Content, progress and pedagogy of the module

Objectives

In designing and developing interactive media systems and technology, one is often faced with looking for interesting patterns and trends. This course presents theoretical concepts and practical tools for analyzing data for multimedia applications and solving machine learning problems, such as classification, in media technology. Many of these methods are used in, e.g., automatic speech recognition, face detection, web page ranking, autonomous driving, etc. The course includes the following topics: multivariate probability density functions, Bayesian classification, estimation, and detection, parametric (e.g., Gaussian density-based) and non-parametric classifiers (e.g. k-nn, parzen, convolutional neural networks), regression, data fitting, evaluation of classifiers and estimators, unsupervised  and supervised learning (e.g., reinforcement learning), feature selection and reduction. The course will contextualize these techniques by how they apply as tools for addressing media creation challenges.

Learning objectives

Knowledge

Students who complete the module will obtain:

• understanding of multivariate statistics and how to model multivariate data, e.g., using probabilistic and parametric descriptions
• understanding of the principles of supervised (e.g., Bayesian classification, SVM, least squares regression, deep learning) and unsupervised learning methods, (e.g., k-means, hierarchical clustering, Gaussian mixture models)
• understanding of features, feature selection, feature learning, and dimensionality reduction (e.g., forward feature selection, principal component analysis, autoencoder)
• knowledge of the application of machine learning techniques and tools to address media creation problems (e.g. visual effects, games, procedural generated content, motion capture etc.)

Skills

Students who complete the module will be able to: 

• choose, implement and apply machine learning methods to solve typical machine learning problems (e.g., classification, detection, regression)
• apply knowledge to compare machine learning methods in terms of performance and complexity
• apply the theory of multivariate statistics to analyze multimedia data (e.g., speech and music, images of faces, gestures, etc.)

Competences

Students who complete the module will be able to:

• apply multivariate statistics to analyze multimedia data, and reflect on a variety of possibilities to recommend a solution to the related machine learning problem(s)
• apply machine learning methods to such problems and evaluate, discuss and generalize the results and reflect on their implications regarding the problems and the data

Type of instruction

Refer to the overview of instruction types listed in § 17.

Exam

Prerequisite for enrollment for the exam

• To be eligible to take the exam, the student must timely have handed in any mandatory assignments

Exams