\n \uc0dd\uc131\uc801 \uc801\ub300 \uc2e0\uacbd\ub9dd(Generative Adversarial Networks, GAN)\uc740 Geoffrey Hinton, Ian Goodfellow, Yoshua Bengio\uac00 \uc81c\uc548\ud55c
\n \ub525\ub7ec\ub2dd\uc758 \ud601\uc2e0\uc801\uc778 \ubaa8\ub378\ub85c, \uc0dd\uc131 \ubaa8\ub378\uacfc \ud310\ubcc4 \ubaa8\ub378\uc774\ub77c\ub294 \ub450 \uac1c\uc758 \uc2e0\uacbd\ub9dd\uc774 \uacbd\uc7c1\ud558\uba70 \ud559\uc2b5\ud558\ub294 \uad6c\uc870\ub97c \uac00\uc9c0\uace0 \uc788\uc2b5\ub2c8\ub2e4.
\n GAN\uc740 \uc774\ubbf8\uc9c0 \uc0dd\uc131, \ubca1\ud130 \uc774\ubbf8\uc9c0 \ubcc0\ud658, \uc2a4\ud0c0\uc77c \ubcc0\ud658 \ub4f1 \ub2e4\uc591\ud55c \ubd84\uc57c\uc5d0\uc11c \uc0ac\uc6a9\ub418\uace0 \uc788\uc73c\uba70, \uadf8 \uac00\ub2a5\uc131\uc740 \ubb34\uad81\ubb34\uc9c4\ud569\ub2c8\ub2e4.
\n \uadf8\ub7ec\ub098 GAN\uc740 \ub2e4\uc591\ud55c \ub3c4\uc804 \uacfc\uc81c\uc5d0 \uc9c1\uba74\ud574 \uc788\uc2b5\ub2c8\ub2e4. \uc774 \uae00\uc5d0\uc11c\ub294 GAN\uc758 \uae30\ubcf8 \uac1c\ub150 \ubc0f \uad6c\uc870\ub97c \uc124\uba85\ud558\uace0,
\n \ud30c\uc774\ud1a0\uce58(PyTorch)\ub97c \ud65c\uc6a9\ud55c \uae30\ubcf8\uc801\uc778 GAN \uad6c\ud604 \uc608\uc81c\uc640 \ud568\uaed8 \uc5ec\ub7ec \ub3c4\uc804 \uacfc\uc81c\uc5d0 \ub300\ud574 \uc54c\uc544\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n
GAN\uc758 \uae30\ubcf8 \uac1c\ub150<\/h2>\n
\n GAN\uc740 \ub450 \uac1c\uc758 \ub124\ud2b8\uc6cc\ud06c\ub85c \uad6c\uc131\ub429\ub2c8\ub2e4. \uccab \ubc88\uc9f8 \ub124\ud2b8\uc6cc\ud06c\ub294 \ub370\uc774\ud130 \uc0d8\ud50c\uc744 \uc0dd\uc131\ud558\ub294 \uc5ed\ud560\uc744 \ud558\ub294 \uc0dd\uc131\uc790(Generator)\uc774\uba70,
\n \ub450 \ubc88\uc9f8 \ub124\ud2b8\uc6cc\ud06c\ub294 \uc0dd\uc131\ub41c \ub370\uc774\ud130\uc640 \uc2e4\uc81c \ub370\uc774\ud130(\ud6c8\ub828 \ub370\uc774\ud130)\ub97c \uad6c\ubd84\ud558\ub294 \uc5ed\ud560\uc744 \ud558\ub294 \ud310\ubcc4\uc790(Discriminator)\uc785\ub2c8\ub2e4.
