Question

In 2019, Chhavi Yadav and Léon Bottou recovered 50,000 lost examples of these things by recreating the imprecisely documented preprocessing steps used to generate them. In his USENIX (“USE-nix”) Security keynote, James Mickens joked that “three explorers…went into a haunted house” and found these things in an “ancient book of evil”. Researchers at clothing retailer Zalando argued that these things are overused, and released photos of various fashion products as a drop-in replacement. Some of these things sourced from high-school students and Census Bureau employees were collected by Corinna Cortes, Christopher Burges, and (*) Yann LeCun into a database sometimes called the “Hello World” of deep learning. (10[2]-5[1])70,000 (10[1])small grayscale images of these things make up the MNIST (“EM-nist”) dataset. For 10 points, a simple OCR task involves classifying pictures of what things into their ten (10[1])possible categories? ■END■

Back to tossups