Applied Multi-Output Machine Learning

Multi-output learning is grounded on simultaneously predict multiple outputs given an input. Its modelling algorithms are very important to support decision-making, since making decisions in the real world often involves multiple complex factors and criteria. Beyond classification and regression solutions, Multi-Output research area deals with all steps of a Data Mining pipeline, e.g. selecting features with a multiple output constraint. Target: To introduce students to classic and state-of-the-art algorithms of Multi-Output based on applications and real-life case studies, as well as general questions related to analyzing and handling datasets with several outputs. Method: The course is split between theoretical foundations and practical exercises: Introduction to Multi-Output Learning - Multi-label problems (Classification Scenario) - Multi-target problems (Regression Scenario) - Mining multi-output scenarios. The practical exercises ask students to write original programs, as well as modify pre-coded examples in R or Python. Each meeting provides 4 hours of a subject. Exame: The student needs to deliver a Multi-Output project (prototype level) with preliminary discussion and insights.