Machine Intelligence

Machine intelligence is the business of teaching devices to learn a task without being told explicitly how to do it — voice recognition, route planning, recommender systems, medical diagnosis, robot control, web search. The course works through the standard supervised toolkit and ends on reinforcement learning.

It starts with linear regression because almost everything later is a variation on the same argument: choose a model, write down what "wrong" means, then move the parameters downhill. Classification, kernels, and networks all reuse that shape.