François Chollet on Twitter: "This new paper could potentially upend batch normalization / renorm. Let's wait and see. Self-Normalizing NNs: https://t.co/AjG6njMAhJ"

Country	Code	For customers of
United States	40404	(any)
Canada	21212	(any)
United Kingdom	86444	Vodafone, Orange, 3, O2
Brazil	40404	Nextel, TIM
Haiti	40404	Digicel, Voila
Ireland	51210	Vodafone, O2
India	53000	Bharti Airtel, Videocon, Reliance
Indonesia	89887	AXIS, 3, Telkomsel, Indosat, XL Axiata
Italy	4880804	Wind
3424486444	Vodafone
» See SMS short codes for other countries

1. ❂ noah rosenberg ❂‏ @nrose 9 Jun 2017
  Replying to @fchollet
  
  This is crazy if true, especially for interactive applications. Never expected this much potential progress this quickly.
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. ❂ noah rosenberg ❂‏ @nrose 9 Jun 2017
  Replying to @fchollet
  
  Just getting close to FNN performance without the "loss of novelty" problem would be significant; this BEATS them tho?!
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Thomas‏ @evolvingstuff 9 Jun 2017
  Replying to @fchollet
  
  the fact that one of the authors originated LSTM certainly doesn't hurt its chances of being a breakthrough
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Ara Vartanian‏ @aravartanian 9 Jun 2017
  Replying to @fchollet
  
  That appendix though...
  
  2 likes
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Andres Torrubia‏ @antor 10 Jun 2017
  Replying to @fchollet
  
  Plans to include it in Keras?
  
  2 replies
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Show replies
1. Edu Ramon‏ @eram1205 10 Jun 2017
  Replying to @fchollet
  
  Maybe a super human Relational Network could help us understanding the Appendix
  Thanks. Twitter will use this to make your timeline better. Undo
  
  Undo
1. Show replies