Skip to content

Comment on PathNet: Evolution Channels Gradient Descent in Super Neural Networksparent

Comments

I assume you're being sarcastic; they do point out in the intro and at the end that a deep RL agent could be trained to do the topology selections, but that would be more work to get going than some simple evolutionary operators, and is left to future work. Don't worry, I'm sure it'll be A3C all the way down eventually...

Well yes, you could use neural net for the tournament selection, but I was thinking of a much dumber competition that involves a whole lot more distributed GPU power.

I think you might have to actually explain what you're thinking of

loss function tournament winner replacing hash winner in something like bitcoin mining

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.