In this post, we will take a closer look at Generative Adversarial Networks (GAN)<\/strong>. GAN is a generative model proposed by Ian Goodfellow in 2014, which uses two neural networks (Generator and Discriminator) to generate data. The key aspect of GAN that we focus on is that the two neural networks compete with each other, which allows for the generation of more advanced data.<\/p>\n GAN consists of the following two components:<\/p>\n The Generator and Discriminator are trained through the following loss functions:<\/p>\n The training process of the GAN model consists of the following steps:<\/p>\n Now, let’s implement a simple GAN using PyTorch. In this example, we will implement a GAN model that generates digit images using the MNIST dataset.<\/p>\n GANs are used not only for image generation but also in various fields:<\/p>\n1. Basic Structure of GAN<\/h2>\n
\n
\n
2. Training Process of GAN<\/h2>\n
\n
3. Implementing GAN Using PyTorch<\/h2>\n
3.1 Installing Required Libraries<\/h3>\n
\n# Install required libraries\n!pip install torch torchvision matplotlib\n<\/code><\/pre>\n3.2 Loading and Preprocessing the Dataset<\/h3>\n
\nimport torch\nimport torchvision\nimport torchvision.transforms as transforms\nimport matplotlib.pyplot as plt\n\n# Download and preprocess the MNIST dataset\ntransform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.5,), (0.5,))])\ntrain_set = torchvision.datasets.MNIST(root='.\/data', train=True, download=True, transform=transform)\ntrain_loader = torch.utils.data.DataLoader(train_set, batch_size=64, shuffle=True)\n<\/code><\/pre>\n3.3 Defining Generator and Discriminator Models<\/h3>\n
\nimport torch.nn as nn\n\n# Define Generator model\nclass Generator(nn.Module):\n def __init__(self):\n super(Generator, self).__init__()\n self.fc = nn.Sequential(\n nn.Linear(100, 256),\n nn.ReLU(),\n nn.Linear(256, 512),\n nn.ReLU(),\n nn.Linear(512, 1024),\n nn.ReLU(),\n nn.Linear(1024, 28 * 28),\n nn.Tanh(),\n )\n\n def forward(self, x):\n x = self.fc(x)\n return x.view(-1, 1, 28, 28)\n\n# Define Discriminator model\nclass Discriminator(nn.Module):\n def __init__(self):\n super(Discriminator, self).__init__()\n self.fc = nn.Sequential(\n nn.Linear(28 * 28, 512),\n nn.LeakyReLU(0.2),\n nn.Linear(512, 256),\n nn.LeakyReLU(0.2),\n nn.Linear(256, 1),\n nn.Sigmoid(),\n )\n\n def forward(self, x):\n x = x.view(-1, 28 * 28)\n return self.fc(x)\n<\/code><\/pre>\n3.4 Model Training<\/h3>\n
\n# Setting hyperparameters\nnum_epochs = 200\nlearning_rate = 0.0002\nbeta1 = 0.5\n\n# Initialize models\ngenerator = Generator()\ndiscriminator = Discriminator()\n\n# Define loss function and optimization algorithm\ncriterion = nn.BCELoss()\noptimizerG = torch.optim.Adam(generator.parameters(), lr=learning_rate, betas=(beta1, 0.999))\noptimizerD = torch.optim.Adam(discriminator.parameters(), lr=learning_rate, betas=(beta1, 0.999))\n\n# Training loop\nfor epoch in range(num_epochs):\n for i, (data, _) in enumerate(train_loader):\n # Setting labels for real and fake data\n real_labels = torch.ones(data.size(0), 1)\n fake_labels = torch.zeros(data.size(0), 1)\n\n # Training Discriminator\n optimizerD.zero_grad()\n outputs = discriminator(data)\n lossD_real = criterion(outputs, real_labels)\n lossD_real.backward()\n\n noise = torch.randn(data.size(0), 100)\n fake_data = generator(noise)\n outputs = discriminator(fake_data.detach())\n lossD_fake = criterion(outputs, fake_labels)\n lossD_fake.backward()\n optimizerD.step()\n\n # Training Generator\n optimizerG.zero_grad()\n outputs = discriminator(fake_data)\n lossG = criterion(outputs, real_labels)\n lossG.backward()\n optimizerG.step()\n\n if (epoch+1) % 10 == 0:\n print(f'Epoch [{epoch+1}\/{num_epochs}], Loss D: {lossD_real.item() + lossD_fake.item():.4f}, Loss G: {lossG.item():.4f}')\n<\/code><\/pre>\n3.5 Visualizing Results<\/h3>\n
\n# Function to visualize generated images\ndef visualize(generator):\n noise = torch.randn(64, 100)\n fake_data = generator(noise)\n fake_data = fake_data.detach().numpy()\n fake_data = (fake_data + 1) \/ 2 # Normalize to [0, 1]\n\n plt.figure(figsize=(8, 8))\n for i in range(fake_data.shape[0]):\n plt.subplot(8, 8, i+1)\n plt.axis('off')\n plt.imshow(fake_data[i][0], cmap='gray')\n plt.show()\n\n# Visualize results\nvisualize(generator)\n<\/code><\/pre>\n4. Applications of GAN<\/h2>\n