1 datasets-fetch

Racket

1 datasets-fetch🔗ℹ

13 datasets. Licenses: CC-BY-4.0; CC-BY-SA-4.0; CC0-1.0; MIT; NOASSERTION.

1.1 california-housing🔗ℹ

20640 rows × 9 columns, fetch tier. Downloads 0.44 MB (cal_housing.tgz) and converts it to scikit-learn’s fetch_california_housing frame: total rooms, bedrooms and population divided by households, and the median house value in units of 100,000 dollars. License NOASSERTION: neither the authors nor StatLib state terms; the file is fetched for use from the mirror scikit-learn uses and never redistributed.

Pace, R. K. and Barry, R. (1997). Sparse spatial autoregressions. Statistics and Probability Letters, 33(3), 291-297. Data from the 1990 US census, one row per block group, as collected by L. Torgo.

Torgo’s cal_housing.tgz, the file scikit-learn 1.8.0 pins (hash pinned); NOASSERTION

Fetch tier: converted on first use from 1 upstream file by california-housing-rows in datasets/fetch-private/convert: cal_housing.tgz (SHA-256 aaa5c9a6afe2225cc2aed2723682ae403280c4a3695a2ddda4ffb5d8215ea681); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
med-inc	MedInc	real	Median household income in the block group; tens of thousands of US dollars; ()
house-age	HouseAge	real	Median house age in the block group; years; ()
ave-rooms	AveRooms	real	Average number of rooms per household; rooms per household; ()
ave-bedrms	AveBedrms	real	Average number of bedrooms per household; bedrooms per household; ()
population	Population	real	Block group population; persons; ()
ave-occup	AveOccup	real	Average number of household members; persons per household; ()
latitude	Latitude	real	Block group latitude; degrees; ()
longitude	Longitude	real	Block group longitude; degrees; ()
med-house-val	MedHouseVal	real	Median house value in the block group; hundreds of thousands of US dollars; ()

1.2 cifar-10🔗ℹ

60000 rows × 3074 columns, fetch tier. Downloads 170.1 MB (cifar-10-binary.tar.gz) and converts it to one row per 32 x 32 colour image: the 50,000 training images of data_batch_1 to data_batch_5, then the 10,000 test images of test_batch, in file order, as torchvision’s CIFAR10 with train=True then train=False. Pixels keep the file’s plane order, red1 to red1024, then green, then blue, each plane row-major, so a row’s pixels reshape to image-shape (3 32 32). License NOASSERTION: the authors state no terms and ask only that the technical report be cited; the archive is fetched for use from their page and never redistributed.

Krizhevsky, A. (2009). Learning multiple layers of features from tiny images. Technical report, University of Toronto.

cifar-10-binary.tar.gz, the binary version on the authors’ page (hash pinned); NOASSERTION

Fetch tier: converted on first use from 1 upstream file by cifar-10-rows in datasets/fetch-private/convert: cifar-10-binary.tar.gz (SHA-256 c4a38c50a1bc5f3a1c5537f2155ab9d68f9f25eb1ed8d9ddda3db29a59bca1dd); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
red1 … red1024	red1 … red1024	integer	Red level, row-major over the 32 x 32 image, 0 to 255; #f; ()
green1 … green1024	green1 … green1024	integer	Green level, row-major over the 32 x 32 image, 0 to 255; #f; ()
blue1 … blue1024	blue1 … blue1024	integer	Blue level, row-major over the 32 x 32 image, 0 to 255; #f; ()
label	label	integer	Class: 0 airplane, 1 automobile, 2 bird, 3 cat, 4 deer, 5 dog, 6 frog, 7 horse, 8 ship, 9 truck; #f; ()
split	split	categorical	train: the 50,000 training images (data_batch_1 to data_batch_5); test: the 10,000 test images (test_batch); #f; (train test)

1.3 cifar-100🔗ℹ

60000 rows × 3075 columns, fetch tier. Downloads 168.5 MB (cifar-100-binary.tar.gz) and converts it to one row per 32 x 32 colour image: the 50,000 training images of train.bin, then the 10,000 test images of test.bin, in file order, as torchvision’s CIFAR100 with train=True then train=False. label is the fine class (torchvision’s target) and coarse-label its superclass. Pixels keep the file’s plane order, red1 to red1024, then green, then blue, each plane row-major, so a row’s pixels reshape to image-shape (3 32 32). License NOASSERTION: the authors state no terms and ask only that the technical report be cited; the archive is fetched for use from their page and never redistributed.

