Tweeted By @hardmaru
Deep Double Descent: Where Bigger Models and More Data Hurt
— hardmaru (@hardmaru) December 6, 2019
They conduct an empirical study of the ‘double descent’ phenomenon in neural nets, and investigate this behavior in a range of architectures and its relationship to model size and training time.https://t.co/C9RfiRLlO4 pic.twitter.com/GnLyaExkQ1