This course begins with an overview of what machine learning is and why it is important. This is illustrated with several real applications in various media, e.g., text summarization, sound and music recommendation and image retrieval. The course then presents the workflow of machine learning development that serves as an overview of the remainder of the course. The course presents the two general classes of machine learning methods: supervised learning (for example closest neighbour, decision tree) and unsupervised learning (e g k-means clustering, principal component analysis). For these, the course presents different types of modelling: parametric (e.g. Bayes, least squares) and non-parametric (for example closest neighbours, decision trees). The course reviews common methods for evaluation of machine learning models (e g holdout, bootstrap). Finally, best practices are presented (e.g. partition) together with common pitfalls (e g over fitting).

After passing the course, the students should be able to:

• develop and modify media technology applications that use machine learning and evaluate them in an appropriate manner,
• recommend methods for machine learning for particular media technology applications,
• describe and explain the machine learning pipeline,
• explain and contrast supervised and unsupervised learning methods,
• explain and contrast parametric and non-parametric methods,
• explain training validation and testing of machine learning models,
• summarize best practice and pitfalls in applied machine learning for media technology.

in order to

• being able to apply and evaluate machine learning models and methods in media technology.