\ub525\ub7ec\ub2dd\uc740 \uae30\uacc4 \ud559\uc2b5\uc758 \ud55c \ubd84\uc57c\ub85c, \uc778\uacf5 \uc2e0\uacbd\ub9dd\uc744 \ud1b5\ud574 \ub370\uc774\ud130\ub97c \ubaa8\ub378\ub9c1\ud558\uace0 \uc608\uce21\ud558\ub294 \ubc29\ubc95\ub860\uc785\ub2c8\ub2e4. \ub9ce\uc740 \ubd84\uc57c\uc5d0\uc11c \ud601\uc2e0\uc801\uc778 \ubc1c\uc804\uc744 \uc774\ub8e9\ud558\uac8c \ud55c \ub525\ub7ec\ub2dd\uc740 \ud2b9\ud788 \uc774\ubbf8\uc9c0 \uc778\uc2dd, \uc790\uc5f0\uc5b4 \ucc98\ub9ac, \ucd94\ucc9c \uc2dc\uc2a4\ud15c \ub4f1\uc5d0\uc11c \ub6f0\uc5b4\ub09c \uc131\ub2a5\uc744 \ubcf4\uc774\uace0 \uc788\uc2b5\ub2c8\ub2e4. \ubcf8 \uac15\uc88c\uc5d0\uc11c\ub294 \ud30c\uc774\ud1a0\uce58(PyTorch)\ub97c \uc774\uc6a9\ud558\uc5ec \ub525\ub7ec\ub2dd\uc758 \uac1c\ub150\uacfc \uc774\uc810\uc744 \uc790\uc138\ud788 \ub2e4\ub8e8\uc5b4 \ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
\ub525\ub7ec\ub2dd\uc740 \uc5ec\ub7ec \uce35(layer)\uc73c\ub85c \uad6c\uc131\ub41c \uc778\uacf5 \uc2e0\uacbd\ub9dd\uc744 \uc0ac\uc6a9\ud558\uc5ec \ub370\uc774\ud130 \ud2b9\uc131\uc744 \ud559\uc2b5\ud569\ub2c8\ub2e4. \uc774 \uacfc\uc815\uc5d0\uc11c \uc54c\uace0\ub9ac\uc998\uc740 \uc785\ub825 \ub370\uc774\ud130\uc640 \uc815\ub2f5(label) \uac04\uc758 \uad00\uacc4\ub97c \ud559\uc2b5\ud558\uac8c \ub429\ub2c8\ub2e4. \ub525\ub7ec\ub2dd\uc758 \uc8fc\uc694 \uad6c\uc131 \uc694\uc18c\ub294 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:<\/p>\n
\ud30c\uc774\ud1a0\uce58(PyTorch)\ub294 \ud398\uc774\uc2a4\ubd81\uc774 \uac1c\ubc1c\ud55c \uc720\uc5f0\ud558\uace0 \uac15\ub825\ud55c \ub525\ub7ec\ub2dd \ud504\ub808\uc784\uc6cc\ud06c\uc785\ub2c8\ub2e4. \ud30c\uc774\ud1a0\uce58\ub294 \ub3d9\uc801 \uacc4\uc0b0 \uadf8\ub798\ud504(dynamic computation graph)\ub97c \uc9c0\uc6d0\ud558\uba70, \uc774\ub85c \uc778\ud574 \ubaa8\ub378\uc744 \uc9c1\uad00\uc801\uc73c\ub85c \uad6c\uc131\ud558\uace0 \ub514\ubc84\uae45\ud560 \uc218 \uc788\ub294 \uc7a5\uc810\uc774 \uc788\uc2b5\ub2c8\ub2e4. \ub610\ud55c, \ub2e4\uc591\ud55c \uc2e0\uacbd\ub9dd \uad6c\uc131 \uc694\uc18c\ub97c \uc27d\uac8c \uc815\uc758\ud560 \uc218 \uc788\ub294 API\ub97c \uc81c\uacf5\ud558\uc5ec \uc5f0\uad6c\uc790\uc640 \uac1c\ubc1c\uc790 \ubaa8\ub450\uc5d0\uac8c \uc778\uae30\uac00 \ub192\uc2b5\ub2c8\ub2e4.<\/p>\n
\ub525\ub7ec\ub2dd\uc740 \uc804\ud1b5\uc801\uc778 \uae30\uacc4 \ud559\uc2b5 \uc54c\uace0\ub9ac\uc998\uc5d0 \ube44\ud574 \uc5ec\ub7ec \uac00\uc9c0 \uc7a5\uc810\uc744 \uc81c\uacf5\ud569\ub2c8\ub2e4. \uc8fc\uc694 \uc774\uc810\uc740 \ub2e4\uc74c\uacfc \uac19\uc2b5\ub2c8\ub2e4:<\/p>\n