\n \uc774 \ub450 \ub124\ud2b8\uc6cc\ud06c\ub294 \uac8c\uc784 \uc774\ub860\uc801 \ub9e5\ub77d\uc5d0\uc11c \uc11c\ub85c \ub300\ub9bd\ud558\ub294 \uad00\uacc4\uc5d0 \uc788\uc2b5\ub2c8\ub2e4. \uc0dd\uc131\uc790\uc758 \ubaa9\ud45c\ub294 \ud310\ubcc4\uc790\ub97c \uc18d\uc5ec\uc11c
\n \uc0dd\uc131\ud55c \ub370\uc774\ud130\uac00 \uc2e4\uc81c \ub370\uc774\ud130\uc640 \uad6c\ubd84\ub418\uc9c0 \uc54a\ub3c4\ub85d \ud558\ub294 \uac83\uc774\uace0, \ud310\ubcc4\uc790\uc758 \ubaa9\ud45c\ub294 \uc0dd\uc131\uc790\uac00 \ub9cc\ub4e0 \ub370\uc774\ud130\ub97c \uc815\ud655\ud788
\n \ubd84\ub958\ud558\ub294 \uac83\uc785\ub2c8\ub2e4.\n <\/p>\n
GAN\uc758 \uad6c\uc870<\/h2>\n\n- \uc0dd\uc131\uc790(Generator)<\/strong>:\n
\ubb34\uc791\uc704 \ub178\uc774\uc988 \ubca1\ud130\ub97c \uc785\ub825\ubc1b\uc544 \uadf8\uac83\uc73c\ub85c\ubd80\ud130 \uc810\ucc28 \uc2e4\uc81c \ub370\uc774\ud130\uc640 \uc720\uc0ac\ud55c \uc0d8\ud50c\uc744 \uc0dd\uc131\ud569\ub2c8\ub2e4.<\/p>\n<\/li>\n
- \ud310\ubcc4\uc790(Discriminator)<\/strong>:\n
\uc9c4\uc9dc \ub370\uc774\ud130\uc640 \uc0dd\uc131\ub41c \ub370\uc774\ud130\ub97c \uc785\ub825\ubc1b\uc544, \uc785\ub825\uc774 \uc9c4\uc9dc\uc778\uc9c0 \uac00\uc9dc\uc778\uc9c0 \uad6c\ubcc4\ud558\ub294 \ud655\ub960\uc744 \ucd9c\ub825\ud569\ub2c8\ub2e4.<\/p>\n<\/li>\n<\/ul>\n
\ud30c\uc774\ud1a0\uce58\ub97c \ud65c\uc6a9\ud55c GAN\uc758 \uad6c\ud604<\/h2>\n
\n \ub2e4\uc74c\uc740 \ud30c\uc774\ud1a0\uce58\ub97c \uc0ac\uc6a9\ud558\uc5ec GAN\uc744 \uad6c\ud604\ud558\ub294 \uac04\ub2e8\ud55c \uc608\uc81c\uc785\ub2c8\ub2e4. MNIST \uc22b\uc790 \ub370\uc774\ud130\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc22b\uc790 \uc774\ubbf8\uc9c0\ub97c \uc0dd\uc131\ud558\ub294
\n GAN \ubaa8\ub378\uc744 \uad6c\ud604\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n
\n \nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torchvision import datasets, transforms\nfrom torchvision.utils import save_image\n\n# \ud558\uc774\ud37c\ud30c\ub77c\ubbf8\ud130 \uc124\uc815\nlatent_size = 64\nbatch_size = 128\nnum_epochs = 100\nlearning_rate = 0.0002\n\n# \ubcc0\ud658 \uc124\uc815 \ubc0f \ub370\uc774\ud130 \ub85c\ub4dc\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\nmnist = datasets.MNIST(root='data\/', train=True, transform=transform, download=True)\ntrain_loader = torch.utils.data.DataLoader(dataset=mnist, batch_size=batch_size, shuffle=True)\n\n# \uc0dd\uc131\uc790 \ubaa8\ub378 \uc815\uc758\nclass Generator(nn.Module):\n def __init__(self):\n super(Generator, self).