simple-gan.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "04d98c3a-df39-474d-8cb4-294181f94b49",
   "metadata": {},
   "source": [
    "# Generative Adversarial Networks (GANs)\n",
    "\n",
    "GANs are a pair of generator and discriminator models. The **generator** is responsible for **generating new samples** from the domain, and the **discriminator** is responsible for **classifying** whether samples are real or fake (generated). Importantly, the discriminator model is used to update both the model weights of the discriminator itself and the generator model. This means that **the generator never actually sees examples from the domain** and is adapted based on how well the discriminator predicts.\n",
    "\n",
    "Paper that introduced GAN: [Generative Adversarial Nets](https://arxiv.org/pdf/1406.2661.pdf)\n",
    "\n",
    "## Very Simple One Layer GAN\n",
    "\n",
    "In order to understand the basic concept of GANs, we build a pair of simple one-layer GANs which generate simple 2x2 black and white images.\n",
    "\n",
    "Inspired by: [gans by Luis Serrano](https://github.com/luisguiserrano/gans)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "a177476d-14d6-45f7-b70a-cf3dcbeae60b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import mlplot"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8ec17b05-ba29-40e3-ac09-48cf7c7652bd",
   "metadata": {},
   "source": [
    "### Real Data\n",
    "\n",
    "Let us suppose, our real data are `2x2` squared images with the top-left and bottom-right square being black:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "9edc8a60-dfb1-4940-a992-0e799ce87a33",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x144 with 5 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "real_data = [np.array([1.0, 0.0, 0.0, 1.0]),\n",
    "             np.array([0.9, 0.1, 0.2, 0.8]),\n",
    "             np.array([0.9, 0.2, 0.1, 0.8]),\n",
    "             np.array([0.8, 0.1, 0.2, 0.9]),\n",
    "             np.array([0.8, 0.2, 0.1, 0.9])]\n",
    "    \n",
    "mlplot.plot_gan_samples(real_data);"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6c35e436-71f9-4fc5-86f1-9c34db8d3c34",
   "metadata": {},
   "source": [
    "### Exemplarey Noise Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "a45b7dc8-0876-489a-a800-146aaea89178",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAj8AAABBCAYAAAA+G8WvAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAADsklEQVR4nO3csUpcCRQG4HOXgEJmmiFRgpEdiH0EfYA8QCq7tClshZCUeZHYmDdIo42FVlamiIV1YhENO4YpRIQh3G22WrLOFfbsRs/3lXL4OYeZO/w4ME3btgEAUMVv//cCAAD/JeUHAChF+QEASlF+AIBSlB8AoBTlBwAo5d5NhgeDQbuwsJC1S0wmk7Tss7OzGI/HzXUz9+/fbweDQdoO8/PzadkRER8/fhy1bfvwupkHDx60w+Ewc4e07IiItm2vfQ0jImZnZ9ter5e2w/n5eVp2xK9x48XFRVr2ZDKJHz9+XHtj0zSpv8Hx9OnTzPj49OnT1Gcx+8bl5eW07JOTkzg/P5/6Ph0MBu3i4mLaHkdHR2nZEd2exezXcWVlJS378+fPMRqNrr1xdna27ff7aTtkG41GP30Wb1R+FhYW4sOHD//aUn/37du3tOyXL19OnRkMBrGxsZG2w+vXr9OyIyKapvkybWY4HMbh4WHmDmnZXfV6vXj+/Hla/vv379Oyu8q+8eDgIC375OQkLbur3d3d1Py5ubmpz2K2/f39tOxnz551mltcXIydnZ20PR4/fpyW/avI/LxeXV2dOtPv92NtbS1th+zfGtzc3Pzps+hrLwCgFOUHAChF+QEASlF+AIBSlB8AoBTlBwAoRfkBAEpRfgCAUpQfAKAU5QcAKEX5AQBKUX4AgFKUHwCgFOUHAChF+QEASlF+AIBSlB8AoBTlBwAoRfkBAEpRfgCAUpQfAKCUezcZnpmZiSdPnmTtEktLS2nZXYzH49je3k7Lf/PmTVp2V1+/fo23b9+m5c/NzaVlf//+vdPccDiMra2ttD1evXqVlv3ixYtOc9k3Nk2Tlt3FyspKHB4epuX3+/207K6yb9zb20vLvry87DQ3mUzi9PQ0bY+2bdOyV1dXO809evQo1tfX0/Y4Pj5Oy766upo6MxqN4t27d2k7ZL6GERGbm5s//bv//AAApSg/AEApyg8AUIryAwCUovwAAKUoPwBAKcoPAFCK8gMAlKL8AAClKD8AQCnKDwBQivIDAJSi/AAApSg/AEApyg8AUIryAwCUovwAAKUoPwBAKcoPAFCK8gMAlKL8AAClKD8AQCnKDwBQStO2bffhpvkjIr7krZPq97ZtH143cMvvi7j7N069L8KNt8Bdf59G3P0bvU//ctdvvOX3RfzDjTcqPwAAt52vvQCAUpQfAKAU5QcAKEX5AQBKUX4AgFKUHwCgFOUHAChF+QEASlF+AIBS/gRmJu04inUQjAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x144 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "noise = [np.random.randn(2,2) for i in range(10)]\n",
