4 ms·
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 de
by lucidrains 10y ago
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)?
- bytefactory 10y agoSo...you mean a grad student? ;)
- Ar-Curunir 10y agoIt's called Grad student descent (referring to the grad student's descent into despair :P)
- aisofteng 10y agoLearning to learn by gradient descent by gradient descent: https://arxiv.org/abs/1606.04474 https://arxiv.org/abs/1606.04474