_each_data_test), np.ones(n_each_data_test)\n",
    "\n",
    "x1 = np.concatenate([x1_t0_train, x1_t1_train])\n",
    "x2 = np.concatenate([x2_t0_train, x2_t1_train])\n",
    "t = np.concatenate([t0_train, t1_train])\n",
    "\n",
    "x1_test = np.concatenate([x1_t0_test, x1_t1_test])\n",
    "x2_test = np.concatenate([x2_t0_test, x2_t1_test])\n",
    "t_test = np.concatenate([t0_test, t1_test])\n",
    "\n",
    "print(x1)\n",
    "print(t)\n",
    "\n",
    "plt.figure(figsize=(8, 7))\n",
    " \n",
    "plt.scatter(x1_t0, x2_t0, marker='o', s=25, label=\"t = 0\")\n",
    "plt.scatter(x1_t1, x2_t1, marker='o', s=25, label=\"t = 1\")\n",
    "plt.legend()\n",
    "plt.xlabel(\"x1\")\n",
    "plt.ylabel(\"x2\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "827061e8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 cost 1.3488193114770146\n",
      "2 cost 1.2430388234832306\n",
      "3 cost 1.0734784649163904\n",
      "4 cost 0.9650035745185005\n",
      "5 cost 0.7661514589201431\n",
      "6 cost 0.5234552028390919\n",
      "7 cost 0.37585724466152115\n",
      "8 cost 0.31622525708430127\n",
      "9 cost 0.24165283081675973\n",
      "10 cost 0.19581839067419246\n",
      "11 cost 0.18414020187955008\n",
      "12 cost 0.1735587076010086\n",
      "13 cost 0.1689126590631057\n",
      "14 cost 0.16153037388999722\n",
      "15 cost 0.15035577561722635\n",
      "16 cost 0.14054395849371787\n",
      "17 cost 0.12878519199576582\n",
      "18 cost 0.11966929338823844\n",
      "19 cost 0.10134885263750337\n",
      "20 cost 0.09496872660442629\n",
      "21 cost 0.09412828418260227\n",
      "22 cost 0.08682575518912429\n",
      "23 cost 0.0760132009452873\n",
      "24 cost 0.06992204738734413\n",
      "25 cost 0.0676172941615202\n",
      "26 cost 0.06373890252043131\n",
      "27 cost 0.061121755646047576\n",
      "28 cost 0.0580052183894978\n",
      "29 cost 0.05628686842712363\n",
      "30 cost 0.05580756229548504\n",
      "31 cost 0.05565208479935763\n",
      "32 cost 0.055486684735660007\n",
      "33 cost 0.05540469107147525\n",
      "34 cost 0.055263900326132404\n",
      "35 cost 0.055019199185377746\n",
      "36 cost 0.054495873897143665\n",
      "37 cost 0.05351376206179172\n",
      "38 cost 0.05278012619490444\n",
      "39 cost 0.05270352501146327\n",
      "40 cost 0.052675851397064585\n",
      "41 cost 0.05266346017709728\n",
      "42 cost 0.052644533787716774\n",
      "43 cost 0.05261747829710166\n",
      "44 cost 0.052570683440349215\n",
      "45 cost 0.05248754196909561\n",
      "46 cost 0.052331895007214815\n",
      "47 cost 0.05155760879682283\n",
      "48 cost 0.05139777085087476\n",
      "49 cost 0.050635535297119005\n",
      "50 cost 0.05002974970588766\n",
      "51 cost 0.04898008880739102\n",
      "52 cost 0.047386570792395416\n",
      "53 cost 0.04723857892833722\n",
      "54 cost 0.047041445993287745\n",
      "55 cost 0.046889465118780935\n",
      "56 cost 0.04682183422759068\n",
      "57 cost 0.04677769137288891\n",
      "58 cost 0.046222006875536514\n",
      "59 cost 0.046163892060114016\n",
      "60 cost 0.04608187488427653\n",
      "61 cost 0.04545419373677975\n",
      "62 cost 0.04465573957432427\n",
      "63 cost 0.04420584955262803\n",
      "64 cost 0.04352697602866468\n",
      "65 cost 0.043304260553541286\n",
      "66 cost 0.04303592886590819\n",
      "67 cost 0.04287199612996789\n",
      "68 cost 0.042671086277139884\n",
      "69 cost 0.042337861564259054\n",
      "70 cost 0.04191676130176112\n",
      "71 cost 0.041757413315882584\n",
      "72 cost 0.04158837284792274\n",
      "73 cost 0.041531924851153104\n",
      "74 cost 0.041481957186880046\n",
      "75 cost 0.041429457153489294\n",
      "76 cost 0.041415115800123366\n",
      "77 cost 0.041406503906578775\n",
      "78 cost 0.041402320604546\n",
      "79 cost 0.041394247915979926\n",
      "80 cost 0.04138050124729596\n",
      "81 cost 0.04135991278713659\n",
      "82 cost 0.0413220700573776\n",