Krizhevsky, A. (2009). Learning multiple layers of features from tiny images. Technical report, University of Toronto.

cifar-100-binary.tar.gz, the binary version on the authors’ page (hash pinned); NOASSERTION

Fetch tier: converted on first use from 1 upstream file by cifar-100-rows in datasets/fetch-private/convert: cifar-100-binary.tar.gz (SHA-256 58a81ae192c23a4be8b1804d68e518ed807d710a4eb253b1f2a199162a40d8ec); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
red1 … red1024	red1 … red1024	integer	Red level, row-major over the 32 x 32 image, 0 to 255; #f; ()
green1 … green1024	green1 … green1024	integer	Green level, row-major over the 32 x 32 image, 0 to 255; #f; ()
blue1 … blue1024	blue1 … blue1024	integer	Blue level, row-major over the 32 x 32 image, 0 to 255; #f; ()
label	label	integer	Fine class: 0 apple, 1 aquarium_fish, 2 baby, 3 bear, 4 beaver, 5 bed, 6 bee, 7 beetle, 8 bicycle, 9 bottle, 10 bowl, 11 boy, 12 bridge, 13 bus, 14 butterfly, 15 camel, 16 can, 17 castle, 18 caterpillar, 19 cattle, 20 chair, 21 chimpanzee, 22 clock, 23 cloud, 24 cockroach, 25 couch, 26 crab, 27 crocodile, 28 cup, 29 dinosaur, 30 dolphin, 31 elephant, 32 flatfish, 33 forest, 34 fox, 35 girl, 36 hamster, 37 house, 38 kangaroo, 39 keyboard, 40 lamp, 41 lawn_mower, 42 leopard, 43 lion, 44 lizard, 45 lobster, 46 man, 47 maple_tree, 48 motorcycle, 49 mountain, 50 mouse, 51 mushroom, 52 oak_tree, 53 orange, 54 orchid, 55 otter, 56 palm_tree, 57 pear, 58 pickup_truck, 59 pine_tree, 60 plain, 61 plate, 62 poppy, 63 porcupine, 64 possum, 65 rabbit, 66 raccoon, 67 ray, 68 road, 69 rocket, 70 rose, 71 sea, 72 seal, 73 shark, 74 shrew, 75 skunk, 76 skyscraper, 77 snail, 78 snake, 79 spider, 80 squirrel, 81 streetcar, 82 sunflower, 83 sweet_pepper, 84 table, 85 tank, 86 telephone, 87 television, 88 tiger, 89 tractor, 90 train, 91 trout, 92 tulip, 93 turtle, 94 wardrobe, 95 whale, 96 willow_tree, 97 wolf, 98 woman, 99 worm; #f; ()
coarse-label	coarse-label	integer	Coarse class (superclass): 0 aquatic_mammals, 1 fish, 2 flowers, 3 food_containers, 4 fruit_and_vegetables, 5 household_electrical_devices, 6 household_furniture, 7 insects, 8 large_carnivores, 9 large_man-made_outdoor_things, 10 large_natural_outdoor_scenes, 11 large_omnivores_and_herbivores, 12 medium_mammals, 13 non-insect_invertebrates, 14 people, 15 reptiles, 16 small_mammals, 17 trees, 18 vehicles_1, 19 vehicles_2; #f; ()
split	split	categorical	train: the 50,000 training images (train.bin); test: the 10,000 test images (test.bin); #f; (train test)

1.4 covtype🔗ℹ

581012 rows × 55 columns, fetch tier. Downloads 11.2 MB (covtype.data.gz) and converts it to 30 m cells of Roosevelt National Forest, Colorado, in file order, as scikit-learn’s fetch_covtype with shuffle=False. Column names follow scikit-learn’s, with 0-based wilderness-area-0 to -3 and soil-type-0 to -39.

