News for hackers. Stuff that matters.

A comment on A group of ex-NSA and Amazon engineers are building a ‘GitHub for data’. Parent comment

Re:A group of ex-NSA and Amazon engineers are building a ‘Gi... by Nition on Thursday February 20, 2020 @02:31PM (#22377816)
Going from the parent comment's assertion, the answer would be that if you want to train something to predict physics, the original physics simulation that's generating the data is either:

1. Already doing that, in which case it's already the superior model.

or

2. Not doing that correctly, in which case the synthetic data output is equally poor.

[ Parent | on HN ]
Re:A group of ex-NSA and Amazon engineers are building a ‘Gi... by slavik81 on Thursday February 20, 2020 @07:19PM (#22379891)
Consider computer graphics and computer vision. If you had a photorealistic renderer, it could be useful for generating training data for a vision algorithm. However, you can't just run your rendering pipeline backwards to create a scene out of an image.
[ Parent | on HN ]
Re:A group of ex-NSA and Amazon engineers are building a ‘Gi... by mehrdadn on Thursday February 20, 2020 @02:33PM (#22377838)
I mean, if (1) is true but you get faster performance, then it's useful, right?
[ Parent | on HN ]
Re:A group of ex-NSA and Amazon engineers are building a ‘Gi... by aghillo on Thursday February 20, 2020 @03:28PM (#22378269)
Yes. A typical example would be the use of statistical emulators, emulating process (physical) based models that are computationally complex in the environmental sciences.
[ Parent | on HN ]