      "83 cost 0.041010735992936524\n",
      "84 cost 0.040491782525721706\n",
      "85 cost 0.039731980381368895\n",
      "86 cost 0.03857338434809077\n",
      "87 cost 0.03362797099405776\n",
      "88 cost 0.03357264500819446\n",
      "89 cost 0.03312134076960676\n",
      "90 cost 0.03238294513629014\n",
      "91 cost 0.03104446742762286\n",
      "92 cost 0.029172965323084433\n",
      "93 cost 0.028310574780692148\n",
      "94 cost 0.026868215181490015\n",
      "95 cost 0.0265553647754621\n",
      "96 cost 0.026243580861308433\n",
      "97 cost 0.025687231735311775\n",
      "98 cost 0.02506071084236599\n",
      "99 cost 0.02450348054306488\n",
      "100 cost 0.024095031459290376\n",
      "Warning: Maximum number of iterations has been exceeded.\n",
      "         Current function value: 0.024095\n",
      "         Iterations: 100\n",
      "         Function evaluations: 1170\n",
      "         Gradient evaluations: 130\n",
      "time 125.5984674 s\n"
     ]
    },
    {
     "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": [
    "depth = 1\n",
    "\n",
    "# 学習\n",
    "params = np.random.normal(0, 0.1, [7*depth + 1]) # 最初のパラメータ\n",
    "maxiter = 100\n",
    "\n",
    "params_history = []\n",
    "params_history.append(params)\n",
    "cost_history = []\n",
    "cost_history.append(cost(params))\n",
    "\n",
    "method = \"BFGS\"\n",
    "options = {\"disp\": True, \"maxiter\": maxiter}\n",
    "\n",
    "start = time.perf_counter()\n",
    "opt = scipy.optimize.minimize(cost, params, method=method,\n",
    "                              callback=lambda x: callback(x), options=options)\n",
    "end = time.perf_counter()\n",
    "\n",
    "print('time', end - start, 's')\n",
    "nit = opt['nit']\n",
    "plt.plot(np.linspace(0, nit, nit+1), cost_history)\n",
    "plt.yscale('log')\n",
    "plt.xlabel('step')\n",
    "plt.ylabel('cost function')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "b60d0bdc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "initial parameters: [ 0.05754508 -0.08629276  0.11298462 -0.15267042  0.06580368 -0.04502428\n",
      "  0.07229795 -0.0570869 ]\n",
      "learnt_parameters: [-4.06215077  1.23163464 -9.91093203  4.86303663  0.70067597  1.81575965\n",
      "  6.35614786  4.54954638]\n",
      "-------------------------------\n",
      "train dataset\n",
      "teacher label [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "learnt label  [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "number of mislabel: 1.0\n",
      "-------------------------------\n",
      "test dataset\n",
      "teacher label [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "learnt label  [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "number of mislabel: 1.0\n"
     ]
    }
   ],
   "source": [
    "print('initial parameters:', params_history[0])\n",
    "print('learnt_parameters:', params_history[-1])\n",
    "\n",
    "print('-------------------------------')\n",
    "print('train dataset')\n",
    "\n",
    "learnt_label_train = np.array([0.0 for i in range(len(t))])\n",
    "for i in range(len(learnt_label_train)):\n",
    "    learnt_label_train[i] = discriminator(x1[i], x2[i], params_history[-1])\n",
    "learnt_label_train = np.round(learnt_label_train).astype(int)\n",
    "n_miss_train = sum(abs(t - learnt_label_train))\n",
    "t = t.astype(int)\n",
    "\n",
    "print('teacher label', t)\n",
    "print('learnt label ', learnt_label_train)\n",
    "print('number of mislabel:', n_miss_train)\n",
    "\n",
    "print('-------------------------------')\n",
    "print('test dataset')\n",
    "\n",
    "learnt_label_test = np.array([0.0 for i in range(len(t_test))])\n",
    "for i in range(len(learnt_label_test)):\n",
    "    learnt_label_test[i] = discriminator(x1_test[i], x2_test[i], params_history[-1])\n",
    "learnt_label_test = np.round(learnt_label_test).astype(int)\n",
    "n_miss_test = sum(abs(t_test - learnt_label_test))\n",
    "t_test = t_test.astype(int)\n",
