Deep Learning Rules of Thumb

Really good article covering some of the basic architecture/hyperparameter configurations