Gated Linear Networks (GLNs) are backpropagation-free neural nets designed for efficient online learning via local convex optimisation. We show their performance, interpretability & robustness to catastrophic forgetting on classification tasks - read here: https://t.co/8NTNDkXZDT
— DeepMind (@DeepMind) June 15, 2020
from Twitter https://twitter.com/DeepMind
June 15, 2020 at 02:24AM
via IFTTT
No comments:
Post a Comment