What does backward feature elimination do?

A. Starts with a broad feature set and removes features step by step B. Starts with no features and adds random variables C. Converts all features into one-hot indicators D. Deletes the training observations in reverse order