toy_example.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "toy_example.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyP7mSdPTU5pFIVyI0nMAYMw",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/peisuke/GenPU/blob/main/toy_example.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ZZK2v3Evvc7L"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import torch.optim as optim\n",
        "\n",
        "from sklearn import datasets\n",
        "from sklearn import preprocessing\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "X, z = datasets.make_moons(n_samples=2000, noise=0.05, random_state=0)\n",
        "\n",
        "sc = preprocessing.StandardScaler()\n",
        "sc.fit(X)\n",
        "X = sc.transform(X) / 2"
      ],
      "metadata": {
        "id": "-XbIv6Rf0n0F"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "s = z.copy()\n",
        "s0 = np.zeros_like(s[s == 1])\n",
        "s0[np.random.rand(len(s0)) > 0.8] = 1\n",
        "s[s == 1] = s0"
      ],
      "metadata": {
        "id": "fzVnr2MD1Ly0"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "plt.scatter(X[s==0,0], X[s==0,1], s=3)\n",
        "plt.scatter(X[s==1,0], X[s==1,1], s=3)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "id": "hEtseTHm1R3t",
        "outputId": "c9fae0ca-cde7-407c-ccd3-815c08a27a31"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.collections.PathCollection at 0x7f3b8defa650>"
            ]
          },
          "metadata": {},
          "execution_count": 4
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "class Dataset(torch.utils.data.Dataset):\n",
        "    def __init__(self, data):\n",
        "        self.data = torch.tensor(data, dtype=torch.float32)\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.data)\n",
        "    \n",
        "    def __getitem__(self, idx):\n",
        "        return self.data[idx]"
      ],
      "metadata": {
        "id": "epvnGBf82Foh"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "batchsize = 50\n",
        "\n",
        "labeled_dataset = Dataset(X[s == 1])\n",
        "unlabeled_dataset = Dataset(X[s == 0])\n",
        "\n",
        "labeled_loader = torch.utils.data.DataLoader(labeled_dataset, batch_size = batchsize, shuffle = True, num_workers = 2)\n",
        "unlabeled_loader = torch.utils.data.DataLoader(unlabeled_dataset, batch_size = batchsize, shuffle = True, num_workers = 2)"
      ],
      "metadata": {
        "id": "Z2l72snbf5N5"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "class Generator(nn.Module):\n",
        "    def __init__(self, input_size):\n",
        "        super(Generator, self).__init__()\n",
        "        self.fc1 = nn.Linear(input_size, 128)\n",
        "        self.fc2 = nn.Linear(128, 2)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = F.relu(self.fc1(x))\n",
        "        return torch.tanh(self.fc2(x))\n",
        "\n",
        "class Discriminator(nn.Module):\n",
        "    def __init__(self):\n",
        "        super(Discriminator, self).__init__()\n",
        "        self.fc1 = nn.Linear(2, 128)\n",
        "        self.fc2 = nn.Linear(128, 1)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = F.relu(self.fc1(x))\n",
        "        return torch.sigmoid(self.fc2(x))"
      ],
      "metadata": {
        "id": "MvZDGBHd2nB0"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "use_cuda = True\n",
        "device = torch.device(\"cuda\" if use_cuda else \"cpu\")"
      ],
      "metadata": {
        "id": "B_kr4PWrQUEE"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "device"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YgZAGpNw9cin",
        "outputId": "9f049743-afd4-479e-a535-dfeeaa34c4a5"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "device(type='cuda')"
            ]
          },
          "metadata": {},
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "input_size = 64\n",
        "gen_p = Generator(input_size=input_size).to(device)\n",