    "mlplot.plot_gan_samples(noise);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "8430b34b-6f78-4961-b14b-91dfd275047f",
   "metadata": {},
   "outputs": [],
   "source": [
    "def sigmoid(x):\n",
    "    return np.exp(x) / (1.0 + np.exp(x))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b81ccac0-f10d-4ba9-89ce-c09ca71b291f",
   "metadata": {},
   "source": [
    "### The Discriminator\n",
    "\n",
    "The discriminator in a GAN is a **classifier** $D(x, w, c)$ which models the **probability**\n",
    "\n",
    "$$D(x) = \\sigma(x_1 w_1 + x_2 w_2 + x_3 w_3 + x_4 w_4 + c)$$\n",
    "\n",
    "that the data $x$ came from the **real dataset** or not. \n",
    "\n",
    "This is a **binary classification problem**.\n",
    "\n",
    "![GAN Discriminator](images/gan-discriminator.svg)\n",
    "\n",
    "It’s weights $w$ are updated as to **maximize** the probability $D(x)$ that any **real** data input $x$ is classified as belonging to the **real** dataset, while **minimizing** the probability $D(x) = D(G(z))$ that any **fake** data $G(z)$ is classified as belonging to the **real** dataset.\n",
    "\n",
    "For the **real data** $x_r$, we want the discriminator to **identify real** data, hence we want the **prediction to be $1$**, so we use the **loss function**\n",
    "\n",
    "$$E = -ln(D(x_r))$$\n",
    "\n",
    "with its derivative\n",
    "\n",
    "$$\n",
    "\\begin{aligned}\n",
    "   \\frac{\\partial E}{\\partial w_i} &= \\frac{\\partial E}{\\partial D} \\frac{\\partial D}{\\partial w_i} \\\\[5pt]\n",
    "                                   &= \\frac{-1}{D(x)} \\cdot D(x) \\, \\big(1 - D(x) \\big) \\, x_i \\\\[5pt]\n",
    "                                   &= - \\big(1 - D(x) \\big) \\, x_i\n",
    "\\end{aligned}\n",
    "$$\n",
    "\n",
    "For **fake data** $x_f = G(z)$, we want the discriminator to **identify** the data as being **not real** hence we want the **prediction to be $0$**, so we use the **loss function**\n",
    "\n",
    "$$E = -ln(1 - D(x_f))$$\n",
    "\n",
    "with its derivative\n",
    "\n",
    "$$\n",
    "\\begin{aligned}\n",
    "   \\frac{\\partial E}{\\partial w_i} &= \\frac{\\partial E}{\\partial D} \\frac{\\partial D}{\\partial w_i} \\\\[5pt]\n",
    "                                   &= \\frac{1}{1 - D(x)} \\cdot D(x) \\, \\big(1 - D(x) \\big) \\, x_i \\\\[5pt]\n",
    "                                   &= D(x) \\, x_i\n",
    "\\end{aligned}\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "dc5a307c-edac-4ba8-adbf-1decacaed83d",
   "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": [
    "mlplot.plot_log();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "3f50d1b5-aab3-4656-be0d-c91c819a4334",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Discriminator:\n",
    "    def __init__(self, lr):\n",
    "        self.weights = np.array([np.random.normal() for i in range(4)])\n",
    "        self.bias = np.random.normal()\n",
    "        self.lr = lr\n",
    "    \n",
    "    def forward(self, x):\n",
    "        return sigmoid(np.dot(x, self.weights) + self.bias)\n",
    "    \n",
    "    def error_real_data(self, data):\n",
    "        prediction = self.forward(data)\n",
    "        return - np.log(prediction)\n",
    "    \n",
    "    def backward_real_data(self, data):\n",
    "        prediction = self.forward(data)\n",
    "        weights = - (1 - prediction) * data\n",
    "        bias = - (1 - prediction)\n",
    "        self.weights -= self.lr * weights\n",
    "        self.bias -= self.lr * bias\n",
    "\n",
    "    def error_fake_data(self, data):\n",
    "        prediction = self.forward(data)\n",
    "        return - np.log(1 - prediction)\n",
    "    \n",
    "    def backward_fake_data(self, data):\n",
