Statistical data analysis 2 1000-718SAD

Syllabus
1. frequentist vs bayesian approach in statistical modeling
2. bayesian networks (probabilistic graphical models)
3. parameter inference in probabilistic graphical models with fully observed data
4. EM algorithm (parameter estimation in models with hidden variables)
5. Markov chains and Hidden Markov 6. Exact inference in graphical models (factor graphs, the sum product algorithm, Cluster trees, potentials, Message passing, Junction tree algorithm)
7. model selection, model evidence, learning model structure, tree models, general models, structural EM
8. Sampling (MCMC, Gibbs sampling)
9. variational inference.

Course coordinators

Konrad Szczawiński

Type of course

obligatory courses

Prerequisites (description)

R programming, passed statistical data Analysis 1, English (advanced)

Learning outcomes

Machine learning and statistical inference, focused on probabilistic graphical models

Assessment criteria

Rules for passing the course:
Scoring:
50% exam at the end (a test)
15% computational project 1
15% computational project 2
Mid-term test 15%
5% lab activity
Required to pass: 50%
Zero exam: oral, the date is agreed individually, no later than a week before the final exam.
Criteria for admission to the zero exam: 45 points for projects and test.

Bibliography

Pattern Recognition and Machine Learning, C. Bishop

Probabilistic Modeling in Bioinformatics and Medical Informatics, D. Husmeier, R. Dybowski and S, Roberts