Spring semester 2015  
02433
Hidden Markov Models
 

5 ects point
Possible applications: Speech recognition, gene sequencing, object tracking, financial volatility models, animal behaviour, count data, wind power forecasting, etc.

The course is web-based and will primarily rely on self study by the student of the textbook, slides, and written exercises – the confrontation form and hours (between teacher and student) will be based on agreement between teacher and students. Questions can be posed on the DTU campusnet.

Course overview
Prerequisites
: 02417 – Time series analysis
Optional: 02407 – Stochastic processes

Textbook: Hidden Markov Models for Time Series – W. Zucchini & I. L. MacDonald, 2009

R source code from appendix and data sets.

Week 1:   Supplementary slides and solutions for selected exercises.
Week 2:  
Supplementary slides and solutions for selected exercises.
Week 3:   Supplementary slides and solutions for selected exercises.  Written exercise 1.
Week 4:   Supplementary slides and solutions for selected exercises.
Week 5:   Supplementary slides and solutions for selected exercises.
Week 6:   Supplementary slides and solutions for selected exercises.
Week 7:   Notes on state-space modelling with HMMs. Written exercise 2.  Data 1 - Data 2.  
Week 8:  
Supplementary slides and solutions for selected exercises.
Week 9:  
Notes on DNA copy No. data.
Week 10: Example on wind power analysis and forecasting. Winddata for exercise.     Solution to exercise.
Week 11: Wind direction at Koeberg (Chapter 12). Written exercise 3, wind power data.
Week 12:
Example on animal behaviour with feedback (Chapter 16). Cursory reading: Altman, R.M. (2007). Mixed hidden Markov models: An extension of the hidden Markov model to the longitudinal data setting. JASA, 102, 201-210.
Week 13:
Example on fish tracking using HMM.


For more information contact:

Jan Kloppenborg Møller   jkmo@dtu.dk