Stockholm

Based on the topics discussed in first part of the course, we continue with deepening the theoretical basis for stochastic optimisation algorithms. Specifically, we discuss theory around Stochastic Gradient Ascent (including momentum and adaptive step sizes), Simulated Annealing, and Particle Swarm Optimisation. Theoretical results on convergence and speed will be discussed.

See http://gauss.stat.su.se/phd/oasi/optimisation2.html for more information.

We will discuss algorithms to compute minima or maxima which are frequently needed in statistics and machine learning. Topics of lectures on four occasions: gradient based algorithms, stochastic gradient based algorithms, gradient free algorithms (e.g. particle swarm optimisation), handling of restrictions during optimisation. Course homepage: http://www.adoptdesign.de/optimisation1.html

This is an advanced course in Bayesian statistics for PhD students in statistics, computer science, the engineering sciences and other related fields.
The course is divided into 4 contemporary topics in Bayesian analysis, and the choice of topics can vary from year to year depending on the research frontier. Students can pick and choose among the topics and will be given 2 credits for each selected topic.
The planned topics for the current year are:

1. Gaussian Processes with Applications
2. Bayesian Nonparametrics
3. Variational Inference

The aim is to provide some theory concerning asymptotic methods of statistics and probability with applications to inference problems. Knowledge of measure theory is not needed. Basic convergence concepts and results (with proofs) such as the Lindeberg-Lévy central limit theorem should have been covered in previous courses.
To sign up for the course send an e-mail to per.gosta.andersson@stat.su.se

Teacher and examiner: Dan Hedlin, Department of Statistics, Stockholm University. .

Topic

We focus on design-based sampling with the Horvitz-Thompson and the generalised regression estimators. (Advanced Survey Sampling II will go further into model assisted estimation). The main course book is Särndal, Swensson and Wretman (1992); for part I we limit ourselves to Chapters 1-7. Although the focus is on design-based sampling we will put this inferential framework into some wider context.

We are happy to invite to the course on Longitudinal Data Analysis (3hp). The course is funded by GRAPES  but all interested in Longitudinal Data Analysis can participate. The course will present material on Repeated Measurements, Mixed Linear Models, Generalized Linear Models, Model Validation and Data Analysis. All together the course comprises 12h lectures and 12h computer classes.

GRAPES will refund travel and accomodation costs for PhD students at Swedish universities.

This masters course introduces basic concepts, ideas and methods of Bayesian inference. It is given in the late fall at Stockholm University as part of the masters programme and in the late spring at Linköping University (then under the name Bayesian Learning) as part of the master's course in the programme 'Statistics and Machine Learning'. The focus is on models and methods in the field of machine learning, but most of the course content is of general statistical interest.
The course will be in English.