Supervised and Online Learning Algorithms

Algorithms that learn from data to predict a value or choose an action. The k-nearest neighbors algorithm assigns a label or value to a new item from its closest training examples, sequential minimal optimization solves the quadratic programming problem that arises in training support vector machines, symbolic regression searches the space of mathematical expressions for the formula that best fits a dataset, and the universal portfolio algorithm learns adaptively from historical data to rebalance an investment portfolio in each trading period. An algorithm belongs here when its cited article describes learning a predictive model from known examples or updating decisions from data that arrives over time. It excludes clustering, which works without labels, decision tree and ensemble methods, and learning from rewards by trial and error, which have their own groups.