Blackard, J. (1998). Covertype. UCI Machine Learning Repository. doi:10.24432/C50K5N. https://archive.ics.uci.edu/dataset/31/covertype Described in: Blackard, J. A. and Dean, D. J. (1999). Comparative accuracies of artificial neural networks and discriminant analysis in predicting forest cover types from cartographic variables. Computers and Electronics in Agriculture, 24(3), 131-151.

UCI covtype.data.gz, the file scikit-learn 1.8.0 pins (hash pinned); CC-BY-4.0

Fetch tier: converted on first use from 1 upstream file by covtype-rows in datasets/fetch-private/convert: covtype.data.gz (SHA-256 614360d0257557dd1792834a85a1cdebfadc3c4f30b011d56afee7ffb5b15771); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
elevation	Elevation	integer	Elevation; m; ()
aspect	Aspect	integer	Aspect; degrees azimuth; ()
slope	Slope	integer	Slope; degrees; ()
horizontal-distance-to-hydrology	Horizontal_Distance_To_Hydrology	integer	Horizontal distance to the nearest surface water; m; ()
vertical-distance-to-hydrology	Vertical_Distance_To_Hydrology	integer	Vertical distance to the nearest surface water; m; ()
horizontal-distance-to-roadways	Horizontal_Distance_To_Roadways	integer	Horizontal distance to the nearest roadway; m; ()
hillshade-9am	Hillshade_9am	integer	Hillshade at 9am, summer solstice; index 0 to 255; ()
hillshade-noon	Hillshade_Noon	integer	Hillshade at noon, summer solstice; index 0 to 255; ()
hillshade-3pm	Hillshade_3pm	integer	Hillshade at 3pm, summer solstice; index 0 to 255; ()
horizontal-distance-to-fire-points	Horizontal_Distance_To_Fire_Points	integer	Horizontal distance to the nearest wildfire ignition point; m; ()
wilderness-area-0	Wilderness_Area_0	integer	1 in the Rawah Wilderness Area, else 0; #f; ()
wilderness-area-1	Wilderness_Area_1	integer	1 in the Neota Wilderness Area, else 0; #f; ()
wilderness-area-2	Wilderness_Area_2	integer	1 in the Comanche Peak Wilderness Area, else 0; #f; ()
wilderness-area-3	Wilderness_Area_3	integer	1 in the Cache la Poudre Wilderness Area, else 0; #f; ()
soil-type-0 … soil-type-39	Soil_Type_0 … Soil_Type_39	integer	1 for this soil type, else 0; covtype.info lists the 40 types by ELU code; #f; ()
cover-type	Cover_Type	integer	Forest cover type: 1 Spruce/Fir, 2 Lodgepole Pine, 3 Ponderosa Pine, 4 Cottonwood/Willow, 5 Aspen, 6 Douglas-fir, 7 Krummholz; #f; ()

1.5 emnist-balanced🔗ℹ

131600 rows × 786 columns, fetch tier. Downloads 561.8 MB (gzip.zip, which holds all six EMNIST splits; emnist-balanced, emnist-letters and emnist-digits each download it) and converts the balanced split to one row per 28 x 28 grayscale image: the 112,800 training images, then the 18,800 test images, in file order, as torchvision’s EMNIST with split="balanced" and train=True then train=False. Pixels are pixel1 to pixel784 in row-major order as stored, untransposed as torchvision and TensorFlow Datasets read them: each image is the transpose of the upright character. License NOASSERTION: NIST states no terms and asks that Cohen et al. (2017) be cited; the archive is fetched for use and never redistributed.

Cohen, G., Afshar, S., Tapson, J. and van Schaik, A. (2017). EMNIST: an extension of MNIST to handwritten letters. arXiv:1702.05373.

gzip.zip, NIST’s archive of the six EMNIST splits as gzipped IDX files, the file torchvision downloads (hash pinned); NOASSERTION