\ub525\ub7ec\ub2dd\uc740 \ub2e4\uce35 \uc2e0\uacbd\ub9dd\uc744 \ud1b5\ud574 \ube44\uc120\ud615 \ub370\uc774\ud130\ub97c \ud6a8\uacfc\uc801\uc73c\ub85c \ucc98\ub9ac\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, \uc774\ubbf8\uc9c0 \uc778\uc2dd \ubb38\uc81c\uc5d0\uc11c \uc0ac\uc9c4\uc758 \ubc30\uacbd\uc774\ub098 \uc870\uba85\uc774 \ub2ec\ub77c\ub3c4 \ub525\ub7ec\ub2dd \ubaa8\ub378\uc740 \ud2b9\uc815 \uac1d\uccb4\ub97c \uc2dd\ubcc4\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\uc804\ud1b5\uc801\uc778 \ubc29\ubc95\uc5d0\uc11c\ub294 \uc804\ubb38\uac00\uac00 \uc9c1\uc811 \ud2b9\uc9d5(feature)\uc744 \ucd94\ucd9c\ud574\uc57c \ud588\uc73c\ub098, \ub525\ub7ec\ub2dd\uc740 \uc790\ub3d9\uc73c\ub85c \ud2b9\uc9d5\uc744 \ud559\uc2b5\ud558\uc5ec \uc131\ub2a5\uc744 \uac1c\uc120\ud569\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, \uc774\ubbf8\uc9c0 \ub370\uc774\ud130\ub97c \uc0ac\uc6a9\ud560 \uacbd\uc6b0 \uc800\ub834\ud55c \uc218\uc758 \ub808\uc774\uc5b4\ub85c \uace0\uae09 \ud2b9\uc9d5\uc744 \uc0dd\uc131\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\ub525\ub7ec\ub2dd\uc740 \ub300\ub7c9\uc758 \ub370\uc774\ud130\ub97c \ucc98\ub9ac\ud558\ub294 \ub370 \ub6f0\uc5b4\ub09c \uc131\ub2a5\uc744 \ubcf4\uc785\ub2c8\ub2e4. \uc2dc\uc2a4\ud15c\uc740 \ud559\uc2b5 \ub370\uc774\ud130\uc758 \uc591\uc774 \ub298\uc5b4\ub0a0\uc218\ub85d \uc77c\ubc18\ud654 \uc131\ub2a5\uc774 \uc810\uc810 \ud5a5\uc0c1\ub429\ub2c8\ub2e4. \uc774\ub294 \ucd94\ucc9c \uc2dc\uc2a4\ud15c\uc774\ub098 \uc790\uc5f0\uc5b4 \ucc98\ub9ac\uc640 \uac19\uc740 \ub300\uaddc\ubaa8 \uc751\uc6a9 \ud504\ub85c\uadf8\ub7a8\uc5d0\uc11c \ud2b9\ud788 \uc911\uc694\ud569\ub2c8\ub2e4.<\/p>\n
\ud30c\uc774\ud1a0\uce58\ub294 \uc0ac\uc6a9\uc790 \ub9de\ucda4\ud615 \uc544\ud0a4\ud14d\ucc98\ub97c \uc27d\uac8c \uc124\uacc4\ud560 \uc218 \uc788\uc5b4 \ub2e4\uc591\ud55c \ubb38\uc81c\ub97c \ub2e4\ub8f0 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc608\ub97c \ub4e4\uc5b4, \ud3ec\ud568\ud558\ub294 \uce35(Layer), \ub274\ub7f0\uc758 \uc218 \ub4f1 \uc0ac\uc6a9\uc790 \uc815\uc758\uac00 \uac00\ub2a5\ud558\uc5ec \ub2e4\uc591\ud55c \ubaa8\ub378\uc744 \uc2e4\ud5d8\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/p>\n
\ub2e4\uc74c\uc740 \ud30c\uc774\ud1a0\uce58\ub97c \uc0ac\uc6a9\ud558\uc5ec \uac04\ub2e8\ud55c \uc2e0\uacbd\ub9dd \ubaa8\ub378\uc744 \uad6c\ud604\ud558\ub294 \uc608\uc81c\uc785\ub2c8\ub2e4. \uc774 \uc608\uc81c\uc5d0\uc11c\ub294 MNIST \ub370\uc774\ud130\uc14b\uc744 \uc0ac\uc6a9\ud558\uc5ec \uc22b\uc790\ub97c \ubd84\ub958\ud558\ub294 \uc791\uc5c5\uc744 \uc218\ud589\ud569\ub2c8\ub2e4.<\/p>\n
!pip install torch torchvision<\/code><\/pre>\n4.2 MNIST \ub370\uc774\ud130 \ub2e4\uc6b4\ub85c\ub4dc \ubc0f \uc804\ucc98\ub9ac<\/h3>\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torchvision import datasets, transforms\n\n# \ub370\uc774\ud130 \uc804\ucc98\ub9ac\ntransform = transforms.Compose([\n transforms.ToTensor(),\n transforms.Normalize((0.5,), (0.5,))\n])\n\n# \ub370\uc774\ud130 \ub85c\ub529\ntrain_dataset = datasets.MNIST(root='.\/data', train=True, download=True, transform=transform)\ntest_dataset = datasets.MNIST(root='.