    "\n",
    "print('teacher label', t_test)\n",
    "print('learnt label ',learnt_label_test)\n",
    "print('number of mislabel:', n_miss_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "d8d50cbc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "be622a2eb0024317bb1960888e1e5fd0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/100 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "x1_lin = np.linspace(-1.5, 2.5, 100)\n",
    "x2_lin = np.linspace(-1, 1.5, 100)\n",
    "learnt_label = np.zeros((len(x1_lin), len(x2_lin)))\n",
    "value = np.zeros_like(learnt_label)\n",
    "for i in tqdm(range(len(x1_lin))):\n",
    "    for j in range(len(x2_lin)):\n",
    "        value[j,i] = discriminator(x1_lin[i], x2_lin[j], params_history[-1])\n",
    "        if value[j,i] > 0.5:\n",
    "            learnt_label[j,i] = 1\n",
    "        else:\n",
    "            learnt_label[j,i] = 0\n",
    "\n",
    "plt.pcolor(x1_lin, x2_lin, learnt_label, shading=\"nearest\", cmap=\"gray\")\n",
    "plt.colorbar()\n",
    "plt.scatter(x1_t0, x2_t0, color='blue')\n",
    "plt.scatter(x1_t1, x2_t1, color='red')\n",
    "plt.xlabel(r\"$x_1$\")\n",
    "plt.ylabel(r\"$x_2$\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b79a791c",
   "metadata": {},
   "source": [
    "## Blobs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "af776dfa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 4.44097997e-01  1.83773582e-01  2.92533998e-01  1.12471455e+00\n",
      "  9.45121887e-01  3.47186679e-01  4.60360226e-01 -2.35350010e-01\n",
      "  8.29560666e-01  9.08591004e-02  2.35299052e-01  3.29843646e-01\n",
      " -5.07217751e-01  1.01457075e+00 -1.14082639e-01  9.25900977e-01\n",
      " -2.20061366e-03  1.06207278e+00  4.09271686e-01  9.99162743e-01\n",
      "  2.42687954e-01  4.69258545e-01 -9.49889850e-01  1.03205278e+00\n",
      "  2.06536474e-01  1.50222387e+00 -1.02471934e-01  4.86069941e-01\n",
      "  7.26952331e-01  1.31311629e-01  8.80507159e-01  1.49966127e-01\n",
      "  1.06996801e+00  5.84451877e-01  5.09201107e-01  1.68872686e-01\n",
      "  2.13728216e-01  8.20601509e-01 -1.89902726e-01  1.62998687e-01\n",
      " -3.91354422e-03  2.45346703e-01  5.80510633e-01  1.02745508e+00\n",
      "  7.87335628e-01  4.43298521e-01  1.37268774e+00  1.70615271e-01\n",
      "  7.68247823e-02  8.31893924e-01 -2.10534543e+00 -1.08112729e+00\n",
      " -1.91554469e+00 -2.73406727e+00 -1.55734484e+00 -3.06348874e+00\n",
      " -1.88176210e+00 -2.04847058e+00 -2.05140651e+00 -2.04543536e+00\n",
      " -2.04642593e+00 -1.89438997e+00 -2.02982774e+00 -2.27554148e+00\n",
      " -2.38456277e+00 -2.54699807e+00 -2.60580737e+00 -2.64734319e+00\n",
      " -2.11792654e+00 -1.53899178e+00 -1.83255383e+00 -1.94717379e+00\n",
      " -1.66397093e+00 -6.85013423e-01 -2.75319696e+00 -1.93135452e+00\n",
      " -2.02420409e+00 -1.70647003e+00 -2.39850689e+00 -2.32087569e+00\n",
      " -1.55605343e+00 -1.63868837e+00 -2.02729835e+00 -2.66050101e+00\n",
      " -1.06778622e+00 -2.40415986e+00 -1.66027380e+00 -1.17137667e+00\n",
      " -2.02194528e+00 -2.41958135e+00 -2.16921202e+00 -2.41034806e+00\n",
      " -2.09410200e+00 -1.97072422e+00 -1.51732021e+00 -2.20962135e+00\n",
      " -2.11187472e+00 -1.53298002e+00 -1.94712400e+00 -2.58884973e+00]\n",
      "[0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.\n",
      " 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.\n",
      " 0. 0. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1.]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.datasets import make_blobs\n",
    "\n",
    "n_each_data_train = 50\n",
    "n_each_data_test = 50\n",
    " \n",
    "x, y = make_blobs(n_samples=2*(n_each_data_train+n_each_data_test), n_features=2, centers=2, \n",
    "                  cluster_std=0.5, center_box=(-5.0,5.0), shuffle=True, random_state=3,)\n",
    "\n",
    "x1_t0, x2_t0 = x[y==0,0], x[y==0,1]\n",
    "x1_t1, x2_t1 = x[y==1,0], x[y==1,1]\n",
    "\n",