Fetch tier: converted on first use from 1 upstream file by emnist-balanced-rows in datasets/fetch-private/convert: gzip.zip (SHA-256 fb9bb67e33772a9cc0b895e4ecf36d2cf35be8b709693c3564cea2a019fcda8e); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
pixel1 … pixel784	pixel1 … pixel784	integer	Gray level, row-major over the 28 x 28 image, 0 background to 255 foreground, as stored: the transpose of the upright character; #f; ()
label	label	integer	0-9 digits, 10-35 A-Z, 36-46 a b d e f g h n q r t; #f; ()
split	split	categorical	train: the 112,800 training images (emnist-balanced-train-* members); test: the 18,800 test images (emnist-balanced-test-* members); #f; (train test)

1.6 emnist-digits🔗ℹ

280000 rows × 786 columns, fetch tier. Downloads 561.8 MB (gzip.zip, which holds all six EMNIST splits; emnist-balanced, emnist-letters and emnist-digits each download it) and converts the digits split to one row per 28 x 28 grayscale image: the 240,000 training images, then the 40,000 test images, in file order, as torchvision’s EMNIST with split="digits" and train=True then train=False. Pixels are pixel1 to pixel784 in row-major order as stored, untransposed as torchvision and TensorFlow Datasets read them: each image is the transpose of the upright character. License NOASSERTION: NIST states no terms and asks that Cohen et al. (2017) be cited; the archive is fetched for use and never redistributed.

Cohen, G., Afshar, S., Tapson, J. and van Schaik, A. (2017). EMNIST: an extension of MNIST to handwritten letters. arXiv:1702.05373.

gzip.zip, NIST’s archive of the six EMNIST splits as gzipped IDX files, the file torchvision downloads (hash pinned); NOASSERTION

Fetch tier: converted on first use from 1 upstream file by emnist-digits-rows in datasets/fetch-private/convert: gzip.zip (SHA-256 fb9bb67e33772a9cc0b895e4ecf36d2cf35be8b709693c3564cea2a019fcda8e); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
pixel1 … pixel784	pixel1 … pixel784	integer	Gray level, row-major over the 28 x 28 image, 0 background to 255 foreground, as stored: the transpose of the upright character; #f; ()
label	label	integer	Digit 0 to 9; #f; ()
split	split	categorical	train: the 240,000 training images (emnist-digits-train-* members); test: the 40,000 test images (emnist-digits-test-* members); #f; (train test)

1.7 emnist-letters🔗ℹ

145600 rows × 786 columns, fetch tier. Downloads 561.8 MB (gzip.zip, which holds all six EMNIST splits; emnist-balanced, emnist-letters and emnist-digits each download it) and converts the letters split to one row per 28 x 28 grayscale image: the 124,800 training images, then the 20,800 test images, in file order, as torchvision’s EMNIST with split="letters" and train=True then train=False. Pixels are pixel1 to pixel784 in row-major order as stored, untransposed as torchvision and TensorFlow Datasets read them: each image is the transpose of the upright character. License NOASSERTION: NIST states no terms and asks that Cohen et al. (2017) be cited; the archive is fetched for use and never redistributed.

Cohen, G., Afshar, S., Tapson, J. and van Schaik, A. (2017). EMNIST: an extension of MNIST to handwritten letters. arXiv:1702.05373.

gzip.zip, NIST’s archive of the six EMNIST splits as gzipped IDX files, the file torchvision downloads (hash pinned); NOASSERTION

Fetch tier: converted on first use from 1 upstream file by emnist-letters-rows in datasets/fetch-private/convert: gzip.zip (SHA-256 fb9bb67e33772a9cc0b895e4ecf36d2cf35be8b709693c3564cea2a019fcda8e); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
pixel1 … pixel784	pixel1 … pixel784	integer	Gray level, row-major over the 28 x 28 image, 0 background to 255 foreground, as stored: the transpose of the upright character; #f; ()
label	label	integer	1 to 26: a to z, upper and lower case merged; #f; ()
split	split	categorical	train: the 124,800 training images (emnist-letters-train-* members); test: the 20,800 test images (emnist-letters-test-* members); #f; (train test)

1.8 fashion-mnist🔗ℹ

70000 rows × 786 columns, fetch tier. Downloads 30.9 MB (four IDX files) and converts them to one row per 28 x 28 grayscale image of a Zalando article: the 60,000 training images, then the 10,000 test images, in file order, as torchvision’s FashionMNIST with train=True then train=False. Pixels are pixel1 to pixel784 in row-major order.