        "gen_n = Generator(input_size=input_size).to(device)\n",
        "dis_p = Discriminator().to(device)\n",
        "dis_n = Discriminator().to(device)\n",
        "dis_u = Discriminator().to(device)\n",
        "\n",
        "dp_optimizer = torch.optim.Adam(dis_p.parameters(), lr = 1e-3)\n",
        "dn_optimizer = torch.optim.Adam(dis_n.parameters(), lr = 1e-3)\n",
        "du_optimizer = torch.optim.Adam(dis_u.parameters(), lr = 1e-3)\n",
        "gp_optimizer = torch.optim.Adam(gen_p.parameters(), lr = 1e-3)\n",
        "gn_optimizer = torch.optim.Adam(gen_n.parameters(), lr = 1e-3)"
      ],
      "metadata": {
        "id": "rWtACgVY4NiS"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def infiniteloop(dataloader):\n",
        "    while True:\n",
        "        for x in iter(dataloader):\n",
        "            yield x"
      ],
      "metadata": {
        "id": "jj75i1VYhbAU"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "real_label = 1.\n",
        "fake_label = 0.\n",
        "criterion = nn.BCELoss()"
      ],
      "metadata": {
        "id": "GkqMdCdlT10Y"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "labeled_looper = infiniteloop(labeled_loader)\n",
        "num_epoch = 1000\n",
        "\n",
        "loss_D_epoch = []\n",
        "loss_G_epoch = []\n",
        "\n",
        "for epoch in range(num_epoch):\n",
        "    G_loss = 0\n",
        "    D_loss = 0\n",
        "\n",
        "    for i, unlabeled_x in enumerate(unlabeled_loader):\n",
        "        z = torch.randn(batchsize, input_size, device=device)\n",
        "        xp = next(labeled_looper).to(device)\n",
        "        xu = unlabeled_x.to(device)\n",
        "\n",
        "        if len(xp) != batchsize:\n",
        "            continue\n",
        "        if len(xu) != batchsize:\n",
        "            continue\n",
        "\n",
        "        dis_p.zero_grad()\n",
        "        dis_n.zero_grad()\n",
        "        dis_u.zero_grad()\n",
        "        gen_p.zero_grad()\n",
        "        gen_n.zero_grad()\n",
        "\n",
        "        label = torch.full((batchsize,), real_label, dtype=torch.float32, device=device)\n",
        "\n",
        "        # update discriminator (positive)\n",
        "        label.fill_(real_label)\n",
        "        out = dis_p(xp).view(-1)\n",
        "        errDp_real = criterion(out, label)\n",
        "        errDp_real.backward()\n",
        "        Dp_x = out.mean().item()\n",
        "        \n",
        "        label.fill_(fake_label)\n",
        "        fake_p = gen_p(z)\n",
        "        out = dis_p(fake_p.detach()).view(-1)\n",
        "        errDp_fake = criterion(out, label)\n",
        "        errDp_fake.backward()\n",
        "        Dp_Gp_z1 = out.mean().item()\n",
        "\n",
        "        errDp = errDp_real + errDp_fake\n",
        "        dp_optimizer.step()\n",
        "        \n",
        "        # update discriminator (negative)\n",
        "        label.fill_(real_label)\n",
        "        out = dis_n(xp).view(-1)\n",
        "        errDn_real = criterion(out, label)\n",
        "        errDn_real.backward()\n",
        "        Dn_x = out.mean().item()\n",
        "\n",
        "        label.fill_(fake_label)\n",
        "        fake_n = gen_n(z)\n",
        "        out = dis_n(fake_n.detach()).view(-1)\n",
        "        errDn_fake = criterion(out, label)\n",
        "        errDn_fake.backward()\n",
        "        Dn_Gn_z1 = out.mean().item()\n",
        "\n",
        "        errDn = errDn_real + errDn_fake\n",
        "        dn_optimizer.step()\n",
        "\n",
        "        # update discriminator (unlabeled)\n",
        "        label.fill_(real_label)\n",
        "        out = dis_u(xu).view(-1)\n",
        "        errDu_real = criterion(out, label)\n",
        "        errDu_real.backward()\n",