    "x1_t0_train = x1_t0[0:n_each_data_train]\n",
    "x2_t0_train = x2_t0[0:n_each_data_train]\n",
    "x1_t1_train = x1_t1[0:n_each_data_train]\n",
    "x2_t1_train = x2_t1[0:n_each_data_train]\n",
    "\n",
    "x1_t0_test = x1_t0[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x2_t0_test = x2_t0[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x1_t1_test = x1_t1[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x2_t1_test = x2_t1[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "\n",
    "t0_train, t1_train = np.zeros(n_each_data_train), np.ones(n_each_data_train)\n",
    "t0_test, t1_test = np.zeros(n_each_data_test), np.ones(n_each_data_test)\n",
    "\n",
    "x1 = np.concatenate([x1_t0_train, x1_t1_train])\n",
    "x2 = np.concatenate([x2_t0_train, x2_t1_train])\n",
    "t = np.concatenate([t0_train, t1_train])\n",
    "\n",
    "x1_test = np.concatenate([x1_t0_test, x1_t1_test])\n",
    "x2_test = np.concatenate([x2_t0_test, x2_t1_test])\n",
    "t_test = np.concatenate([t0_test, t1_test])\n",
    "\n",
    "print(x1)\n",
    "print(t)\n",
    "\n",
    "plt.figure(figsize=(8, 7))\n",
    " \n",
    "plt.scatter(x1_t0, x2_t0, marker='o', s=25, label=\"t = 0\")\n",
    "plt.scatter(x1_t1, x2_t1, marker='o', s=25, label=\"t = 1\")\n",
    "plt.legend()\n",
    "plt.xlabel(\"x1\")\n",
    "plt.ylabel(\"x2\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "d3a8dd86",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 cost 0.6290411445265331\n",
      "2 cost 0.3683598031517619\n",
      "3 cost 0.2620479551681123\n",
      "4 cost 0.14261612938507348\n",
      "5 cost 0.10792263602364763\n",
      "6 cost 0.05325053492394152\n",
      "7 cost 0.025084841783995313\n",
      "8 cost 0.011611227593632756\n",
      "9 cost 0.0032925348021670027\n",
      "10 cost 0.002040126931041781\n",
      "11 cost 0.001057162509190959\n",
      "12 cost 0.0006166620829519055\n",
      "13 cost 0.0003609063074740786\n",
      "14 cost 0.00022243352683559575\n",
      "15 cost 0.0001522901084177009\n",
      "16 cost 0.0001366408706859382\n",
      "17 cost 0.00011956605074702423\n",
      "18 cost 9.57738180664704e-05\n",
      "19 cost 8.859275211801717e-05\n",
      "20 cost 8.692764267063672e-05\n",
      "21 cost 8.481575896911949e-05\n",
      "22 cost 8.113456088035307e-05\n",
      "23 cost 7.458594282324235e-05\n",
      "24 cost 6.320159061377377e-05\n",
      "25 cost 4.492330835457307e-05\n",
      "26 cost 2.5385442347891008e-05\n",
      "27 cost 1.4078579874999502e-05\n",
      "28 cost 6.883512694109119e-06\n",
      "29 cost 3.0107708216788206e-06\n",
      "30 cost 1.4980277253062824e-06\n",
      "31 cost 9.792099130377636e-07\n",
      "32 cost 7.149222266448509e-07\n",
      "33 cost 4.886380398046159e-07\n",
      "34 cost 1.9118704795390073e-07\n",
      "35 cost 1.7771962736463384e-07\n",
      "36 cost 1.7631371884138994e-07\n",
      "37 cost 1.7487447236724662e-07\n",
      "38 cost 1.724519736013546e-07\n",
      "39 cost 1.6814788683720643e-07\n",
      "40 cost 1.603805508541443e-07\n",
      "41 cost 1.4652245665591003e-07\n",
      "Optimization terminated successfully.\n",
      "         Current function value: 0.000000\n",
      "         Iterations: 41\n",
      "         Function evaluations: 414\n",
      "         Gradient evaluations: 46\n",
      "time 43.00566739999999 s\n"
     ]
    },
    {
     "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": [
    "depth = 1\n",
    "\n",
    "# 学習\n",
    "params = np.random.normal(0, 0.1, [7*depth + 1]) # 最初のパラメータ\n",
    "maxiter = 100\n",
    "\n",
    "params_history = []\n",
    "params_history.append(params)\n",
    "cost_history = []\n",
    "cost_history.append(cost(params))\n",
    "\n",
    "method = \"BFGS\"\n",
    "options = {\"disp\": True, \"maxiter\": maxiter}\n",
    "\n",
    "start = time.perf_counter()\n",
    "opt = scipy.optimize.minimize(cost, params, method=method,\n",
    "                              callback=lambda x: callback(x), options=options)\n",