__init__()\n self.model = nn.Sequential(\n nn.Linear(latent_size, 256),\n nn.ReLU(True),\n nn.Linear(256, 512),\n nn.ReLU(True),\n nn.Linear(512, 1024),\n nn.ReLU(True),\n nn.Linear(1024, 28 * 28),\n nn.Tanh() # \ucd9c\ub825 \ubc94\uc704 [-1, 1]\n )\n\n def forward(self, z):\n return self.model(z).view(z.size(0), 1, 28, 28)\n\n# \ud310\ubcc4\uc790 \ubaa8\ub378 \uc815\uc758\nclass Discriminator(nn.Module):\n def __init__(self):\n super(Discriminator, self).__init__()\n self.model = nn.Sequential(\n nn.Linear(28 * 28, 1024),\n nn.LeakyReLU(0.2),\n nn.Linear(1024, 512),\n nn.LeakyReLU(0.2),\n nn.Linear(512, 256),\n nn.LeakyReLU(0.2),\n nn.Linear(256, 1),\n nn.Sigmoid() # \ucd9c\ub825 \ud655\ub960\n )\n\n def forward(self, img):\n return self.model(img.view(img.size(0), -1))\n\n# \uc0dd\uc131\uc790 \ubc0f \ud310\ubcc4\uc790 \ucd08\uae30\ud654\ngenerator = Generator()\ndiscriminator = Discriminator()\n\n# \uc190\uc2e4 \ud568\uc218 \ubc0f \ucd5c\uc801\ud654 \uae30\ubc95 \uc124\uc815\ncriterion = nn.BCELoss()\noptimizer_G = optim.Adam(generator.parameters(), lr=learning_rate)\noptimizer_D = optim.Adam(discriminator.parameters(), lr=learning_rate)\n\n# \ubaa8\ub378 \ud559\uc2b5\nfor epoch in range(num_epochs):\n for i, (imgs, _) in enumerate(train_loader):\n # \uc9c4\uc9dc \uc774\ubbf8\uc9c0\uc5d0 \ub300\ud55c \ub77c\ubca8\n real_labels = torch.ones(imgs.size(0), 1)\n # \uac00\uc9dc \uc774\ubbf8\uc9c0\uc5d0 \ub300\ud55c \ub77c\ubca8\n fake_labels = torch.zeros(imgs.size(0), 1)\n\n # \ud310\ubcc4\uc790 \ud559\uc2b5\n optimizer_D.zero_grad()\n outputs = discriminator(imgs)\n d_loss_real = criterion(outputs, real_labels)\n \n z = torch.randn(imgs.size(0), latent_size)\n fake_imgs = generator(z)\n outputs = discriminator(fake_imgs.detach())\n d_loss_fake = criterion(outputs, fake_labels)\n\n d_loss = d_loss_real + d_loss_fake\n d_loss.backward()\n optimizer_D.step()\n\n # \uc0dd\uc131\uc790 \ud559\uc2b5\n optimizer_G.zero_grad()\n outputs = discriminator(fake_imgs)\n g_loss = criterion(outputs, real_labels)\n\n g_loss.backward()\n optimizer_G.step()\n\n # \uc774\ubbf8\uc9c0 \uc800\uc7a5\n if (epoch+1) % 10 == 0:\n save_image(fake_imgs.data, f'images\/fake_images-{epoch+1}.png', nrow=8, normalize=True)\n print(f'Epoch [{epoch+1}\/{num_epochs}], d_loss: {d_loss.item():.4f}, g_loss: {g_loss.item():.4f}')\n <\/code>\n <\/pre>\nGAN\uc758 \ub3c4\uc804 \uacfc\uc81c<\/h2>\n
\n GAN\uc740 \uc5ec\ub7ec \uac00\uc9c0 \ub3c4\uc804 \uacfc\uc81c\uc5d0 \uc9c1\uba74\ud574 \uc788\uc2b5\ub2c8\ub2e4. \uc774 \uc139\uc158\uc5d0\uc11c\ub294 \uadf8 \uc911 \uba87 \uac00\uc9c0\ub97c \uc0b4\ud3b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n
1. \ubaa8\ub4dc \ubd95\uad34(Mode Collapse)<\/h3>\n
\n \ubaa8\ub4dc \ubd95\uad34\ub294 \uc0dd\uc131\uc790\uac00 \uc81c\ud55c\ub41c \uc218\uc758 \ucd9c\ub825\ub9cc \uc0dd\uc131\ud558\ub3c4\ub85d \ud559\uc2b5\ud558\uc138\uc694 \ubc1c\uc0dd\ud558\ub294 \ud604\uc0c1\uc785\ub2c8\ub2e4.
