Noisy Activation Functions — mentioned by vintermann

Thu, 15 Sep 2016 06:47:43 GMT

In this case, it's DeepMind's idea anyway :) I get more disappointed when the opposite happens. I think something like, "Yeah, I'm totally going to add support in torch for noisy activation functions like in this paper!" (https://arxiv.org/pdf/1603.00391.pdf). Then I procrastinate and put it off. Then I think, "No matter, someone else has surely done it by now". Then they haven't.

Paper and discussion history. A mention is not an endorsement.

Noisy Activation Functions · HackerPapers discussions

Noisy Activation Functions — mentioned by vintermann