    "end = time.perf_counter()\n",
    "\n",
    "print('time', end - start, 's')\n",
    "nit = opt['nit']\n",
    "plt.plot(np.linspace(0, nit, nit+1), cost_history)\n",
    "plt.yscale('log')\n",
    "plt.xlabel('step')\n",
    "plt.ylabel('cost function')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "6a602e6b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "initial parameters: [ 0.00663932  0.14379017  0.06598695 -0.01028692 -0.1418387  -0.06650375\n",
      "  0.02874182 -0.12619529]\n",
      "learnt_parameters: [-2.35053151e-01  5.62435792e-01  5.64241303e-01  2.30829114e-03\n",
      " -9.92056585e-02  1.27793806e-01 -6.04099954e-01  3.74278207e+00]\n",
      "-------------------------------\n",
      "train dataset\n",
      "teacher label [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "learnt label  [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "number of mislabel: 0.0\n",
      "-------------------------------\n",
      "test dataset\n",
      "teacher label [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "learnt label  [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "number of mislabel: 0.0\n"
     ]
    }
   ],
   "source": [
    "print('initial parameters:', params_history[0])\n",
    "print('learnt_parameters:', params_history[-1])\n",
    "\n",
    "print('-------------------------------')\n",
    "print('train dataset')\n",
    "\n",
    "learnt_label_train = np.array([0.0 for i in range(len(t))])\n",
    "for i in range(len(learnt_label_train)):\n",
    "    learnt_label_train[i] = discriminator(x1[i], x2[i], params_history[-1])\n",
    "learnt_label_train = np.round(learnt_label_train).astype(int)\n",
    "n_miss_train = sum(abs(t - learnt_label_train))\n",
    "t = t.astype(int)\n",
    "\n",
    "print('teacher label', t)\n",
    "print('learnt label ', learnt_label_train)\n",
    "print('number of mislabel:', n_miss_train)\n",
    "\n",
    "print('-------------------------------')\n",
    "print('test dataset')\n",
    "\n",
    "learnt_label_test = np.array([0.0 for i in range(len(t_test))])\n",
    "for i in range(len(learnt_label_test)):\n",
    "    learnt_label_test[i] = discriminator(x1_test[i], x2_test[i], params_history[-1])\n",
    "learnt_label_test = np.round(learnt_label_test).astype(int)\n",
    "n_miss_test = sum(abs(t_test - learnt_label_test))\n",
    "t_test = t_test.astype(int)\n",
    "\n",
    "print('teacher label', t_test)\n",
    "print('learnt label ',learnt_label_test)\n",
    "print('number of mislabel:', n_miss_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "6b7b0278",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d1ab44a3bb5543858317e5854dcfd4fa",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/100 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "x1_lin = np.linspace(-3.5, 2, 100)\n",
    "x2_lin = np.linspace(-1.5, 3.5, 100)\n",
    "learnt_label = np.zeros((len(x1_lin), len(x2_lin)))\n",
    "value = np.zeros_like(learnt_label)\n",
    "for i in tqdm(range(len(x1_lin))):\n",
    "    for j in range(len(x2_lin)):\n",
    "        value[j,i] = discriminator(x1_lin[i], x2_lin[j], params_history[-1])\n",
    "        if value[j,i] > 0.5:\n",
    "            learnt_label[j,i] = 1\n",
    "        else:\n",
    "            learnt_label[j,i] = 0\n",
    "\n",
    "plt.pcolor(x1_lin, x2_lin, learnt_label, shading=\"nearest\", cmap=\"gray\")\n",
    "plt.colorbar()\n",
    "plt.scatter(x1_t0, x2_t0, color='blue')\n",
    "plt.scatter(x1_t1, x2_t1, color='red')\n",
    "plt.xlabel(r\"$x_1$\")\n",
    "plt.ylabel(r\"$x_2$\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7a7f6ace",
   "metadata": {},
   "source": [
    "## Circles"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "c9c0ab3c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[-0.03190127  0.08274053 -0.16058388  0.21113828 -0.00505293 -0.15642995\n",
      "  0.16609505  0.08273599 -0.18499502 -0.1427308  -0.17153917  0.09551509\n",
