The Lottery Ticket Hypothesis suggests that, even before training, a neural network can be pruned to a fraction of its original size, and still achieve comparable results. It's implications are very important for architecture design and training efficiency.
I'll be creating a video on the update to StyleGAN 2 (ADA) soon, so be sure to watch for that!
Sources:
Paper: https://arxiv.org/abs/1803.03635
Music:
With a Stamp by Twin Musicom is licensed under a Creative Commons Attribution license (https://creativecommons.org/licenses/...)
Artist: http://www.twinmusicom.org/