\n \uc774\ub294 \uc0dd\uc131\uc790\uac00 \ud558\ub098\uc758 \uc774\ubbf8\uc9c0\ub97c \uc5ec\ub7ec \ubc88 \uc0dd\uc131\ud558\ub294 \uacb0\uacfc\ub97c \ub0b3\uc544 \ub2e4\ubcc0\ud654\ub41c \uacb0\uacfc\ub97c \uc81c\uacf5\ud560 \uc218 \uc5c6\uc2b5\ub2c8\ub2e4.
\n \uc774 \ubb38\uc81c\ub97c \ud574\uacb0\ud558\uae30 \uc704\ud574 \uc5ec\ub7ec \uac00\uc9c0 \uae30\ubc95\uc774 \uc81c\uc548\ub418\uc5c8\uc73c\uba70, \uadf8 \uc911 \ud558\ub098\ub294 \ub2e4\uc218\uc758 \ub2e4\uc591\ud55c \uac00\uc9dc \ub370\uc774\ud130\ub97c \uc0dd\uc131\ud560 \uc218 \uc788\ub3c4\ub85d
\n \ud558\ub294 \uac83\uc785\ub2c8\ub2e4.\n <\/p>\n
2. \ubd88\uc548\uc815\ud55c \ud6c8\ub828<\/h3>\n
\n GAN\uc758 \ud6c8\ub828\uc740 \uc885\uc885 \ubd88\uc548\uc815\ud558\uba70, \ud310\ubcc4\uc790\uc640 \uc0dd\uc131\uc790\uc758 \ud559\uc2b5 \uacfc\uc815\uc774 \ubd88\uade0\ud615\ud558\uba74 \ud6c8\ub828\uc774 \uc81c\ub300\ub85c \uc9c4\ud589\ub418\uc9c0 \uc54a\uc744 \uc218 \uc788\uc2b5\ub2c8\ub2e4.
\n \uc774\ub97c \ud574\uacb0\ud558\uae30 \uc704\ud574 \ub2e4\uc591\ud55c \ucd5c\uc801\ud654 \ubc29\ubc95\uacfc \ud6c8\ub828 \uc804\ub7b5\uc744 \uc0ac\uc6a9\ud558\ub294 \uac83\uc774 \ud544\uc694\ud569\ub2c8\ub2e4.\n <\/p>\n
3. \ubd80\uc815\ud655\ud55c \ud310\ubcc4<\/h3>\n
\n \ud310\ubcc4\uc790\uac00 \ub108\ubb34 \uac15\ub825\ud558\uba74 \uc0dd\uc131\uc790\uac00 \ud559\uc2b5\ud558\ub294 \ub370 \uc5b4\ub824\uc6c0\uc744 \uacaa\uc744 \uc218 \uc788\uc73c\uba70, \uc0dd\uc131\uc790\uac00 \ub108\ubb34 \uc57d\ud558\uba74 \ud310\ubcc4\uc790\uac00
\n \uc27d\uac8c \uc18d\uc774\ub294 \uacb0\uacfc\ub97c \ub0b3\uc744 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc801\uc808\ud55c \ud6c8\ub828 \uade0\ud615\uc744 \uc720\uc9c0\ud558\ub294 \uac83\uc774 \uc911\uc694\ud569\ub2c8\ub2e4.\n <\/p>\n
4. \uace0\ucc28\uc6d0 \uacf5\uac04\uc5d0\uc11c\uc758 \ubb38\uc81c<\/h3>\n
\n GAN\uc758 \ud6c8\ub828\uc740 \uace0\ucc28\uc6d0\uc758 \ub370\uc774\ud130\uc5d0\uc11c \uc9c4\ud589\ub418\uba70, \uc774\ub85c \uc778\ud574 \ud559\uc2b5\uc774 \uc5b4\ub824\uc6cc\uc9c0\ub294 \uacbd\uc6b0\uac00 \ub9ce\uc2b5\ub2c8\ub2e4.