      " -0.10491614 -0.00713407 -0.14732465 -0.49370938 -0.02084201 -0.11415317\n",
      " -0.15729459  0.05505298  0.05868738 -0.11744406 -0.06106753 -0.10070878\n",
      " -0.01960367 -0.03171476 -0.57959966  0.19682313  0.21265156  0.0786386\n",
      " -0.04619745  0.29571994 -0.02663912 -0.2246165   0.0162749   0.16937464\n",
      " -0.24716181  0.04931894  0.17050611  0.29897871  0.08792805  0.05344897\n",
      "  0.10008946  0.31092548  0.01378217 -0.17504021 -0.05402217 -0.25082317\n",
      " -0.18503418 -0.25200824 -0.94465953 -1.14869369  0.39216127 -0.85684052\n",
      " -0.40046915  0.50735524 -0.14320516  0.70687613  0.98277608 -0.98941462\n",
      " -0.90443069  0.47824105  1.26419112  0.63244487  0.84521441 -0.2915405\n",
      " -1.08451305 -0.73989865 -0.59940267  0.94303075 -0.09822512 -0.08405574\n",
      "  0.81723821 -0.69286316  0.34321514 -0.46322952  0.73697352  0.42683365\n",
      "  0.25335868  0.08567518  0.08810988 -0.02790555  0.30609894 -0.66234701\n",
      "  0.50212892  0.47959475  0.94026695  0.20188732  0.89125224  0.93283874\n",
      " -0.85897167  0.24249714 -1.02784877 -1.12738328  0.48803371  0.67466233\n",
      " -0.63645575 -0.95681365 -0.18385512 -0.41551373]\n",
      "[0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.\n",
      " 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.\n",
      " 0. 0. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1. 1.\n",
      " 1. 1. 1. 1.]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.datasets import make_circles\n",
    "\n",
    "n_each_data_train = 50\n",
    "n_each_data_test = 50\n",
    " \n",
    "x, y = make_circles(n_samples=2*(n_each_data_train+n_each_data_test), shuffle=True, noise = 0.2, \n",
    "                    random_state=3, factor=0.1)\n",
    "\n",
    "x1_t0, x2_t0 = x[y==1,0], x[y==1,1]\n",
    "x1_t1, x2_t1 = x[y==0,0], x[y==0,1]\n",
    "\n",
    "x1_t0_train = x1_t0[0:n_each_data_train]\n",
    "x2_t0_train = x2_t0[0:n_each_data_train]\n",
    "x1_t1_train = x1_t1[0:n_each_data_train]\n",
    "x2_t1_train = x2_t1[0:n_each_data_train]\n",
    "\n",
    "x1_t0_test = x1_t0[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x2_t0_test = x2_t0[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x1_t1_test = x1_t1[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "x2_t1_test = x2_t1[n_each_data_train:n_each_data_train+n_each_data_test]\n",
    "\n",
    "t0_train, t1_train = np.zeros(n_each_data_train), np.ones(n_each_data_train)\n",
    "t0_test, t1_test = np.zeros(n_each_data_test), np.ones(n_each_data_test)\n",
    "\n",
    "x1 = np.concatenate([x1_t0_train, x1_t1_train])\n",
    "x2 = np.concatenate([x2_t0_train, x2_t1_train])\n",
    "t = np.concatenate([t0_train, t1_train])\n",
    "\n",
    "x1_test = np.concatenate([x1_t0_test, x1_t1_test])\n",
    "x2_test = np.concatenate([x2_t0_test, x2_t1_test])\n",
    "t_test = np.concatenate([t0_test, t1_test])\n",
    "\n",
    "print(x1)\n",
    "print(t)\n",
    "\n",
    "plt.figure(figsize=(8, 7))\n",
    " \n",
    "plt.scatter(x1_t0, x2_t0, marker='o', s=25, label=\"t = 0\")\n",
    "plt.scatter(x1_t1, x2_t1, marker='o', s=25, label=\"t = 1\")\n",
    "plt.legend()\n",
    "plt.xlabel(\"x1\")\n",
    "plt.ylabel(\"x2\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "612f33ef",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1 cost 0.9450733964854521\n",
      "2 cost 0.5304859999834993\n",
      "3 cost 0.45950245141587603\n",
      "4 cost 0.4159376514125389\n",
      "5 cost 0.3961104233120545\n",
      "6 cost 0.3687974280928467\n",
      "7 cost 0.32918413536902025\n",
      "8 cost 0.3156941692362028\n",
      "9 cost 0.2989535571522099\n",
      "10 cost 0.2838008984208515\n",
      "11 cost 0.276879622019449\n",
      "12 cost 0.26591123242006715\n",
      "13 cost 0.25025365929946564\n",
      "14 cost 0.239414441254861\n",
      "15 cost 0.22772204621601422\n",
      "16 cost 0.22666817660238997\n",
      "17 cost 0.22494303686139777\n",
      "18 cost 0.22212156204197575\n",
