News for hackers. Stuff that matters.

A comment on Unsupervised Learning with Even Less Supervision Using Bayesian Optimization. Parent comment

Re:Unsupervised Learning with Even Less Supervision Using Ba... by pizza on Friday March 11, 2016 @04:02PM (#11269505)
Actually, out of curiosity, would there be some way to use the inverse coloring transform + a lil noise to generate some kind of equivalence class of free training examples, sort of a la skip-gram?
[ Parent | on HN ]
Re:Unsupervised Learning with Even Less Supervision Using Ba... by idewanck on Friday March 11, 2016 @04:12PM (#11269566)
Author here, you can definitely talk about augmenting your training data with some slight transforms of the labelled set you have. Common strategies are adding noise, rotation etc for images
[ Parent | on HN ]
Re:Unsupervised Learning with Even Less Supervision Using Ba... by pizza on Friday March 11, 2016 @04:49PM (#11269793)
Interesting. I guess the question then becomes what constitutes a "big" transformation that preserves relevant invariants.
[ Parent | on HN ]