Neural Networks and Deep Learning (Andrew Ng, DeepLearning.AI)

Course 1 of the Deep Learning Specialization in Andrew Ng's own lecture videos: 43 lessons, about 6 hours. It explains what a neural network is and why deep learning took off, then treats logistic regression as a tiny neural network: cost function, gradient descent, derivatives, computation graphs and vectorisation in Python. Later weeks build shallow and then deep networks, covering activation functions, random initialisation, forward and backward propagation and the difference between parameters and hyperparameters. Still one of the clearest foundations for anyone moving from machine learning into deep learning.