    "        prediction = self.forward(data)\n",
    "        weights = data * prediction\n",
    "        bias = prediction\n",
    "        self.weights -= self.lr * weights\n",
    "        self.bias -= self.lr * bias "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4ebf2d04-2ff9-4b63-9696-037d3e8c0656",
   "metadata": {},
   "source": [
    "### The Generator\n",
    "\n",
    "The generator is trained to fool the discriminator by generating data as similar as possible to the real data. \n",
    "\n",
    "The generators weights $v$ are updated as to **maximize** the probability $D(G(z))$ that any **fake** data $G(z)$ is classified by the discriminator as belonging to the **real** dataset.\n",
    "\n",
    "![GAN Generator](images/gan-generator.svg)\n",
    "\n",
    "For the **fake data** $G(z) = (g_1, g_2, g_3, g_4)$ with $g_i = \\sigma(v_i z + b_i)$, we want the **prediction** of the **discriminator to be $1$**, so we use the **loss function**\n",
    "\n",
    "$$E = -ln(D(G(z)))$$\n",
    "\n",
    "with\n",
    "\n",
    "$$\n",
    "\\begin{aligned}\n",
    "G(z) &= (g_1, g_2, g_3, g_4) \\\\\n",
    "g_i &= \\sigma(v_i z + b_i) \\\\\n",
    "D(G(z)) &= \\sigma \\big(\\sigma(v_1 z + c_1) \\cdot w_1 + \\sigma(v_2 z + c_2) \\cdot w_2 + \\sigma(v_3 z + c_3) \\cdot w_3 + \\sigma(v_4 z + c_4) \\cdot w_4 + c\\big)\n",
    "\\end{aligned}\n",
    "$$\n",
    "\n",
    "and its derivatives\n",
    "\n",
    "$$\n",
    "\\begin{aligned}\n",
    "   \\frac{\\partial E}{\\partial v_i} &= \\frac{\\partial E}{\\partial D} \\frac{\\partial D}{\\partial G} \\frac{\\partial G}{\\partial v_i} = - \\big(1 - D(G(z)) \\big) \\cdot g_i (1 - g_i) \\cdot z \\\\[5pt]\n",
    "   \\frac{\\partial E}{\\partial b_i} &= \\frac{\\partial E}{\\partial D} \\frac{\\partial D}{\\partial G} \\frac{\\partial G}{\\partial b_i} = - \\big(1 - D(G(z)) \\big) \\cdot g_i (1 - g_i) \\\\[5pt]\n",
    "\\end{aligned}\n",
    "$$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "02dc859b-d173-47c4-84f9-0d137541d3cf",
   "metadata": {},
   "outputs": [],
   "source": [
    "class Generator:\n",
    "    def __init__(self, lr):\n",
    "        self.weights = np.array([np.random.normal() for i in range(4)])\n",
    "        self.biases = np.array([np.random.normal() for i in range(4)])\n",
    "        self.lr = lr\n",
    "\n",
    "    def forward(self, z):\n",
    "        return sigmoid(z * self.weights + self.biases)\n",
    "\n",
    "    def error(self, z, discriminator):\n",
    "        data = self.forward(z)\n",
    "        prediction = discriminator.forward(data)\n",
    "        return - np.log(prediction)\n",
    "\n",
    "    def backward(self, z, discriminator):\n",
    "        data = self.forward(z)\n",
    "        prediction = discriminator.forward(data)\n",
    "        factor = - (1 - prediction) * discriminator.weights * data * (1 - data)\n",
    "        self.weights -= self.lr * factor * z\n",
    "        self.biases -= self.lr * factor"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "edcb0273-235f-4258-88fd-ec88078ae4cb",
   "metadata": {},
   "source": [
    "### Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "069c29df-9933-4654-902c-b1fa8477ec74",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7a171576af7c47299a7eb78cc7f1721c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/1000 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from tqdm.notebook import trange\n",
    "\n",
    "np.random.seed(3583795213)\n",
    "\n",
    "# hyperparameters\n",
    "learning_rate = 0.01\n",
    "epochs = 1000\n",
    "\n",
    "# the GAN\n",
    "D = Discriminator(learning_rate)\n",
    "G = Generator(learning_rate)\n",
    "\n",
    "# error plots\n",
    "discriminator_errors = []\n",
    "generator_errors = []\n",
    "\n",
    "# generated images plot\n",
    "images = []\n",
