“Deep Learning has outlived its usefulness as a buzz-phrase”

Re:“Deep Learning has outlived its usefulness as a buzz-phrase” by platz on Friday January 05, 2018 @10:01PM (#16083601)

> It's really very much like a regular progam, except it's parameterized, automatically differentiated, and trainable/optimizable.

> People are now actively working on compilers for imperative differentiable programming languages.

Do you have an example of either of these things?

[ Parent | on HN ]

Re:“Deep Learning has outlived its usefulness as a buzz-phrase” by joshuamorton on Friday January 05, 2018 @10:32PM (#16083689)

I believe https://github.com/google/tangent counts, though I'm not 100% sure.

[ Parent | on HN ]

Re:“Deep Learning has outlived its usefulness as a buzz-phrase” by zengid on Tuesday January 09, 2018 @01:04PM (#16108210)

From the thread, LeCun mentions:

"Look at papers by Jeff Siskind and Barak Barak A. Pearlmutter particularly their work on VLAD, Stalingrad and VLD."

[ Parent | on HN ]

Re:“Deep Learning has outlived its usefulness as a buzz-phrase” by chombier on Saturday January 06, 2018 @07:45AM (#16085047)

Erik Meijer gave a talk about this at KotlinConf last year https://www.youtube.com/watch?v=NKeHrApPWlo

[ Parent | on HN ]