Course type: compulsory
Lecturer: Assistant Professor Erik Štrumbelj

For the timeline see Curriculum.

Prerequisites:

• Enrolment into the first year of the programme.

Content (Syllabus outline):

Prediction: linear regression, logistic regression, LDA/QDA, nearest neighbors, evaluating goodness of fit.

Feature and model selection: cross-validation, bootstrap, filter methods, wrapper methods.

Advanced prediction: basis expansions, splines, regularization, decision trees, generalized additive models, local regression.

Combining models: bagging, boosting, random forests, ensemble learning.

Support Vector Machines: for classification, for regression, optimization, duality, RKHS (reproducing kernel Hilbert spaces).

Neural networks: fitting neural networks, overfitting and other computational challenges.

Objectives and competences:
The methods covered in this course are fundamental to prediction, clustering and other quantitative data analysis tasks. Knowledge of these methods is key to applications of machine learning and understanding advanced machine learning methods. The course is also relevant to statisticians that do not specialize in machine learning, because it offers a set of new tools for data analysis.

Intended learning outcomes:

Knowledge and understanding: Understanding the basic concepts of machine learning.

Application: Classical machine learning methods are indispensable in modern data analysis and the foundation on which we can build a good understanding of advanced machine learning methods.

Reflection: understanding of the theory on the basis of examples of application. Razumevanje povezav med strojnim učenjem in statistiko.

Transferable skills: Analytical ability. Ability to solve practical data analysis problems.