News for hackers. Stuff that matters.

A comment on Gaussian Distributions Are Soap Bubbles.

Re:Gaussian Distributions Are Soap Bubbles by Bromskloss on Saturday November 11, 2017 @10:21AM (#15676665)
Those images [0] inputs that were optimised to maximise a certain classification response were cool! Instead of going to this peak of the response function, is there a way to explore the shell where the actual images reside? Would such images look, to our eyes, more like real input than the optimised input? I suspect they won't, but I still would like to see what is between the dogs!

[0] http://www.inference.vc/content/images/2017/11/Screen-Shot-2...

[ Parent | on HN ]
Re:Gaussian Distributions Are Soap Bubbles by lebek on Saturday November 11, 2017 @10:40AM (#15676750)
This paper is sort of about that: https://arxiv.org/abs/1710.11381

They use a gamma distribution which has more probability density near the origin, which causes samples around the origin and interpolations to be more like real input.

[ Parent | on HN ]
Re:Gaussian Distributions Are Soap Bubbles by tobr on Saturday November 11, 2017 @10:49AM (#15676778)
The source is a great read: https://distill.pub/2017/feature-visualization/
[ Parent | on HN ]