Master the training loop — optimizers, regularization, and normalization
Optimizing by following the slope downhill
Step decay, cosine annealing, warmup, and cyclical LR
SGD, momentum, and Adam — smarter gradient descent
Why normalizing activations speeds up deep network training
Xavier, He, and why random weights matter
Why deep networks struggle and how to fix it
Dropout, L1, and L2 — keeping models honest