bild
Skolan för
datavetenskap
och kommunikation
KTH / CSC / Kurser / DD2447 / statmet09 / Kursanalys

Kursanalys: Statistical Methods in Applied Computer Science, statmet09

For the 2008 analysis, see 'andra omgångar'

Statistical Methods in Applied Computer Science, 2D1447 4p, 6 ECTS(hp).

Course given in Spring 2008

Instructor: Stefan Arnborg

Lectures: 18 hours
Recitals: 12 hours
Individual Advising: ca 15 hours

Registered students: 18 Master's students(DD2447)+4 doctoral student(FDD3342)

Course Literature:

Lecture Notes: Statistical Methods in Applied Computer Science
Ed Jaynes Lecture notes, Ch 1-3 and 5
This year material on recommender systems (Bayesian SVD and Amazon style) technology was added. The compendium was updated with more on Gammermann/Vovk scheme. Seems quite popular.

EXAMINATION

Course has one examined moment, labelled as Homework
All 18 Master's students passed, currently 4 with distinction (grades A,B)
most students where non-Swedish speaking, so teaching was in English, Advising in Swedish or English.

Throughput: 100% (14 Master's students had passed to grade C after last lecture, by doing 6 weekly homeworks). One student achieved higher grade B by doing one Master's tests, and three students obtained grade A after two Master's tests or one project.

Course Goals:

This course summarizes statistical and probabilistic methods used in applied Computer Science - statistical aspects of Data Mining, Knowledge Discovery, Machine learning and Information Fusion with a Bayesian outlook.

LEARNING GOALS

After successfully taking this course, you will be able to:

-motivate the use of uncertainty management and statistical methodology in computer science applications, as well as the main methods in use,

-account for algorithms used in the area and use the standard tools,

-critically evaluate the applicability of these methods in new contexts, and design new applications of uncertainty management,

-follow research and development in the area.

Changes made in 2009:

No change, but some extra rehearsal was introduced. Lectures and material on Gammerman/Vovk inference and recommender systems was popular.

SUMMARY

I liked giving the course, few problems except that I think people did not like to see how nuiscance parameters are eliminated by integration in joint inferences of variance and mean for gaussian mixture modelling. I will try to skip that next time.

EXAMINATION

OK. individual examination is however somewhat time-consuming

LITERATURE:

Seemed to work fine. I will do some refinements in the compendium, but just now I have not found new stuff to throw in. Maybe oracle property in feature selection, a hot but difficult topic in high-throughput bioinformatics.

Copyright © Sidansvarig: Stefan Arnborg <stefan@nada.kth.se>
Uppdaterad 2009-09-08