Statistics Using R最新文献

Statistics Using R Pub Date : 2020-08-13 DOI: 10.1017/9781108755351.020

{"title":"Dataset Descriptions","authors":"","doi":"10.1017/9781108755351.020","DOIUrl":"https://doi.org/10.1017/9781108755351.020","url":null,"abstract":"– MNIST and MNIST-test [31] which respectively correspond to full and test subset of MNIST dataset. They depict binary white handwritten digits centered over a black background. – USPS [17] is a handwritten digit dataset from USPS composed of greyscale images. – Fashion-MNIST [47] is a 10-class clothing dataset composed of greyscale images of cloth over black background. Classes are: T-shirt, trouser, pullover, dress, coat, sandal, shirt, sneaker, bag, ankle boot. – FRGC [43] is a colored face dataset. We use a subset of this dataset introduced in [49], where 20 subjects are selected and each image is cropped and resized to a constant size of 32×32. – SVHN [42] is composed of digits extracted from house numbers cropped from Google Street View images. Following standard practice for clustering, we use both labeled samples (99,289) and unlabeled extra samples (~530k) for training and evaluate on the labeled subset only. – affNIST-test is the test split of affNIST (https://www.cs.toronto.edu/ tijmen/affNIST/) an augmented dataset of MNIST where random affine transformations are applied. – MNIST-1k: we randomly sampled 1,000 images from the test split of MNIST. – MNIST-color: we augmented MNIST with random colors for background and foreground.","PeriodicalId":164454,"journal":{"name":"Statistics Using R","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-08-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116940306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}