      "19 cost 0.21779477142061843\n",
      "20 cost 0.212115348900843\n",
      "21 cost 0.20256634334241547\n",
      "22 cost 0.18535552460070404\n",
      "23 cost 0.17375996725801013\n",
      "24 cost 0.16021715556620653\n",
      "25 cost 0.15384577230642346\n",
      "26 cost 0.15067356258413503\n",
      "27 cost 0.1469435803973302\n",
      "28 cost 0.14521639022823082\n",
      "29 cost 0.142513877021876\n",
      "30 cost 0.13959635640019905\n",
      "31 cost 0.13672085869605877\n",
      "32 cost 0.1322850773444503\n",
      "33 cost 0.12578491715871495\n",
      "34 cost 0.12138109765471664\n",
      "35 cost 0.11410789543927176\n",
      "36 cost 0.10617480603542248\n",
      "37 cost 0.09920392693446349\n",
      "38 cost 0.096299453009163\n",
      "39 cost 0.09469764864243058\n",
      "40 cost 0.09286988253894218\n",
      "41 cost 0.09190308395484764\n",
      "42 cost 0.09071490543537906\n",
      "43 cost 0.08622700548114107\n",
      "44 cost 0.08565002669627496\n",
      "45 cost 0.08473003366004374\n",
      "46 cost 0.0833879971735823\n",
      "47 cost 0.08208762476740689\n",
      "48 cost 0.08166676254954451\n",
      "49 cost 0.08090607130862872\n",
      "50 cost 0.0802295085431748\n",
      "51 cost 0.0791310707459421\n",
      "52 cost 0.07868295277548865\n",
      "53 cost 0.07788692289343589\n",
      "54 cost 0.07648646006826963\n",
      "55 cost 0.07426827695841541\n",
      "56 cost 0.07204099865563934\n",
      "57 cost 0.07075414018790084\n",
      "58 cost 0.06850303144238579\n",
      "59 cost 0.06742480368232108\n",
      "60 cost 0.06656246126193123\n",
      "61 cost 0.06599210124356725\n",
      "62 cost 0.06520791939117339\n",
      "63 cost 0.06458538057339909\n",
      "64 cost 0.0642070731787905\n",
      "65 cost 0.06387475457692299\n",
      "66 cost 0.0638408730917844\n",
      "67 cost 0.0638055533553271\n",
      "68 cost 0.06376103539824819\n",
      "69 cost 0.06374686151379441\n",
      "70 cost 0.06372326188211128\n",
      "71 cost 0.0636922488515624\n",
      "72 cost 0.06369066343125858\n",
      "73 cost 0.06368797727035805\n",
      "74 cost 0.06368636582894674\n",
      "75 cost 0.0636861462362968\n",
      "76 cost 0.06368608733592011\n",
      "77 cost 0.06368598559791314\n",
      "78 cost 0.06368581070785105\n",
      "79 cost 0.06368552113760564\n",
      "80 cost 0.0636850329096717\n",
      "81 cost 0.06368417879303044\n",
      "82 cost 0.0636826502186372\n",
      "83 cost 0.06367987873880031\n",
      "84 cost 0.0636749235524233\n",
      "85 cost 0.06366618836416485\n",
      "86 cost 0.06365144385589779\n",
      "87 cost 0.0636282307245556\n",
      "88 cost 0.06359321536339277\n",
      "89 cost 0.06357034476218457\n",
      "90 cost 0.06354265819457398\n",
      "91 cost 0.06351090227236307\n",
      "92 cost 0.06345157177674378\n",
      "93 cost 0.06341804113199835\n",
      "94 cost 0.06336142121538019\n",
      "95 cost 0.06328215708733129\n",
      "96 cost 0.06314641116272177\n",
      "97 cost 0.06297173704877719\n",
      "98 cost 0.06285291644094881\n",
      "99 cost 0.06280278775903425\n",
      "100 cost 0.0627241123430976\n",
      "Warning: Maximum number of iterations has been exceeded.\n",
      "         Current function value: 0.062724\n",
      "         Iterations: 100\n",
      "         Function evaluations: 1053\n",
      "         Gradient evaluations: 117\n",
      "time 110.27807409999997 s\n"
     ]
    },
    {
     "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": [
    "depth = 1\n",
    "\n",
    "# 学習\n",
    "params = np.random.normal(0, 0.1, [7*depth + 1]) # 最初のパラメータ\n",
    "maxiter = 100\n",
    "\n",
    "params_history = []\n",
    "params_history.append(params)\n",
    "cost_history = []\n",
    "cost_history.append(cost(params))\n",
    "\n",
    "method = \"BFGS\"\n",
    "options = {\"disp\": True, \"maxiter\": maxiter}\n",
    "\n",
    "start = time.perf_counter()\n",
    "opt = scipy.optimize.minimize(cost, params, method=method,\n",
