Hacker News new | ask | show | jobs
Revisiting Small Batch Training for Deep NNs; batch_size less than 64 is best (arxiv.org)
1 points by evc123 2982 days ago