Deep Learning Best Practices: Activation Functions & Weight Initialization Methods -- Part 1