Mathematical Foundations and Algorithms for Machine Learning, 7.5 credits

Learning outcomes

Knowledge and Understanding
After completed studies, the student shall be able to

• the most important mathematical concepts and algorithms in the area of machine learning,
• the basic feed forward neural network architectures, and
• the most important algorithms within the area of mathematical optimization and regularization used in training neural networks.

Competence and Skills
After completed studies, the student shall be able to

• apply the algorithms described during the course on simple applied problems, and
• train simple network architectures with respect to overfitting and regularization techniques.

Content

Mathematical aspects of deep learning. Feed forward neural networks. Backpropagation and automatic differentiation. Stochastic gradient descent. Regularization techniques. Linear Regression. Dimensionality reduction. Principal component analysis. Density estimation. Classification. Support vector machines.

Examinations and grades

Examination, 7.5 credits (Code: A001)
Grades used are Fail (F), Sufficient (E), Satisfactory (D), Good (C), Very Good (B) or Excellent (A).

According to the Higher Education Ordinance, Chapter 6, Section 18, a grade is to be awarded on the completion of a course, unless otherwise prescribed by the university. The university may determine which grading system is to be used. The grade must be determined by a teacher specifically nominated by the university (the examiner).

In accordance with university regulations on grading systems for first and second-cycle courses and study programmes (Vice-Chancellor’s decision ORU 2018/00929), one of the following grades is to be used: fail (U), pass (G) or pass with distinction (VG). For courses included in an international master’s programme (60 or 120 credits) or offered to the university’s incoming exchange students, the A to F grading scale is to be used. The vice-chancellor, or a person appointed by them, may decide on exceptions from this provision for a specific course, if there are special grounds for doing so.

The grades used on this course are Fail (F), Sufficient (E), Satisfactory (D), Good (C), Very Good (B) or Excellent (A).

Modes of assessment

• Examination (code A001): Written assignment

For students with a documented disability, the university may approve applications for adapted or other modes of assessment.

For further information, see the university's local examination regulations.

Specific entry requirements

Multivariate Calculus, 9 credits or Multivariate Calculus for MSc in Engineering, 7.5 credits, and Optimization, 7.5 credits, or Optimization for MSc in Engineering, 7.5 credits, as well as Statistics, Statistical Theory, Intermediate Course, 15 credits or Mathematical Statistics and Probability, 7.5 credits.

For further information, see the university's admission regulations.

Other provisions

All or part of the course can be given in English.

Students who have been admitted to and registered on a course have the right to receive tuition and/or supervision for the duration of the time period specified for the particular course to which they were accepted (see, the university's admission regulations (in Swedish)). After that, the right to receive tuition and/or supervision expires.

Reading list and other learning resources

Required Reading

Deisenroth, Marc Peter, Faisal, A. Aldo & Soon Ong, Cheng (2019)
Mathematics for Machine Learning
Cambridge University Press
https://mml-book.github.io/book/mml-book.pdf

Goodfellow, Ian, Bengio, Yoshua & Courville, Aaron
Deep learning
MIT press
https://www.deeplearningbook.org/

Study materials provided by the Mathematics Unit.