BI 123 Irina Rish: Continual Learning
Brain Inspired
00:00
Optimizations for Architectures
In the paper at nerips by a sergey bartuno ah and e jeff hinton they also kind of wore trying different alternatives, like target prop and averiances. And unfortunately, they couldn't show that it would beat or be comparable with a standard back prop ont, say, imagina. So there were the canof many er unsuccessful attempts. There were some limited successes, and the question is still open whether such alternatives can become true state of art. If you make those methods work, you will get benefits of much better paralization and skill ability.
Transcript
Play full episode
