learning-on-hard-examples (WIP) Optimize learning algorithms with use of only examples with a large cost. Inspired by "Accelerating Deep Learning by Focusing on the Biggest Losers" paper Backboan model Described in Lambda Lab's blog.