6. \uacb0\ub860<\/a><\/li>\n<\/ul>\n1. \ub370\uc774\ud130 \uc804\ucc98\ub9ac\ub780?<\/h2>\n
\ub370\uc774\ud130 \uc804\ucc98\ub9ac\ub294 \uba38\uc2e0\ub7ec\ub2dd \ub610\ub294 \ub525\ub7ec\ub2dd \ubaa8\ub378\uc5d0 \uc785\ub825\ud558\uae30 \uc804\uc5d0 \ub370\uc774\ud130\ub97c \ubcc0\ud658\ud558\uace0 \uc815\ub9ac\ud558\ub294 \uacfc\uc815\uc744 \ub9d0\ud569\ub2c8\ub2e4. \uc774 \uacfc\uc815\uc5d0\uc11c \ub370\uc774\ud130\uc758 \uc77c\uad00\uc131, \ubb34\uacb0\uc131, \ud488\uc9c8\uc744 \ubcf4\uc7a5\ud558\uac8c \ub429\ub2c8\ub2e4. \uc804\ucc98\ub9ac \ub2e8\uacc4\uc5d0\uc11c\ub294 \uacb0\uce21\uac12 \ucc98\ub9ac, \uc774\uc0c1\uac12 \uc81c\uac70, \ubc94\uc8fc\ud615 \ubcc0\uc218 \uc778\ucf54\ub529, \ub370\uc774\ud130 \uc815\uaddc\ud654 \ubc0f \ud2b9\uc131 \uc120\ud0dd \ub4f1\uc758 \uc791\uc5c5\uc774 \ud3ec\ud568\ub429\ub2c8\ub2e4.<\/p>\n
1.1 \uacb0\uce21\uac12 \ucc98\ub9ac<\/h3>\n
\uacb0\uce21\uac12\uc740 \ub370\uc774\ud130 \ubd84\uc11d\uc5d0\uc11c \ud2b9\ud788 \ubb38\uc81c\ub97c \uc77c\uc73c\ud0ac \uc218 \uc788\uc2b5\ub2c8\ub2e4. \ud30c\uc774\uc36c\uc758 pandas<\/code> \ub77c\uc774\ube0c\ub7ec\ub9ac\ub97c \uc774\uc6a9\ud558\uc5ec \uacb0\uce21\uac12\uc744 \ud0d0\uc9c0\ud558\uace0 \ucc98\ub9ac\ud558\ub294 \ubc29\ubc95\uc744 \uc54c\uc544\ubcf4\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\nimport pandas as pd\n\n# \ub370\uc774\ud130 \ub85c\ub4dc\ndata = pd.read_csv('data.csv')\n\n# \uacb0\uce21\uac12 \ud655\uc778\nprint(data.isnull().sum())\n\n# \uacb0\uce21\uac12 \uc0ad\uc81c\ndata_cleaned = data.dropna()\n# \ub610\ub294 \uacb0\uce21\uac12\uc744 \ud3c9\uade0\uc73c\ub85c \ub300\uccb4\ndata.fillna(data.mean(), inplace=True)<\/code><\/pre>\n1.2 \uc774\uc0c1\uac12 \ud0d0\uc9c0 \ubc0f \uc81c\uac70<\/h3>\n
\uc774\uc0c1\uac12\uc740 \ubaa8\ub378\uc758 \ud559\uc2b5\uc5d0 \ubd80\uc815\uc801\uc778 \uc601\ud5a5\uc744 \ubbf8\uce60 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774\uc0c1\uac12\uc744 \ud0d0\uc9c0\ud558\uace0 \uc81c\uac70\ud558\ub294 \ubc29\ubc95\uc740 \uc5ec\ub7ec \uac00\uc9c0\uac00 \uc788\ub294\ub370, \uc5ec\uae30\uc11c\ub294 IQR \ubc29\ubc95\uc744 \uc0ac\uc6a9\ud55c \uc608\uc81c\ub97c \ubcf4\uc5ec\ub4dc\ub9ac\uaca0\uc2b5\ub2c8\ub2e4.<\/p>\n
Q1 = data['column_name'].quantile(0.25)\nQ3 = data['column_name'].quantile(0.75)\nIQR = Q3 - Q1\n\n# IQR\uc744 \uc774\uc6a9\ud55c \uc774\uc0c1\uac12 \ud0d0\uc9c0\noutliers = data[(data['column_name'] < (Q1 - 1.5 * IQR)) | (data['column_name'] > (Q3 + 1.5 * IQR))]\ndata_no_outliers = data[~data.index.isin(outliers.index)]<\/code><\/pre>\n2. \uc815\uaddc\ud654\ub780?<\/h2>\n