\n \uace0\ucc28\uc6d0 \uacf5\uac04\uc5d0\uc11c\uc758 \ub370\uc774\ud130 \ud2b9\uc131\uc744 \uc798 \uc774\ud574\ud558\uace0 \uc801\uc808\ud55c \ubc29\ubc95\uc73c\ub85c \ubaa8\ub378\uc744 \uc124\uacc4\ud574\uc57c \ud569\ub2c8\ub2e4.\n <\/p>\n
\uacb0\ub860<\/h2>\n
\n GAN\uc740 \ub9e4\uc6b0 \uac15\ub825\ud55c \uc0dd\uc131 \ubaa8\ub378\uc774\uc9c0\ub9cc, \uc5ec\ub7ec \ub3c4\uc804 \uacfc\uc81c\ub97c \uac00\uc9c0\uace0 \uc788\uc2b5\ub2c8\ub2e4. \ud30c\uc774\ud1a0\uce58\ub97c \uc0ac\uc6a9\ud558\uba74 GAN\uc744 \uc27d\uac8c \uad6c\ud604\ud558\uace0
\n \uc2e4\ud5d8\ud560 \uc218 \uc788\uc73c\uba70, \uc774\ub97c \ud1b5\ud574 GAN\uc758 \uc774\ud574\ub3c4\ub97c \uc99d\uc9c4\uc2dc\ud0ac \uc218 \uc788\uc2b5\ub2c8\ub2e4. GAN\uc758 \ubc1c\uc804 \uac00\ub2a5\uc131\uc740 \ubb34\uad81\ubb34\uc9c4\ud558\uba70, \uc55e\uc73c\ub85c \ub354
\n \ub9ce\uc740 \uc5f0\uad6c\uc640 \uac1c\uc120\uc774 \uc774\ub8e8\uc5b4\uc9c8 \uac83\uc785\ub2c8\ub2e4.\n <\/p>\n<\/article>\n
<\/body><\/p>\n","protected":false},"excerpt":{"rendered":"
\ubb34\uc791\uc704 \ub178\uc774\uc988 \ubca1\ud130\ub97c \uc785\ub825\ubc1b\uc544 \uadf8\uac83\uc73c\ub85c\ubd80\ud130 \uc810\ucc28 \uc2e4\uc81c \ub370\uc774\ud130\uc640 \uc720\uc0ac\ud55c \uc0d8\ud50c\uc744 \uc0dd\uc131\ud569\ub2c8\ub2e4.<\/p>\n<\/li>\n
\uc9c4\uc9dc \ub370\uc774\ud130\uc640 \uc0dd\uc131\ub41c \ub370\uc774\ud130\ub97c \uc785\ub825\ubc1b\uc544, \uc785\ub825\uc774 \uc9c4\uc9dc\uc778\uc9c0 \uac00\uc9dc\uc778\uc9c0 \uad6c\ubcc4\ud558\ub294 \ud655\ub960\uc744 \ucd9c\ub825\ud569\ub2c8\ub2e4.<\/p>\n<\/li>\n<\/ul>\n
\ud30c\uc774\ud1a0\uce58\ub97c \ud65c\uc6a9\ud55c GAN\uc758 \uad6c\ud604<\/h2>\n
\n \ub2e4\uc74c\uc740 \ud30c\uc774\ud1a0\uce58\ub97c \uc0ac\uc6a9\ud558\uc5ec GAN\uc744 \uad6c\ud604\ud558\ub294 \uac04\ub2e8\ud55c \uc608\uc81c\uc785\ub2c8\ub2e4. MNIST \uc22b\uc790 \ub370\uc774\ud130\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc22b\uc790 \uc774\ubbf8\uc9c0\ub97c \uc0dd\uc131\ud558\ub294
\n GAN \ubaa8\ub378\uc744 \uad6c\ud604\ud574 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n
