123 points by ai_guru 7 months ago flag hide 9 comments
john_doe 7 months ago next
Fascinating! I've been following the progress in machine learning, and this seems like a major step forward in neural networks. I'm excited to see where this trains us in the future!
hackin_dave 7 months ago next
Brilliant! I've been working with neural networks for years and this truly changes the game. Thanks for sharing, and I can't wait to see it in action!
jane_q 7 months ago next
@hackin_dave I'm curious how your existing projects would be affected by this. Any insights?
python.coder 7 months ago prev next
From the article, this new training technique seems to speed up training by an order of magnitude, and even facilitate convergent networks. Truly impressive.
tensor_rt 7 months ago next
I wonder if applying the technique to NLP, computer vision, or reinforcement learning could change those fields, too.
science_101 7 months ago prev next
It took a lot of research on various branches of math, like non-convex geometry, to finally reach this point? I would love to read more about this.
math_junkie 7 months ago next
Yes, interdisciplinary research is what brought us these results! We also utilized optimization theory. Feel free to check out our linked papers for more information.
code_monk 7 months ago prev next
I can't believe that this has been hiding in front of us for the last decade! Amazing work. Not sure we really needed another Deep Learning framework, though.
innovative_geek 7 months ago prev next
As a research assistant at DeepLearnInc., a competing enterprise, I have to say I'm impressed with your work. Keep it up!