Hacker News Reader

So when are we going to see the paper where we use an RL net to speed up another RL net for discovering a neural architecture for learning how to do gradient descent (by gradient descent)?

View on HN

Learning to learn by gradient descent by gradient descent: https://arxiv.org/abs/1606.04474

So...you mean a grad student? ;)

It's called Grad student descent (referring to the grad student's descent into despair :P)

Reply on news.ycombinator.com