\n \nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torchvision import datasets, transforms\nfrom torchvision.utils import save_image\n\n# \ud558\uc774\ud37c\ud30c\ub77c\ubbf8\ud130 \uc124\uc815\nlatent_size = 64\nbatch_size = 128\nnum_epochs = 100\nlearning_rate = 0.0002\n\n# \ubcc0\ud658 \uc124\uc815 \ubc0f \ub370\uc774\ud130 \ub85c\ub4dc\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\nmnist = datasets.MNIST(root='data\/', train=True, transform=transform, download=True)\ntrain_loader = torch.utils.data.DataLoader(dataset=mnist, batch_size=batch_size, shuffle=True)\n\n# \uc0dd\uc131\uc790 \ubaa8\ub378 \uc815\uc758\nclass Generator(nn.Module):\n def __init__(self):\n super(Generator, self).__init__()\n self.model = nn.Sequential(\n nn.Linear(latent_size, 256),\n nn.ReLU(True),\n nn.Linear(256, 512),\n nn.ReLU(True),\n nn.Linear(512, 1024),\n nn.ReLU(True),\n nn.Linear(1024, 28 * 28),\n nn.Tanh() # \ucd9c\ub825 \ubc94\uc704 [-1, 1]\n )\n\n def forward(self, z):\n return self.model(z).view(z.size(0), 1, 28, 28)\n\n# \ud310\ubcc4\uc790 \ubaa8\ub378 \uc815\uc758\nclass Discriminator(nn.Module):\n def __init__(self):\n super(Discriminator, self).__init__()\n self.model = nn.Sequential(\n nn.Linear(28 * 28, 1024),\n nn.LeakyReLU(0.2),\n nn.Linear(1024, 512),\n nn.LeakyReLU(0.2),\n nn.Linear(512, 256),\n nn.LeakyReLU(0.2),\n nn.Linear(256, 1),\n nn.Sigmoid() # \ucd9c\ub825 \ud655\ub960\n )\n\n def forward(self, img):\n return self.model(img.view(img.size(0), -1))\n\n# \uc0dd\uc131\uc790 \ubc0f \ud310\ubcc4\uc790 \ucd08\uae30\ud654\ngenerator = Generator()\ndiscriminator = Discriminator()\n\n# \uc190\uc2e4 \ud568\uc218 \ubc0f \ucd5c\uc801\ud654 \uae30\ubc95 \uc124\uc815\ncriterion = nn.BCELoss()\noptimizer_G = optim.Adam(generator.parameters(), lr=learning_rate)\noptimizer_D = optim.Adam(discriminator.parameters(), lr=learning_rate)\n\n# \ubaa8\ub378 \ud559\uc2b5\nfor epoch in range(num_epochs):\n for i, (imgs, _) in enumerate(train_loader):\n # \uc9c4\uc9dc \uc774\ubbf8\uc9c0\uc5d0 \ub300\ud55c \ub77c\ubca8\n real_labels = torch.ones(imgs.size(0), 1)\n # \uac00\uc9dc \uc774\ubbf8\uc9c0\uc5d0 \ub300\ud55c \ub77c\ubca8\n fake_labels = torch.zeros(imgs.size(0), 1)\n\n # \ud310\ubcc4\uc790 \ud559\uc2b5\n optimizer_D.zero_grad()\n outputs = discriminator(imgs)\n d_loss_real = criterion(outputs, real_labels)\n \n z = torch.randn(imgs.size(0), latent_size)\n