Xiao, H., Rasul, K. and Vollgraf, R. (2017). Fashion-MNIST: a novel image dataset for benchmarking machine learning algorithms. arXiv:1708.07747.

The four IDX files torchvision downloads, identical to data/fashion at commit 658ec8c of the repository (hash pinned); MIT

Fetch tier: converted on first use from 4 upstream files by idx-image-rows in datasets/fetch-private/convert: train-images-idx3-ubyte.gz (SHA-256 3aede38d61863908ad78613f6a32ed271626dd12800ba2636569512369268a84), train-labels-idx1-ubyte.gz (SHA-256 a04f17134ac03560a47e3764e11b92fc97de4d1bfaf8ba1a3aa29af54cc90845), t10k-images-idx3-ubyte.gz (SHA-256 346e55b948d973a97e58d2351dde16a484bd415d4595297633bb08f03db6a073), t10k-labels-idx1-ubyte.gz (SHA-256 67da17c76eaffca5446c3361aaab5c3cd6d1c2608764d35dfb1850b086bf8dd5); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
pixel1 … pixel784	pixel1 … pixel784	integer	Gray level, row-major over the 28 x 28 image, 0 background to 255 foreground; #f; ()
label	label	integer	Article class: 0 T-shirt/top, 1 Trouser, 2 Pullover, 3 Dress, 4 Coat, 5 Sandal, 6 Shirt, 7 Sneaker, 8 Bag, 9 Ankle boot; #f; ()
split	split	categorical	train: the 60,000 training images (train-* files); test: the 10,000 test images (t10k-* files); #f; (train test)

1.9 kmnist🔗ℹ

70000 rows × 786 columns, fetch tier. Downloads 21.2 MB (four IDX files) and converts them to one row per 28 x 28 grayscale image of a cursive (Kuzushiji) hiragana character: the 60,000 training images, then the 10,000 test images, in file order, as torchvision’s KMNIST with train=True then train=False. Pixels are pixel1 to pixel784 in row-major order. License CC-BY-SA-4.0, as the rois-codh/kmnist repository states; the files are fetched for use and never redistributed.

Clanuwat, T., Bober-Irizar, M., Kitamoto, A., Lamb, A., Yamamoto, K. and Ha, D. (2018). Deep learning for classical Japanese literature. arXiv:1812.01718. KMNIST is derived from the Kuzushiji dataset created by the National Institute of Japanese Literature and edited by ROIS-DS Center for Open Data in the Humanities.

The four IDX files torchvision and TensorFlow Datasets download from CODH (hash pinned); CC-BY-SA-4.0

Fetch tier: converted on first use from 4 upstream files by idx-image-rows in datasets/fetch-private/convert: train-images-idx3-ubyte.gz (SHA-256 51467d22d8cc72929e2a028a0428f2086b092bb31cfb79c69cc0a90ce135fde4), train-labels-idx1-ubyte.gz (SHA-256 e38f9ebcd0f3ebcdec7fc8eabdcdaef93bb0df8ea12bee65224341c8183d8e17), t10k-images-idx3-ubyte.gz (SHA-256 edd7a857845ad6bb1d0ba43fe7e794d164fe2dce499a1694695a792adfac43c5), t10k-labels-idx1-ubyte.gz (SHA-256 20bb9a0ef54c7db3efc55a92eef5582c109615df22683c380526788f98e42a1c); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
pixel1 … pixel784	pixel1 … pixel784	integer	Gray level, row-major over the 28 x 28 image, 0 background to 255 foreground; #f; ()
label	label	integer	Hiragana class: 0 o, 1 ki, 2 su, 3 tsu, 4 na, 5 ha, 6 ma, 7 ya, 8 re, 9 wo; #f; ()
split	split	categorical	train: the 60,000 training images (train-* files); test: the 10,000 test images (t10k-* files); #f; (train test)

1.10 linnerud🔗ℹ

20 rows × 6 columns, fetch tier. Downloads 0.0004 MB (two CSV files) and joins them column by column into scikit-learn’s load_linnerud(as_frame=True).frame: the three exercise variables, then the three physiological targets, for twenty middle-aged men in a fitness club, in file order. License NOASSERTION: scikit-learn bundles the files with no stated data terms and cites only Tenenhaus (1998); its BSD-3-Clause licence covers its code. The files are fetched for use from scikit-learn’s repository at a pinned commit and never redistributed.