    "\n",
    "for epoch in trange(epochs):\n",
    "    \n",
    "    discriminator_error = []\n",
    "    generator_error = []\n",
    "    fake_datas = []\n",
    "    \n",
    "    for data in real_data:\n",
    "        \n",
    "        # update the discriminator weights from the real data\n",
    "        D.backward_real_data(data)\n",
    "    \n",
    "        # use some noise for input\n",
    "        z = np.random.randn()\n",
    "\n",
    "        # calculate discriminator error for plot\n",
    "        error = np.sum(D.error_real_data(data) + D.error_fake_data(z))\n",
    "        discriminator_error.append(error)\n",
    "        \n",
    "        # calculate generator error for plot\n",
    "        error = G.error(z, D)\n",
    "        generator_error.append(error)\n",
    "        \n",
    "        # generate fake data\n",
    "        fake_data = G.forward(z)\n",
    "        fake_datas.append(fake_data)\n",
    "        \n",
    "        # update the discriminator weights from the fake data\n",
    "        D.backward_fake_data(fake_data)\n",
    "    \n",
    "        # update the generator weights from the fake data\n",
    "        G.backward(z, D)\n",
    "        \n",
    "    discriminator_errors.append(np.mean(discriminator_error))\n",
    "    generator_errors.append(np.mean(generator_error))\n",
    "    \n",
    "    if epoch % 10 == 0:\n",
    "        figs = mlplot.plot_generator(fake_datas, epoch=epoch)\n",
    "        image = mlplot.fig2img(figs[0])\n",
    "        images.append(image)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "00c00e99-f32a-48fd-9be8-1750c423f85b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mlplot.plot_errors(generator_errors, discriminator_errors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "870586c5-d2de-423e-866c-c8056ea58e69",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Weights: [-0.41938194  0.11301774 -0.03043566  0.1294328 ]\n",
      "Biases:  [ 2.07643064 -3.14111999 -3.32244774  2.51017606]\n"
     ]
    }
   ],
   "source": [
    "print(f'Weights: {G.weights}\\nBiases:  {G.biases}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "38749315-7e1d-4caf-a602-527f20afc4b6",
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import Image\n",
    "\n",
    "images[0].save('images/simple_gan.gif', save_all=True, append_images=images[1:], optimize=False, duration=80, loop=0)\n",
    "#Image(filename='images/simple_gan.gif')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0d346bbc-456d-4633-9e6c-d4ac0c624301",
   "metadata": {},
   "source": [
    "![gan](images/simple_gan.gif)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b96a1427-21b0-4116-9612-1572f14e19de",
   "metadata": {},
   "source": [
    "### Generated Data\n",
    "\n",
    "Noise data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "ad15c520-f87c-40c0-9fee-8a83a3a76c63",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x144 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "noise = [np.random.randn(2,2) for i in range(10)]\n",
    "mlplot.plot_gan_samples(noise);"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "36975ca3-336d-41a6-92e7-f5347e7146cb",
   "metadata": {},
   "source": [
    "Generated data of trained generator with noise data as input:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "06937d06-614e-4ff4-9cd2-3fa0eb3708ba",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x144 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "generated_data = [sigmoid(np.random.randn() * G.weights + G.biases) for i in range(10)]\n",
    "mlplot.plot_gan_samples(generated_data);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "22fdd0bf-b7c9-4201-ba4d-6bd979a92890",
   "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": "3.10.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}