Course content

The course is a follow-up to TDT4173 Machine Learning. It gives thorough coverage of deep learning. The course covers both mathematical and computational foundation for deep learning, practical applications such as processing of images, text, and other modalities. Modern software frameworks for deep learning will be introduced and used for some projects, while other projects will require relatively low-level coding in Python or similar languages.

Learning outcome

Knowledge: General principles for learning/adaptive systems Mathematical and computational foundation for deep learning How to use deep learning in diverse practical applications Skills: Analyze different frameworks for deep learning in specific application domains Ability to analyze the mathematical foundation for diverse deep learning published in the literature Build computational systems that achieve deep learning General competences: Understand deep learning's basis in mathematics and cognitive science Understand possibilities and limitations of deep learning in practical settings

Learning methods and activities

Lectures, self study. Assignments will be published during the semester, from which a subset must be solved successfully to be accepted to be accepted for the exam.

Compulsory assignments

• Compulsory assignments

Specific conditions

Recommended previous knowledge

Some experience in Python programming

Required previous knowledge

TMA4115 Mathematics 3, TDT4120 Algorithms and data-structures, TDT4171 Methods in Artificial Intelligence, and TDT4173 Machine learning.

Course materials

Ian Goodfellow, Yoshua Bengio and Aaron Courville: Deep Learning (2016). MIT Press. Supplementary articles will be handed out as needed.