1234 points by deepmind_ai 5 months ago flag hide 9 comments
johnsmith 5 months ago next
Fascinating! I've been waiting for a breakthrough in neural network compression. This could really help improve deep learning on resource-constrained devices.
machinelearing 5 months ago next
@johnsmith I completely agree! This method could have a significant impact on real-time AI systems.
mlresearcher 5 months ago next
@machinelearing So far the research is mostly focused on supervised-learning networks, but that doesn't mean it cannot be adapted for unsupervised-learning.
anonymous 5 months ago prev next
Would this method also be applicable for unsupervised-learning neural networks or will it work only with supervised-learning networks?
hackingai 5 months ago next
@anonymous The authors reported some successful results for unsupervised-learning networks, but further investigation is indeed warranted.
aienthusiast 5 months ago next
@hackingai Thanks for clarifying! I hope someone from the community explores this further and share the findings.
codingfanatic 5 months ago prev next
This approach seems to be very innovative, combining pruning and quantization. Will be looking forward to testing this on my projects.
deeplearner 5 months ago next
@codingfanatic I've started implementing this on my models, and the initial results are promising. Good choice!
quantumcode 5 months ago next
@deeplearner It's great to hear the promising outcome. Looking forward to more updates!