I'm curious why this claim (distributed minibatch is hard) continues to be repeated. Am I misreading the state-of-the-art? There was a paper at Interspeech this

I'm curious why this claim (distributed minibatch is hard) continues to be repeated. Am I misreading the state-of-the-art? There was a paper at Interspeech this year [0] where Amazon showed excellent scaling (54X speed up when distributed across 80 nodes) without loss of accuracy. [0] http://www.nikkostrom.com/publications/interspeech2015/strom_interspeech2015.pdf http://www.nikkostrom.com/publications/interspeech2015/strom...