Tweeted By @dennybritz

Overfitting is always a good first step. Overfitting on training data and making sure the results look reasonable and that there are no obvious bugs is what everyone should do first for a new task/model :)

— Denny Britz (@dennybritz) September 17, 2018