Research Seminar in Statistics: Bayesian computation: state of the art and recent developments (SE)
Course Description
Bayesian computation mainly revolves around computation of the posterior distribution. Often these cannot be computed analytically in closed form, especially for big and complex datasets. Hence, approximations became more and more popular. Recent decades have seen enormous improvements in computational inference for statistical models, with both theoretical and algorithmic innovations opening new opportunities to practitioners.
In this seminar, we will discuss state of the art methods like MCMC algorithms but also recent developments in approximate Bayesian techniques, such as ABC or variational Bayes. We will aim for both: A good theoretical understanding of the presented methods as well as showing how helpful such methods are in Economics, Life Sciences, and other fields.
Prerequisites
Good knowledge of Statistics is recommended for this course. Basic knowledge of Bayesian Statistics is desired, but not mandatory since there will be a brief introduction at the beginning of the term.
Course Learning Objectives
Literature will be given later, a nice overview can be found in https://link.springer.com/article/10.1007/s11222-015-9574-5
Organizational matters
Max. 20 participants. Please register with Lucas Kock (lucas.kock@hu-berlin.de) until Oktober 1st, 2019 (with student number and the degree program).
First meeting, including topic assignments: Thursday, 10th of October at 12 pm (s.t.) in room 112 (Spandauer Str 1). If you are unable to attend, please send an e-mail to Lucas Kock. You will then be considered for the topic assignment.
Each student can contribute at most one seminar 7010319 (Prof. Greven) and one seminar 7010324 (Prof. Klein) to the module "Research Seminar in Statistics", regardless of the varying topics.
Course Structure
For details please see Moodle.