    "                              callback=lambda x: callback(x), options=options)\n",
    "end = time.perf_counter()\n",
    "\n",
    "print('time', end - start, 's')\n",
    "nit = opt['nit']\n",
    "plt.plot(np.linspace(0, nit, nit+1), cost_history)\n",
    "plt.yscale('log')\n",
    "plt.xlabel('step')\n",
    "plt.ylabel('cost function')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "c0c9acce",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "initial parameters: [ 0.01302884  0.00653518 -0.02140593 -0.12317383  0.04460896 -0.15475837\n",
      "  0.07742512  0.00507051]\n",
      "learnt_parameters: [-0.11615354 -1.19413608 -0.39810006 -1.09980372 -0.92421184  1.48313568\n",
      "  3.32976866  3.81909046]\n",
      "-------------------------------\n",
      "train dataset\n",
      "teacher label [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "learnt label  [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "number of mislabel: 2.0\n",
      "-------------------------------\n",
      "test dataset\n",
      "teacher label [0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "learnt label  [1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n",
      " 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 0 1 1 1 1 1 1 1 1\n",
      " 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1]\n",
      "number of mislabel: 4.0\n"
     ]
    }
   ],
   "source": [
    "print('initial parameters:', params_history[0])\n",
    "print('learnt_parameters:', params_history[-1])\n",
    "\n",
    "print('-------------------------------')\n",
    "print('train dataset')\n",
    "\n",
    "learnt_label_train = np.array([0.0 for i in range(len(t))])\n",
    "for i in range(len(learnt_label_train)):\n",
    "    learnt_label_train[i] = discriminator(x1[i], x2[i], params_history[-1])\n",
    "learnt_label_train = np.round(learnt_label_train).astype(int)\n",
    "n_miss_train = sum(abs(t - learnt_label_train))\n",
    "t = t.astype(int)\n",
    "\n",
    "print('teacher label', t)\n",
    "print('learnt label ', learnt_label_train)\n",
    "print('number of mislabel:', n_miss_train)\n",
    "\n",
    "print('-------------------------------')\n",
    "print('test dataset')\n",
    "\n",
    "learnt_label_test = np.array([0.0 for i in range(len(t_test))])\n",
    "for i in range(len(learnt_label_test)):\n",
    "    learnt_label_test[i] = discriminator(x1_test[i], x2_test[i], params_history[-1])\n",
    "learnt_label_test = np.round(learnt_label_test).astype(int)\n",
    "n_miss_test = sum(abs(t_test - learnt_label_test))\n",
    "t_test = t_test.astype(int)\n",
    "\n",
    "print('teacher label', t_test)\n",
    "print('learnt label ',learnt_label_test)\n",
    "print('number of mislabel:', n_miss_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "474f792c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3dfc1085a3564b34a12d38a4ced10e25",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/100 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "x1_lin = np.linspace(-1.5, 1.5, 100)\n",
    "x2_lin = np.linspace(-1.5, 1.5, 100)\n",
    "learnt_label = np.zeros((len(x1_lin), len(x2_lin)))\n",
    "value = np.zeros_like(learnt_label)\n",
    "for i in tqdm(range(len(x1_lin))):\n",
    "    for j in range(len(x2_lin)):\n",
    "        value[j,i] = discriminator(x1_lin[i], x2_lin[j], params_history[-1])\n",
    "        if value[j,i] > 0.5:\n",
    "            learnt_label[j,i] = 1\n",
    "        else:\n",
    "            learnt_label[j,i] = 0\n",
    "\n",
    "plt.pcolor(x1_lin, x2_lin, learnt_label, shading=\"nearest\", cmap=\"gray\")\n",
    "plt.colorbar()\n",
    "plt.scatter(x1_t0, x2_t0, color='blue')\n",
    "plt.scatter(x1_t1, x2_t1, color='red')\n",
    "plt.xlabel(r\"$x_1$\")\n",
    "plt.ylabel(r\"$x_2$\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "80aec622",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d49ab983",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}