Statistical learning with high-dimensional data

Education facts

Code: 5MS084

This course provides comprehensive knowledge, both regarding breadth and depth, about data science and statistical learning. In the course, both traditional and state of the art methods and algorithms in these fields are discussed. The related fundamental theories are also covered. After passing the course, the students should have a strong ability to solve problems through data. Meanwhile, students are also expected to have a strong self-study ability for understanding and learning any newly developed methods and algorithms. Module 1 (3hp): Theory Three families of approaches for dimensionality reduction are covered: spectral based learning (multi-dimensional Scaling, Isomap, Kernel PCA, etc.), manifold learning (Locally linear Embedding, Hessian Eigen-mapping, t-distributed stochastic neighbor embedding, etc.), and deep neural network-based methods (Autoencoders, Variational autoencoder, etc.). As special cases of dimensionality reduction, different feature selection methods, such as Ridge regression, LASSO, and Feature importance are also discussed. Supervised learning approaches including the Kernel-based methods (Kernel ridge regression, Support Vector Machine, etc.), Ensemble methods (Random Forest and Adaboost), Neural Networks, and different Deep Learning approaches and architectures are discussed. Furthermore unsupervised learning approaches including different clustering analysis algorithms, such as Density-based methods and Spectral clustering analysis are included. Deep learning-based unsupervised learning methods, such as Generative adversarial networks and its variations are also covered. Finally, fundamental mathematical theories about kernel methods, ensemble methods, penalty approaches, shallow network, gradient descent algorithm, universal estimator, and fundamental theorem of learning, etc. are discussed. Module 2 (4.5hp): Computer labs The module covers the analysis of several data sets, using the statistical methods that are included in the course. The analyses are conducted in one of the the programming languages R or Python. In the module, students write thorough reports of the analyses and the results from them.

Entry requirements

The course requires 90 ECTS including 7,5 ECTS Computer Programming, 7,5 ECTS Multivariate Data Analysis and 12 ECTS Mathematical Statistics or equivalent. Proficiency in English and Swedish equivalent to the level required for basic eligibility for higher studies.

Education offerings

Each offering has its own dates and conditions. Closed offerings are retained as history and do not mean that a new application is open.

• Statistical learning with high-dimensional data

  Umeå University

  Umeå

  Start date: 2 November 2026

  End date: 17 January 2027

  Pace of study: 50 %