Wide-minima Density Hypothesis and the Explore-Exploit Learning Rate Schedule

Open paper

Benchmark	Model	Rank	Results
machine-translation-on-iwslt2014-german	Cutoff+Knee	#7	BLEU score: 37.78
machine-translation-on-wmt2014-german-english	MAT+Knee	#6	BLEU score: 31.9