Progressive Weight Pruning of Deep Neural Networks using ADMM — mentioned by janhenr

Sat, 09 May 2020 15:35:23 GMT

It is, unfortunately, not revolutionary at all. The Lottery Ticket Hypothesis won a Best Paper Award because the idea is very intuitive and elegant. However, the compression rates (9x) of this follow-up work on ImageNet are far from impressive. Take for example this paper: https://arxiv.org/abs/1810.07378 that reaches 30x compression rate on ImageNet. (Disclaimer: I know the original author and co-published a paper with him - not this one however)

Paper and discussion history. A mention is not an endorsement.

Progressive Weight Pruning of Deep Neural Networks using ADMM · HackerPapers discussions

Progressive Weight Pruning of Deep Neural Networks using ADMM — mentioned by janhenr