\uc815\uaddc\ud654\ub294 \uc11c\ub85c \ub2e4\ub978 \ubc94\uc704\ub97c \uac00\uc9c4 \ub370\uc774\ud130\uc758 \uac12\ub4e4\uc744 \uc77c\uc815\ud55c \ubc94\uc704\ub85c \ubcc0\ud658\ud558\ub294 \uacfc\uc815\uc785\ub2c8\ub2e4. \uc774\ub97c \ud1b5\ud574 \ubaa8\ub378\uc758 \uc218\ub834 \uc18d\ub3c4\ub97c \ud5a5\uc0c1\uc2dc\ud0a4\uace0, \ud2b9\uc815 \ud2b9\uc131\uc774 \ubaa8\ub378\uc5d0 \ubbf8\uce58\ub294 \uc601\ud5a5\uc744 \uc904\uc77c \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc8fc\ub85c Min-Max \uc815\uaddc\ud654\uc640 Z-score \uc815\uaddc\ud654\uac00 \uc0ac\uc6a9\ub429\ub2c8\ub2e4.<\/p>\n
2.1 Min-Max \uc815\uaddc\ud654<\/h3>\n
Min-Max \uc815\uaddc\ud654\ub294 \uac01 \ud2b9\uc131\uc758 \uac12\uc744 0\uacfc 1 \uc0ac\uc774\ub85c \ubcc0\ud658\ud569\ub2c8\ub2e4. \uc774 \ubc29\ubc95\uc740 \ub2e4\uc74c \uacf5\uc2dd\uc744 \ub530\ub985\ub2c8\ub2e4:<\/p>\n
X' = (X - X_min) \/ (X_max - X_min)<\/code><\/pre>\n2.2 Z-score \uc815\uaddc\ud654<\/h3>\n
Z-score \uc815\uaddc\ud654\ub294 \uac01 \ud2b9\uc131\uc758 \uac12\uc744 \ud3c9\uade0\uc774 0\uc774\uace0 \ud45c\uc900\ud3b8\ucc28\uac00 1\uc774 \ub418\ub3c4\ub85d \ubcc0\ud658\ud569\ub2c8\ub2e4. \uc774 \ubc29\ubc95\uc740 \ub2e4\uc74c \uacf5\uc2dd\uc744 \ub530\ub985\ub2c8\ub2e4:<\/p>\n
X' = (X - \u03bc) \/ \u03c3<\/code><\/pre>\n\uc5ec\uae30\uc11c \u03bc<\/code>\ub294 \ud3c9\uade0, \u03c3<\/code>\ub294 \ud45c\uc900 \ud3b8\ucc28\uc785\ub2c8\ub2e4.<\/p>\n3. \uc65c \uc804\ucc98\ub9ac\uc640 \uc815\uaddc\ud654\uac00 \ud544\uc694\ud55c\uac00?<\/h2>\n
\ub370\uc774\ud130 \uc804\ucc98\ub9ac\uc640 \uc815\uaddc\ud654 \uacfc\uc815\uc740 \ubaa8\ub378\uc758 \uc131\ub2a5\uc744 \uadf9\ub300\ud654\ud558\ub294 \ub370 \ud544\uc218\uc801\uc785\ub2c8\ub2e4. \uc65c\ub0d0\ud558\uba74:<\/p>\n
\n- \uacb0\uce21\uac12\uc774\ub098 \uc774\uc0c1\uac12\uc774 \uc874\uc7ac\ud560 \uacbd\uc6b0, \ubaa8\ub378\uc758 \uc77c\ubc18\ud654 \uc131\ub2a5\uc774 \ub5a8\uc5b4\uc9c8 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n
- \uc815\uaddc\ud654\uac00 \ub418\uc9c0 \uc54a\uc740 \ub370\uc774\ud130\ub294 \ud559\uc2b5 \uc18d\ub3c4\ub97c \uc800\ud558\uc2dc\ud0a4\uace0, \ucd5c\uc801\ud654 \uc54c\uace0\ub9ac\uc998\uc758 \uc218\ub834 \ubb38\uc81c\ub97c \ubc1c\uc0dd\uc2dc\ud0ac \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n
- \uc11c\ub85c \ub2e4\ub978 \ubc94\uc704\ub97c \uac00\uc9c4 \ud2b9\uc131\uc740 \ubaa8\ub378\uc774 \ud2b9\uc815 \ud2b9\uc131\uc744 \uacfc\ub300 \ub610\ub294 \uacfc\uc18c \ud3c9\uac00\ud558\ub3c4\ub85d \uc720\ub3c4\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4.<\/li>\n<\/ul>\n
4. \ud30c\uc774\ud1a0\uce58\uc5d0\uc11c\uc758 \ub370\uc774\ud130 \uc804\ucc98\ub9ac<\/h2>\n