fake_imgs = generator(z)\n outputs = discriminator(fake_imgs.detach())\n d_loss_fake = criterion(outputs, fake_labels)\n\n d_loss = d_loss_real + d_loss_fake\n d_loss.backward()\n optimizer_D.step()\n\n # \uc0dd\uc131\uc790 \ud559\uc2b5\n optimizer_G.zero_grad()\n outputs = discriminator(fake_imgs)\n g_loss = criterion(outputs, real_labels)\n\n g_loss.backward()\n optimizer_G.step()\n\n # \uc774\ubbf8\uc9c0 \uc800\uc7a5\n if (epoch+1) % 10 == 0:\n save_image(fake_imgs.data, f'images\/fake_images-{epoch+1}.png', nrow=8, normalize=True)\n print(f'Epoch [{epoch+1}\/{num_epochs}], d_loss: {d_loss.item():.4f}, g_loss: {g_loss.item():.4f}')\n <\/code>\n <\/pre>\nGAN\uc758 \ub3c4\uc804 \uacfc\uc81c<\/h2>\n
\n GAN\uc740 \uc5ec\ub7ec \uac00\uc9c0 \ub3c4\uc804 \uacfc\uc81c\uc5d0 \uc9c1\uba74\ud574 \uc788\uc2b5\ub2c8\ub2e4. \uc774 \uc139\uc158\uc5d0\uc11c\ub294 \uadf8 \uc911 \uba87 \uac00\uc9c0\ub97c \uc0b4\ud3b4\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.\n <\/p>\n
1. \ubaa8\ub4dc \ubd95\uad34(Mode Collapse)<\/h3>\n
\n \ubaa8\ub4dc \ubd95\uad34\ub294 \uc0dd\uc131\uc790\uac00 \uc81c\ud55c\ub41c \uc218\uc758 \ucd9c\ub825\ub9cc \uc0dd\uc131\ud558\ub3c4\ub85d \ud559\uc2b5\ud558\uc138\uc694 \ubc1c\uc0dd\ud558\ub294 \ud604\uc0c1\uc785\ub2c8\ub2e4.
\n \uc774\ub294 \uc0dd\uc131\uc790\uac00 \ud558\ub098\uc758 \uc774\ubbf8\uc9c0\ub97c \uc5ec\ub7ec \ubc88 \uc0dd\uc131\ud558\ub294 \uacb0\uacfc\ub97c \ub0b3\uc544 \ub2e4\ubcc0\ud654\ub41c \uacb0\uacfc\ub97c \uc81c\uacf5\ud560 \uc218 \uc5c6\uc2b5\ub2c8\ub2e4.
\n \uc774 \ubb38\uc81c\ub97c \ud574\uacb0\ud558\uae30 \uc704\ud574 \uc5ec\ub7ec \uac00\uc9c0 \uae30\ubc95\uc774 \uc81c\uc548\ub418\uc5c8\uc73c\uba70, \uadf8 \uc911 \ud558\ub098\ub294 \ub2e4\uc218\uc758 \ub2e4\uc591\ud55c \uac00\uc9dc \ub370\uc774\ud130\ub97c \uc0dd\uc131\ud560 \uc218 \uc788\ub3c4\ub85d
\n \ud558\ub294 \uac83\uc785\ub2c8\ub2e4.\n <\/p>\n
2. \ubd88\uc548\uc815\ud55c \ud6c8\ub828<\/h3>\n
\n GAN\uc758 \ud6c8\ub828\uc740 \uc885\uc885 \ubd88\uc548\uc815\ud558\uba70, \ud310\ubcc4\uc790\uc640 \uc0dd\uc131\uc790\uc758 \ud559\uc2b5 \uacfc\uc815\uc774 \ubd88\uade0\ud615\ud558\uba74 \ud6c8\ub828\uc774 \uc81c\ub300\ub85c \uc9c4\ud589\ub418\uc9c0 \uc54a\uc744 \uc218 \uc788\uc2b5\ub2c8\ub2e4.