Tenenhaus, M. (1998). La regression PLS: theorie et pratique. Paris: Editions Technic.

linnerud_exercise.csv and linnerud_physiological.csv of scikit-learn 1.8.0, commit 646da0f (hash pinned); NOASSERTION

Fetch tier: converted on first use from 2 upstream files by linnerud-rows in datasets/fetch-private/convert: linnerud_exercise.csv (SHA-256 cb8d8c24937643fa2459682efb86c5e667bcd6dd93109eef81964d9e9f11bf8c), linnerud_physiological.csv (SHA-256 2bf7e05c1cd7d0adf0eca1e456941f624bed0a4fc96694d60d0ff7853ec5fcf7); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
chins	Chins	integer	Chin-ups, an exercise variable; #f; ()
situps	Situps	integer	Sit-ups, an exercise variable; #f; ()
jumps	Jumps	integer	Jumping jacks, an exercise variable; #f; ()
weight	Weight	integer	Body weight, a physiological target; #f; ()
waist	Waist	integer	Waist circumference, a physiological target; #f; ()
pulse	Pulse	integer	Pulse, a physiological target; #f; ()

1.11 mnist🔗ℹ

70000 rows × 786 columns, fetch tier. Downloads 11.6 MB (four IDX files) and converts them to one row per 28 x 28 grayscale image of a handwritten digit: the 60,000 training images, then the 10,000 test images, in file order, as torchvision’s MNIST with train=True then train=False. Pixels are pixel1 to pixel784 in row-major order. License NOASSERTION: the authors state no terms; the files are fetched for use from the mirrors torchvision and TensorFlow Datasets use and never redistributed. yann.lecun.com no longer serves them.

LeCun, Y., Bottou, L., Bengio, Y. and Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.

The four IDX files torchvision downloads from its ossci-datasets mirror (hash pinned); NOASSERTION

Fetch tier: converted on first use from 4 upstream files by idx-image-rows in datasets/fetch-private/convert: train-images-idx3-ubyte.gz (SHA-256 440fcabf73cc546fa21475e81ea370265605f56be210a4024d2ca8f203523609), train-labels-idx1-ubyte.gz (SHA-256 3552534a0a558bbed6aed32b30c495cca23d567ec52cac8be1a0730e8010255c), t10k-images-idx3-ubyte.gz (SHA-256 8d422c7b0a1c1c79245a5bcf07fe86e33eeafee792b84584aec276f5a2dbc4e6), t10k-labels-idx1-ubyte.gz (SHA-256 f7ae60f92e00ec6debd23a6088c31dbd2371eca3ffa0defaefb259924204aec6); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
pixel1 … pixel784	pixel1 … pixel784	integer	Gray level, row-major over the 28 x 28 image, 0 background to 255 foreground; #f; ()
label	label	integer	Digit 0 to 9; #f; ()
split	split	categorical	train: the 60,000 training images (train-* files); test: the 10,000 test images (t10k-* files); #f; (train test)

1.12 nycflights13-flights🔗ℹ

336776 rows × 19 columns, fetch tier. Downloads 33.2 MB (flights.csv, Rdatasets’ CSV of nycflights13 1.0.2) and converts it to the 336,776 flights that departed EWR, JFK and LGA in 2013, in R’s row order. CRAN’s .rda is bzip2-compressed, which base Racket cannot read, so Rdatasets is the source (code MIT, data unchanged). Its rownames column is dropped and empty fields are missing. Columns keep their R types: integers stay integers, doubles are reals; names normalized to kebab-case. The R POSIXct time_hour (America/New_York) is an ISO 8601 UTC string. The other four tables are bundled in datasets-r: nycflights13-airlines, nycflights13-airports, nycflights13-planes and nycflights13-weather.