\ud30c\uc774\ud1a0\uce58\uc5d0\uc11c\ub294 torchvision.transforms<\/code>\ub97c \uc0ac\uc6a9\ud558\uc5ec \uc774\ubbf8\uc9c0\ub97c \uc804\ucc98\ub9ac\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc77c\ubc18\uc801\uc73c\ub85c \ub370\uc774\ud130\uc14b\uc744 \ub85c\ub4dc\ud560 \ub54c \uc544\ub798\uc640 \uac19\uc740 \ubcc0\ud658\uc744 \uc801\uc6a9\ud569\ub2c8\ub2e4.<\/p>\nimport torchvision.transforms as transforms\nfrom torchvision import datasets\n\ntransform = transforms.Compose([\n transforms.Resize((256, 256)),\n transforms.ToTensor(),\n transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5]),\n])\n\n# \ub370\uc774\ud130\uc14b \ub85c\ub4dc\ntrain_dataset = datasets.CIFAR10(root='.\/data', train=True, download=True, transform=transform)\ntrain_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)<\/code><\/pre>\n5. \ud30c\uc774\ud1a0\uce58\uc5d0\uc11c\uc758 \uc815\uaddc\ud654<\/h2>\n
\ud30c\uc774\ud1a0\uce58\uc5d0\uc11c\ub294 \uc774\ubbf8\uc9c0 \uc815\uaddc\ud654\ub97c \uc27d\uac8c \uc218\ud589\ud560 \uc218 \uc788\ub3c4\ub85d \ubbf8\ub9ac \uc815\uc758\ub41c \uc815\uaddc\ud654 \uacc4\uce35\uc744 \uc81c\uacf5\ud569\ub2c8\ub2e4. \ub2e4\uc74c\uc740 \uc774\ubbf8\uc9c0 \ub370\uc774\ud130\ub97c \uc815\uaddc\ud654\ud558\ub294 \ubc29\ubc95\uc785\ub2c8\ub2e4.<\/p>\n
import torch\nimport torchvision.transforms as transforms\n\n# \uc815\uaddc\ud654 \ubcc0\ud658 \uc815\uc758\nnormalize = transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])\n\n# \uc784\uc758\uc758 \uc774\ubbf8\uc9c0 \ud150\uc11c\nimage = torch.randn(3, 256, 256) # (\ucc44\ub110 \uc218, \ub192\uc774, \ub108\ube44)\n\n# \uc815\uaddc\ud654 \uc801\uc6a9\nnormalized_image = normalize(image)<\/code><\/pre>\n6. \uacb0\ub860<\/h2>\n
\ub525\ub7ec\ub2dd \ubaa8\ub378\uc758 \uc131\ub2a5\uc740 \ub370\uc774\ud130 \ud488\uc9c8\uc5d0 \ud06c\uac8c \uc758\uc874\ud569\ub2c8\ub2e4. \uc804\ucc98\ub9ac\uc640 \uc815\uaddc\ud654\ub294 \ubaa8\ub378\uc774 \ud6a8\uacfc\uc801\uc73c\ub85c \ud559\uc2b5\ud560 \uc218 \uc788\ub3c4\ub85d \ub370\uc774\ud130\ub97c \uc900\ube44\ud558\ub294 \ub370 \ud544\uc218\uc801\uc778 \ub2e8\uacc4\uc785\ub2c8\ub2e4. \ud30c\uc774\ud1a0\uce58\ub97c \ud65c\uc6a9\ud558\uba74 \uc774\ub7ec\ud55c \uc804\ucc98\ub9ac \ubc0f \uc815\uaddc\ud654 \uc791\uc5c5\uc744 \ud6a8\uacfc\uc801\uc73c\ub85c \uc218\ud589\ud560 \uc218 \uc788\uc2b5\ub2c8\ub2e4. \uc774\ubc88 \uac15\uc88c\ub97c \ud1b5\ud574 \ub370\uc774\ud130 \uc804\ucc98\ub9ac\uc640 \uc815\uaddc\ud654\uc758 \ud544\uc694\uc131\uc744 \uc774\ud574\ud558\uace0, \uc2e4\uc81c \ucf54\ub4dc \uc608\uc81c\ub97c \ud1b5\ud574 \ud30c\uc774\ud1a0\uce58\uc5d0\uc11c \uc774\ub97c \uc5b4\ub5bb\uac8c \uad6c\ud604\ud560 \uc218 \uc788\ub294\uc9c0 \uc54c\uc544\ubcf4\uc558\uc2b5\ub2c8\ub2e4. \ud5a5\ud6c4 \ub525\ub7ec\ub2dd \ud504\ub85c\uc81d\ud2b8\uc5d0\uc11c\ub294 \ud56d\uc0c1 \ub370\uc774\ud130 \uc804\ucc98\ub9ac\uc640 \uc815\uaddc\ud654 \uacfc\uc815\uc744 \uc2e0\uacbd\uc368\uc57c \ud569\ub2c8\ub2e4.<\/p>\n