        "        Du_x = out.mean().item()\n",
        "\n",
        "        label.fill_(fake_label)\n",
        "        out1 = dis_u(fake_p.detach()).view(-1)\n",
        "        out2 = dis_u(fake_n.detach()).view(-1)\n",
        "        errDu_fake = criterion(out1, label) + criterion(out2, label)\n",
        "        errDu_fake.backward()\n",
        "        du_optimizer.step()\n",
        "\n",
        "        Du_Gp_z1 = out1.mean().item()\n",
        "        Du_Gn_z1 = out2.mean().item()\n",
        "        errDu = errDu_real + errDu_fake\n",
        "\n",
        "        # update generator (positive)\n",
        "        label.fill_(real_label)\n",
        "        out1 = dis_p(fake_p).view(-1)\n",
        "        out2 = dis_u(fake_p).view(-1)\n",
        "        errGp_real= criterion(out1, label) + criterion(out2, label)\n",
        "        errGp_real.backward()\n",
        "        gp_optimizer.step()\n",
        "\n",
        "        Dp_Gp_z2 = out1.mean().item()\n",
        "        Du_Gp_z2 = out2.mean().item()\n",
        "        \n",
        "        # update generator (negative)\n",
        "        label.fill_(real_label)\n",
        "        label2 = torch.full((batchsize,), fake_label, dtype=torch.float32, device=device)\n",
        "        out1 = dis_n(fake_n).view(-1)\n",
        "        out2 = dis_u(fake_n).view(-1)\n",
        "        errGn_real= criterion(out1, label2) + criterion(out2, label)\n",
        "        errGn_real.backward()\n",
        "        gn_optimizer.step()\n",
        "\n",
        "        Dn_Gn_z2 = out1.mean().item()\n",
        "        Du_Gn_z2 = out2.mean().item()\n",
        "        \n",
        "        total_errD = (errDp + errDn + errDu).item()\n",
        "        total_errG = (errGp_real + errGn_real).item()\n",
        "        G_loss += total_errG\n",
        "        D_loss += total_errD\n",
        "    G_loss /= len(unlabeled_loader)\n",
        "    D_loss /= len(unlabeled_loader)\n",
        "\n",
        "    loss_D_epoch.append(D_loss)\n",
        "    loss_G_epoch.append(G_loss)\n",
        "    if epoch % 10 == 0:\n",
        "        print(f'[{epoch}/{num_epoch}]: LossD={D_loss}, LossG={G_loss}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Z8cfpqLTg7HI",
        "outputId": "728a02a4-7d4f-4104-a738-ff56cc3b4e1d"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[0/1000]: LossD=3.388206739683409, LossG=2.275372724275331\n",
            "[10/1000]: LossD=2.492092074574651, LossG=2.4048326079909867\n",
            "[20/1000]: LossD=2.6108345276600606, LossG=2.103841446541451\n",
            "[30/1000]: LossD=2.3222572545747497, LossG=2.196898067319715\n",
            "[40/1000]: LossD=2.204339755548013, LossG=2.2987127948451684\n",
            "[50/1000]: LossD=2.2660351637247445, LossG=2.2679632612176843\n",
            "[60/1000]: LossD=2.284652039811418, LossG=2.1865467251958073\n",
            "[70/1000]: LossD=2.41138590993108, LossG=2.2090606882765487\n",
            "[80/1000]: LossD=2.3082787926132613, LossG=2.168390312710324\n",
            "[90/1000]: LossD=2.15762811093717, LossG=2.4749279279966614\n",
            "[100/1000]: LossD=2.2759577196997567, LossG=2.269512969094354\n",
            "[110/1000]: LossD=2.348401701128161, LossG=2.1137513598880253\n",
            "[120/1000]: LossD=2.424440255036225, LossG=2.1881272663941256\n",
            "[130/1000]: LossD=2.335577765026608, LossG=2.3706368948962235\n",
            "[140/1000]: LossD=2.3171250562410095, LossG=2.163456175778363\n",
            "[150/1000]: LossD=2.348749173654092, LossG=2.249324573052896\n",
            "[160/1000]: LossD=2.2725938204172493, LossG=2.4237179949476912\n",
            "[170/1000]: LossD=2.323837074073585, LossG=2.2476826745110587\n",
            "[180/1000]: LossD=2.35672326990076, LossG=2.2209077525783227\n",
            "[190/1000]: LossD=2.177343207436639, LossG=2.2630188078493685\n",