\/data', train=False, download=True, transform=transform)\n\ntrain_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)\ntest_loader = torch.utils.data.DataLoader(dataset=test_dataset, batch_size=64, shuffle=False)\n<\/code><\/pre>\n4.3 \uc2e0\uacbd\ub9dd \ubaa8\ub378 \uc815\uc758<\/h3>\n# \uc2e0\uacbd\ub9dd \ud074\ub798\uc2a4 \uc815\uc758\nclass SimpleNN(nn.Module):\n def __init__(self):\n super(SimpleNN, self).__init__()\n self.fc1 = nn.Linear(28 * 28, 128) # \uc785\ub825\uce35\n self.fc2 = nn.Linear(128, 64) # \uc740\ub2c9\uce35\n self.fc3 = nn.Linear(64, 10) # \ucd9c\ub825\uce35\n\n def forward(self, x):\n x = x.view(-1, 28 * 28) # \uc774\ubbf8\uc9c0 \ud3c9\ud0c4\ud654\n x = torch.relu(self.fc1(x))\n x = torch.relu(self.fc2(x))\n x = self.fc3(x)\n return x\n\nmodel = SimpleNN()\n<\/code><\/pre>\n4.4 \uc190\uc2e4 \ud568\uc218 \ubc0f \uc635\ud2f0\ub9c8\uc774\uc800 \uc815\uc758<\/h3>\ncriterion = nn.CrossEntropyLoss() # \uc190\uc2e4 \ud568\uc218\noptimizer = optim.Adam(model.parameters(), lr=0.001) # \uc635\ud2f0\ub9c8\uc774\uc800\n<\/code><\/pre>\n4.5 \ubaa8\ub378 \ud559\uc2b5<\/h3>\nfor epoch in range(5): # 5\ubc88\uc758 epoch \ub3d9\uc548 \ud559\uc2b5\n for data, target in train_loader:\n optimizer.zero_grad() # \uae30\uc6b8\uae30 \ucd08\uae30\ud654\n output = model(data) # \uc608\uce21\n loss = criterion(output, target) # \uc190\uc2e4 \uacc4\uc0b0\n loss.backward() # \uae30\uc6b8\uae30 \uacc4\uc0b0\n optimizer.step() # \uac00\uc911\uce58 \uc5c5\ub370\uc774\ud2b8\n\n print(f'Epoch {epoch+1} completed.')\n<\/code><\/pre>\n4.6 \ubaa8\ub378 \ud3c9\uac00<\/h3>\ncorrect = 0\ntotal = 0\nwith torch.no_grad():\n for data, target in test_loader:\n output = model(data)\n _, predicted = torch.max(output.data, 1) # \ucd5c\ub313\uac12 \uc778\ub371\uc2a4\n total += target.size(0)\n correct += (predicted == target).sum().item()\n\nprint(f'Accuracy: {100 * correct \/ total}%')\n<\/code><\/pre>\n5. \uacb0\ub860<\/h2>\n
\ub525\ub7ec\ub2dd\uc740 \ub9e4\uc6b0 \uac15\ub825\ud55c \ub3c4\uad6c\ub85c, \ud30c\uc774\ud1a0\uce58\ub294 \uc774\ub97c \uc704\ud55c \ud6cc\ub96d\ud55c \ud504\ub808\uc784\uc6cc\ud06c\uc785\ub2c8\ub2e4. \ube44\uc120\ud615 \ub370\uc774\ud130 \ucc98\ub9ac, \uc790\ub3d9 \ud2b9\uc9d5 \ucd94\ucd9c, \ub300\uaddc\ubaa8 \ub370\uc774\ud130 \ucc98\ub9ac, \uc720\uc5f0\ud55c \uad6c\uc870 \uc124\uacc4\ub97c \ud1b5\ud574 \ub2e4\uc591\ud55c \ud604\uc0c1\uc774\ub098 \ubb38\uc81c\ub97c \ud574\uacb0\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774\ubc88 \uac15\uc88c\uc5d0\uc11c\ub294 \uac04\ub2e8\ud55c \uc608\uc81c\ub97c \ud1b5\ud574 \ud30c\uc774\ud1a0\uce58\uc758 \uae30\ubcf8\uc801\uc778 \uc0ac\uc6a9\ubc95\uacfc \ub525\ub7ec\ub2dd\uc758 \uc774\uc810\uc744 \uc124\uba85\ud588\uc2b5\ub2c8\ub2e4. \uc55e\uc73c\ub85c \ub354 \ubc1c\uc804\ub41c \ubaa8\ub378\uacfc \uae30\uc220\uc744 \ub2e4\ub8e8\ub294 \uace0\uae09 \uac15\uc88c\ub4e4\ub3c4 \uc900\ube44\ud560 \uc608\uc815\uc785\ub2c8\ub2e4. \ub9ce\uc740 \uad00\uc2ec \ubd80\ud0c1\ub4dc\ub9bd\ub2c8\ub2e4!<\/p>\n
<\/body><\/p>\n","protected":false},"excerpt":{"rendered":"