News for hackers. Stuff that matters.

A comment on Ask HN: What are your biggest pain points as a data scientist?.

Re:Ask HN: What are your biggest pain points as a data scien... by airza on Thursday April 11, 2019 @01:41PM (#19637135)
Getting the dimensionality correct between layers in an untyped language is incredibly painful. Getting a prototype working on my personal machine and then having to play AMI roulette on AWS to get it running on a GPU is frustrating and expensive. Every time I read the tensorflow documentation I think, "I wish i could pay 500 dollars for a version of this that looked like someone cared about it."
[ Parent | on HN ]
Re:Ask HN: What are your biggest pain points as a data scien... by gtrevize on Thursday April 11, 2019 @04:01PM (#19638550)
You can skip AMI roulette by using AWS Deep Learning AMIs (https://aws.amazon.com/machine-learning/amis/) or the new container equivalent (https://aws.amazon.com/machine-learning/containers/)
[ Parent | on HN ]
Re:Ask HN: What are your biggest pain points as a data scien... by chewxy on Thursday April 11, 2019 @05:12PM (#19639267)
Not related to TF, but the upcoming version of Gorgonia has a solution for that - we basically developed a calculus for dimensionality of data. Gorgonia is deep learning in Go, and is well typed as far as I can tell.

Work on the coq program proving the calculus is taking longer than expected though.

If you're interested in contributing to Gorgonia, lmk.

[ Parent | on HN ]