Datasets for feature screening strategy

There are five datasets for feature screening strategy: AllBooks : It is a text dataset containing 8ligious books (590 chapters). It aims to count the words from the bag of words (8265) preprocessing of the mini-corpus made up of the 8 religious books. NIPSpaper : It is a text dataset including 2483 NIPS conference papers. It aims to count each word (total 14035 words) in each paper. 20Newsgroup : This is a text dataset containing 204 news. It aims to count the word (total 18 845 words) in each new. Encyclopedia : Here is a tiny subset of word counts from some encyclopedia articles. Only the 29515 most common words are used in the vocabulary, and only about 7500 articles are represented. In other words, it includes 7500 samples (new) and 29515 features (word). TasteProfile : It is a song listening count dataset that contains real user-play counts from undisclosed partners, all songs matched to the MSD. In this experiment, we only keep the users who listened to at least 20 songs and the songs that are listened to by at least 50 users. Thus, the final dataset has 6000 songs and 70000 users.