Conversation

JSON Baumgartner

@jasonbaumgartne

What do my fellow data scientists think of

? Do you use it for any of your datasets?

2:33 AM · Jul 1, 2018Twitter Web Client

Sergei Sokolenko

Replying to

@jasonbaumgartne

and

yup, it's a great way to build Tensorflow models, I used it for my Reddit research (using your data). Recommend using Keras as an abstraction layer - makes it really easy. GPUs are free in Kaggle and run about 20 times faster than regular CPU. I am datancoffee on Kaggle.

1

2

JSON Baumgartner

@jasonbaumgartne

Replying to

and

Thank you for that insight!

Replying to

@jasonbaumgartne

and

i don't use it except for browsing the notebooks that present examples of code to do something with a given dataset (so called "kernels"). Those are useful, a little bit like d3 "blocks" For data itself i'd say useful simply b/c it centralizes resources, as github did for code...