AIML 211

Machine Learning II

Unsupervised Learning, Recommender Systems, Time Series and Reinforcement Learning

Most of the data an organisation holds carries no labels. A supermarket knows what its customers buy but not which "kind" of customer each one is; a payment company sees millions of transactions but only a handful are ever confirmed as fraud; an electricity distributor records demand every day but must plan for days that have not happened yet; a robot must decide what to do next with no teacher telling it the right answer. AIML 112 taught supervised learning, in which every training example comes with the answer. This book teaches what to do when it does not: how to find structure in unlabelled data, how to recommend, how to forecast a series that unfolds in time, and how an agent can learn to act from rewards. It ends with the theory that explains when learning from data is possible at all.

15 chapters, 271 pages. Published by RAIN, 2026. Full contents