            "[200/1000]: LossD=2.305899091669031, LossG=2.198311999037459\n",
            "[210/1000]: LossD=2.376764407029023, LossG=2.3273822101386816\n",
            "[220/1000]: LossD=2.3251157902382515, LossG=2.205281367173066\n",
            "[230/1000]: LossD=2.3369754713934823, LossG=2.2292714183394975\n",
            "[240/1000]: LossD=2.419139153248555, LossG=2.2116278828801335\n",
            "[250/1000]: LossD=2.3473924881703145, LossG=2.1788235548380257\n",
            "[260/1000]: LossD=2.3510435594094767, LossG=2.2045205348246806\n",
            "[270/1000]: LossD=2.3430652360658386, LossG=2.1658503300434835\n",
            "[280/1000]: LossD=2.3417562278541357, LossG=2.1870782117585876\n",
            "[290/1000]: LossD=2.358000220479192, LossG=2.210941301809775\n",
            "[300/1000]: LossD=2.386047627474811, LossG=2.1804489316167057\n",
            "[310/1000]: LossD=2.330101006739848, LossG=2.2832680264034786\n",
            "[320/1000]: LossD=2.40895364091203, LossG=2.1384200147680335\n",
            "[330/1000]: LossD=2.3706275643529118, LossG=2.1589943073891305\n",
            "[340/1000]: LossD=2.3785402001561344, LossG=2.1595866680145264\n",
            "[350/1000]: LossD=2.4039228477993526, LossG=2.18018998326482\n",
            "[360/1000]: LossD=2.4153587367083573, LossG=2.145782374046944\n",
            "[370/1000]: LossD=2.4464343882895805, LossG=2.1217325829170846\n",
            "[380/1000]: LossD=2.4213873373495565, LossG=2.1626996542956376\n",
            "[390/1000]: LossD=2.3942286130544304, LossG=2.150638187253797\n",
            "[400/1000]: LossD=2.39462398193978, LossG=2.148933268882133\n",
            "[410/1000]: LossD=2.39590521760889, LossG=2.150245866260013\n",
            "[420/1000]: LossD=2.3866932585432723, LossG=2.1418694805454566\n",
            "[430/1000]: LossD=2.3890405281169995, LossG=2.1397210005167366\n",
            "[440/1000]: LossD=2.395204988685814, LossG=2.129044481225916\n",
            "[450/1000]: LossD=2.388930855570613, LossG=2.1421068101315885\n",
            "[460/1000]: LossD=2.3844423229629927, LossG=2.14595094886986\n",
            "[470/1000]: LossD=2.389974639222429, LossG=2.153272648115416\n",
            "[480/1000]: LossD=2.393867183376003, LossG=2.1331787495999723\n",
            "[490/1000]: LossD=2.3883060571309684, LossG=2.1247866798091577\n",
            "[500/1000]: LossD=2.3900644521455505, LossG=2.1562515336114005\n",
            "[510/1000]: LossD=2.3905791269766317, LossG=2.1474778007816626\n",
            "[520/1000]: LossD=2.3946726322174072, LossG=2.1436899739342765\n",
            "[530/1000]: LossD=2.3878942695823877, LossG=2.1365331314705513\n",
            "[540/1000]: LossD=2.3841803911569954, LossG=2.142592771633251\n",
            "[550/1000]: LossD=2.39182287937886, LossG=2.1307839960665316\n",
            "[560/1000]: LossD=2.384556074400206, LossG=2.1312781282373376\n",
            "[570/1000]: LossD=2.3868360648284086, LossG=2.149122038403073\n",
            "[580/1000]: LossD=2.3931683269706934, LossG=2.125475786827706\n",
            "[590/1000]: LossD=2.39575855152027, LossG=2.143769915039475\n",
            "[600/1000]: LossD=2.387794597728832, LossG=2.1297278468673295\n",
            "[610/1000]: LossD=2.3861030307975977, LossG=2.1194610917890393\n",
            "[620/1000]: LossD=2.3920708991385795, LossG=2.144439594165699\n",
            "[630/1000]: LossD=2.390412627039729, LossG=2.1400481623572274\n",
            "[640/1000]: LossD=2.382832656035552, LossG=2.162938388618263\n",
            "[650/1000]: LossD=2.3869196659809835, LossG=2.1474414000640043\n",
            "[660/1000]: LossD=2.3857653978708626, LossG=2.148450754784249\n",
            "[670/1000]: LossD=2.3852485063913704, LossG=2.144287077156273\n",
            "[680/1000]: LossD=2.3880380939792945, LossG=2.1455821475467167\n",
            "[690/1000]: LossD=2.3914661922970333, LossG=2.1531349839390934\n",