\n \uc774\ub97c \ud574\uacb0\ud558\uae30 \uc704\ud574 \ub2e4\uc591\ud55c \ucd5c\uc801\ud654 \ubc29\ubc95\uacfc \ud6c8\ub828 \uc804\ub7b5\uc744 \uc0ac\uc6a9\ud558\ub294 \uac83\uc774 \ud544\uc694\ud569\ub2c8\ub2e4.\n <\/p>\n
3. \ubd80\uc815\ud655\ud55c \ud310\ubcc4<\/h3>\n
\n \ud310\ubcc4\uc790\uac00 \ub108\ubb34 \uac15\ub825\ud558\uba74 \uc0dd\uc131\uc790\uac00 \ud559\uc2b5\ud558\ub294 \ub370 \uc5b4\ub824\uc6c0\uc744 \uacaa\uc744 \uc218 \uc788\uc73c\uba70, \uc0dd\uc131\uc790\uac00 \ub108\ubb34 \uc57d\ud558\uba74 \ud310\ubcc4\uc790\uac00
\n \uc27d\uac8c \uc18d\uc774\ub294 \uacb0\uacfc\ub97c \ub0b3\uc744 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc801\uc808\ud55c \ud6c8\ub828 \uade0\ud615\uc744 \uc720\uc9c0\ud558\ub294 \uac83\uc774 \uc911\uc694\ud569\ub2c8\ub2e4.\n <\/p>\n
4. \uace0\ucc28\uc6d0 \uacf5\uac04\uc5d0\uc11c\uc758 \ubb38\uc81c<\/h3>\n
\n GAN\uc758 \ud6c8\ub828\uc740 \uace0\ucc28\uc6d0\uc758 \ub370\uc774\ud130\uc5d0\uc11c \uc9c4\ud589\ub418\uba70, \uc774\ub85c \uc778\ud574 \ud559\uc2b5\uc774 \uc5b4\ub824\uc6cc\uc9c0\ub294 \uacbd\uc6b0\uac00 \ub9ce\uc2b5\ub2c8\ub2e4.
\n \uace0\ucc28\uc6d0 \uacf5\uac04\uc5d0\uc11c\uc758 \ub370\uc774\ud130 \ud2b9\uc131\uc744 \uc798 \uc774\ud574\ud558\uace0 \uc801\uc808\ud55c \ubc29\ubc95\uc73c\ub85c \ubaa8\ub378\uc744 \uc124\uacc4\ud574\uc57c \ud569\ub2c8\ub2e4.\n <\/p>\n
\uacb0\ub860<\/h2>\n
\n GAN\uc740 \ub9e4\uc6b0 \uac15\ub825\ud55c \uc0dd\uc131 \ubaa8\ub378\uc774\uc9c0\ub9cc, \uc5ec\ub7ec \ub3c4\uc804 \uacfc\uc81c\ub97c \uac00\uc9c0\uace0 \uc788\uc2b5\ub2c8\ub2e4. \ud30c\uc774\ud1a0\uce58\ub97c \uc0ac\uc6a9\ud558\uba74 GAN\uc744 \uc27d\uac8c \uad6c\ud604\ud558\uace0
\n \uc2e4\ud5d8\ud560 \uc218 \uc788\uc73c\uba70, \uc774\ub97c \ud1b5\ud574 GAN\uc758 \uc774\ud574\ub3c4\ub97c \uc99d\uc9c4\uc2dc\ud0ac \uc218 \uc788\uc2b5\ub2c8\ub2e4. GAN\uc758 \ubc1c\uc804 \uac00\ub2a5\uc131\uc740 \ubb34\uad81\ubb34\uc9c4\ud558\uba70, \uc55e\uc73c\ub85c \ub354
\n \ub9ce\uc740 \uc5f0\uad6c\uc640 \uac1c\uc120\uc774 \uc774\ub8e8\uc5b4\uc9c8 \uac83\uc785\ub2c8\ub2e4.\n <\/p>\n<\/article>\n
<\/body><\/p>\n","protected":false},"excerpt":{"rendered":"