Tweeted By @glagolista
training to overfit a new model on a small dataset (or existing model on new data) is prob the most important and useful rule in practical DL
— helena sarin (@glagolista) September 20, 2018
(just found a bug in my new code for synthetic data generator 😌)