            "[700/1000]: LossD=2.3890093468331, LossG=2.159199946635478\n",
            "[710/1000]: LossD=2.3859136233458647, LossG=2.1356147108851253\n",
            "[720/1000]: LossD=2.3816367806615055, LossG=2.140126402313645\n",
            "[730/1000]: LossD=2.3831623889304496, LossG=2.1289188024160026\n",
            "[740/1000]: LossD=2.3916452188749573, LossG=2.1421393897082353\n",
            "[750/1000]: LossD=2.3777831051800704, LossG=2.1307225356230863\n",
            "[760/1000]: LossD=2.388679349744642, LossG=2.1387329165999955\n",
            "[770/1000]: LossD=2.382695217390318, LossG=2.1452743040548787\n",
            "[780/1000]: LossD=2.380645887271778, LossG=2.1385198348277323\n",
            "[790/1000]: LossD=2.377046726845406, LossG=2.1520331228101575\n",
            "[800/1000]: LossD=2.389324561969654, LossG=2.1328924088864714\n",
            "[810/1000]: LossD=2.374615089313404, LossG=2.150454753154033\n",
            "[820/1000]: LossD=2.3815147361239872, LossG=2.156162474606488\n",
            "[830/1000]: LossD=2.38041986645879, LossG=2.140254342878187\n",
            "[840/1000]: LossD=2.383572513992722, LossG=2.146272169577109\n",
            "[850/1000]: LossD=2.377396416019749, LossG=2.14558794691756\n",
            "[860/1000]: LossD=2.378191715962178, LossG=2.146513687597739\n",
            "[870/1000]: LossD=2.3842395834020667, LossG=2.1515009789853483\n",
            "[880/1000]: LossD=2.373801476246602, LossG=2.1362430598284745\n",
            "[890/1000]: LossD=2.3824863176088074, LossG=2.138340473175049\n",
            "[900/1000]: LossD=2.3709195755623482, LossG=2.139899801563572\n",
            "[910/1000]: LossD=2.385663734899985, LossG=2.1443545238391772\n",
            "[920/1000]: LossD=2.3838687716303646, LossG=2.1580441578014478\n",
            "[930/1000]: LossD=2.3776109283034867, LossG=2.137422078364604\n",
            "[940/1000]: LossD=2.384124472334578, LossG=2.170952661617382\n",
            "[950/1000]: LossD=2.3781667528925716, LossG=2.131013283858428\n",
            "[960/1000]: LossD=2.390983194918246, LossG=2.1413342115041374\n",
            "[970/1000]: LossD=2.3811894623008936, LossG=2.1439165038031502\n",
            "[980/1000]: LossD=2.3931091927193306, LossG=2.1471573211051322\n",
            "[990/1000]: LossD=2.376787791381011, LossG=2.1347341859662854\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.plot(loss_D_epoch)\n",
        "plt.plot(loss_G_epoch)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "id": "GKOxnsDi_yvo",
        "outputId": "cd8340a0-c1f5-4a1d-a690-40aaddd634a6"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7f3b80339610>]"
            ]
          },
          "metadata": {},
          "execution_count": 14
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "with torch.no_grad():\n",
        "    z = torch.randn(256, input_size, device=device)\n",
        "    out_p_fake = gen_p(z).cpu().numpy()\n",
        "    out_n_fake = gen_n(z).cpu().numpy()\n",
        "    xp = torch.tensor(labeled_dataset[0:256]).to(device)\n",
        "    out_p_real = xp.cpu().numpy()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QLTegv52mKx-",
        "outputId": "e5359ee8-85a1-463c-9aa6-654af7f0b990"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:5: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).\n",
            "  \"\"\"\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "plt.scatter(out_p_fake[:,0], out_p_fake[:,1])\n",
        "plt.scatter(out_n_fake[:,0], out_n_fake[:,1])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "id": "FZB2uiQ9RrBs",
        "outputId": "4994b79d-1b32-45a7-d5c6-a8cf2909ea7b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.collections.PathCollection at 0x7f3b8029fc50>"
            ]
          },
          "metadata": {},
          "execution_count": 16
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "Kz6tzAJlbLi_"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}