Wickham, H. (2021). nycflights13: Flights that Departed NYC in 2013. R package version 1.0.2. https://CRAN.R-project.org/package=nycflights13. Data: Bureau of Transportation Statistics, https://www.transtats.bts.gov/DL_SelectFields.asp?Table_ID=236

nycflights13 1.0.2 via Rdatasets commit cc03c29690889dbe83089f5206e2422db8c3f71f (hash pinned); CC0-1.0

Fetch tier: converted on first use from 1 upstream file by csv-header-rows in datasets/fetch-private/convert: flights.csv (SHA-256 b791a2590c9167aa0bd3cb45af3f64101180878cf088c183eae4011e3a654f42); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
year	year	integer	Year of departure; #f; ()
month	month	integer	Month of departure; #f; ()
day	day	integer	Day of departure; #f; ()
dep-time	dep_time	integer	Actual departure time as HHMM or HMM, local time; #f; ()
sched-dep-time	sched_dep_time	integer	Scheduled departure time as HHMM or HMM, local time; #f; ()
dep-delay	dep_delay	real	Departure delay; negative for early departures; minutes; ()
arr-time	arr_time	integer	Actual arrival time as HHMM or HMM, local time; #f; ()
sched-arr-time	sched_arr_time	integer	Scheduled arrival time as HHMM or HMM, local time; #f; ()
arr-delay	arr_delay	real	Arrival delay; negative for early arrivals; minutes; ()
carrier	carrier	string	Two-letter carrier abbreviation; see nycflights13-airlines; #f; ()
flight	flight	integer	Flight number; #f; ()
tailnum	tailnum	string	Plane tail number; see nycflights13-planes; #f; ()
origin	origin	string	Origin airport: EWR, JFK or LGA; see nycflights13-airports; #f; ()
dest	dest	string	Destination airport; see nycflights13-airports; #f; ()
air-time	air_time	real	Time spent in the air; minutes; ()
distance	distance	real	Distance between the airports; miles; ()
hour	hour	real	Hour of the scheduled departure; #f; ()
minute	minute	real	Minute of the scheduled departure; #f; ()
time-hour	time_hour	string	Scheduled date and hour of departure, ISO 8601 UTC; #f; ()

1.13 zip🔗ℹ

9298 rows × 258 columns, fetch tier. Downloads 2.27 MB (two gzipped text files) and converts them to one row per 16 x 16 USPS digit: the 7,291 training rows, then the 2,007 test rows, in file order. Pixels are pixel1 to pixel256 in row-major order, read as R’s read.table reads them, so the 170 -0.0000 values of zip.train.gz stay negative zeros. License NOASSERTION: the book’s data page states no terms; the files are fetched for use and never redistributed. The same values are in CRAN ElemStatLearn 2015.6.26.2 (GPL (>= 2)); its zip.test has 68 more negative zeros.

Le Cun, Y. et al. (1990). Handwritten digit recognition with a back-propagation network. Advances in Neural Information Processing Systems 2. Data as distributed with Hastie, T., Tibshirani, R. and Friedman, J. (2009), The Elements of Statistical Learning, Second Edition, Springer.

zip.train.gz and zip.test.gz from the ESL data page (hash pinned); NOASSERTION

Fetch tier: converted on first use from 2 upstream files by esl-zip-rows in datasets/fetch-private/convert: zip.train.gz (SHA-256 66ebf231ab51f2cf6ac1a892d0f98ccffee1079d7a5db638c1f11f366046ca16), zip.test.gz (SHA-256 8dacc3c6e136af6c3488de9e5113a7daf140c8c2f1f0b61c54ee02f03f1a962f); see Downloaded datasets.

Column	Original	Type	Description / unit / levels
digit	digit	integer	Digit id, 0-9; #f; ()
pixel1 … pixel256	pixel1 … pixel256	real	Gray level from -1 (background) to 1, row-major over the 16 x 16 image; #f; ()
split	split	categorical	train: the 7,291 zip.train rows; test: the 2,007 zip.test rows; #f; (train test)

1.1	california-housing
1.2	cifar-10
1.3	cifar-100
1.4	covtype
1.5	emnist-balanced
1.6	emnist-digits
1.7	emnist-letters
1.8	fashion-mnist
1.9	kmnist
1.10	linnerud
1.11	mnist
1.12	nycflights13-flights
1.13	zip