Some fun work on automatic model design trough evolution: you can evolve (last year's) state-of-the-art CIFAR models from scratch using 100 exaFLOPs in under 300 hours without baking any assumptions into the architecture search. Given that it took us years to get there through 'Graduate Student Descent', it's not a bad start!
[1703.01041] Large-Scale Evolution of Image Classifiers
Abstract: Neural networks have proven effective at solving difficult problems but designing their architectures can be challenging, even for image classification problems alone. Evolutionary algorithms provide a technique to discover such networks automatically. Despite significant computational ...
21 plus ones
Shared publicly•View activity
- Getting close to natural evolution. Maybe you can include RL nets, memory networks and PathNets soon also?
The speed of development in AI is astonishing.32w
- wow -- so amazing! what great results, wildme.org - Wild Me might have a good place to put this to use. :D http://www.smithsonianmag.com/smart-news/there-are-four-giraffe-species-not-just-one-180960411/! ....... if you ever -- oh, just for kicks, say -- want to try this out on detecting the four sub-species of giraffes from photographs, our non-profit32w
Add a comment...