Improving Deep Neural Networks: Tuning, Regularisation and Optimisation (Andrew Ng)

Course 2 of the Deep Learning Specialization in Andrew Ng's lecture videos: 34 lessons, about 5 hours, on making neural networks actually work well. Week 1 covers train, dev and test sets, bias and variance, the basic recipe for machine learning, L2 and dropout regularisation, input normalisation, vanishing and exploding gradients, weight initialisation and gradient checking. Week 2 covers mini-batch gradient descent, exponentially weighted averages, momentum, RMSprop, Adam and learning-rate decay. Week 3 covers hyperparameter tuning, batch normalisation, softmax classifiers and an introduction to TensorFlow. Essential practical knowledge for anyone training deep models.