SF1930 Statistical Learning and Data Analysis
KTH Royal Institute of Technology
Completed basic course in probability and statistics (SF1918, SF1922 or equivalent).
The course gives an introduction to the theory of statistical inference and prediction, which constitute the main goals for modern statistical data analysis and machine learning. Particular attention is given to multidimensional probability distributions and exponential families, which are fundamental tools for modeling data analytical problems, and the theory of graphical models is a powerful means for describing conditional dependencies with a bearing on high-dimensional statistical inference problems. Decision theory provides a framework for making optimal decisions under statistical uncertainty, as well as weighting different statistical approaches against each other. In particular, Bayesian decision theory—in which the inference and learning problems are solved through calculation of the posterior and predictive distributions, respectively—plays a central role in today’s statistical data analysis and is used to construct Bayesian point estimates, hypothesis tests, and credibility intervals. In parallel with the Bayesian approach, likelihood theory is also discussed, and special attention is given to the asymptotic properties of the maximum likelihood estimate as the amount of data grows towards infinity. The course also introduces basic statistical computation methods, such as stochastic gradient methods and Markov chain Monte Carlo (MCMC) methods, which are of great importance in modern computer-intensive statistics. In the course, these are applied to real data-analytical problems within the framework of a computer-based project.
After having passed the course, the student is supposed to be able to:
- formulate and apply concepts in statistical inference and prediction to solve theoretical problems;
- formulate and apply concepts in statistical inference and prediction to solve problems in data analysis;
- design and implement methods in statistical learning for data analysis.
Reviews
Improve accuracy by rating this course