≈ Looking for a Function
Deep Learning is so simple ……
Total Loss:
Find the network parameters ∗ that minimize total loss L
Training Example:
•Problem statement
•Gradient ascent
do not have to be differentiable It can even be a black box.
(|)=?