Accelerating Deep Learning by Focusing on the Biggest Losers - 42Papers