Probabilistic Modelling and Reasoning
Course homepage
This is a course for MSc level students. The course descriptor can
be found
here.
The
introductory
handout for the course is available.
MATHS PREREQUSITES: To do PMR you need a
reasonable background in statistics, calculus, linear algebra.
You should thoroughly review the maths in the following
cribsheet (from Iain Murray, UCL)
before the start of the course.
A handout on
Mathematical
Preliminaries which used to be used for the course LfD1 is also
useful for PMR; all material is relevant.
Lectures
PMR lectures will be
in
7 Bristo
Square LT1
on Tuesdays at 1000-1050 and
in
Appleton Tower LT 1 on Fridays at 1000-1050.
Information taken from
Informatics
2011/12 lecture timetable.
Resources
-
Books. The course textbook is Pattern
Recognition and Machine Learning by Christopher M. Bishop,
Springer (2006).
- Other useful texts for some of the material are
Information Theory, Inference and Learning Algorithms by
David J. C. MacKay (Cambridge University Press, 2003), and
Artificial
Intelligence: A Modern Approach by S. Russell and P. Norvig (Prentice Hall,
second edition, 2002).
-
A web
page giving details of some books with maths background,
and some web-based resources
-
Links
to more advanced mathematical material.
-
A handout Introduction
to MATLAB giving an introduction to MATLAB and the NETLAB neural networks
toolbox. Further MATLAB tutorials are available at
UNH
Matlab Tutorial, US
Navy Matlab Tutorial and MTU
Introduction to Matlab
-
A Brief
Introduction to Graphical Models and Bayesian Networks by Kevin Murphy.
-
Thomas Minka's excellent
Statistical
Learning/Pattern Recognition Glossary
-
Max Welling's Classnotes in Machine Learning
-
The Association for Uncertainty in Artificial
Intelligence homepage is a good place to start looking for interesting
material.
-
The Kalman Filter
website.
-
List
of Public Domain Belief Network Tools from a course on Belief Networks
at Duke University.
- Zoubin Ghahramani's tutorial/overview paper
Unsupervised Learning (original
link), in Bousquet, O., Raetsch, G. and von Luxburg, U. (eds) Advanced Lectures on Machine Learning LNAI 3176, Springer-Verlag (2004).
Assignments
There will be two assignments in this module worth in total 30%
of the mark for the course.
-
Assignment 1 is due in on
Mon 31 Oct by 4pm by manual submission to the ITO.
Here is the javabayes file
racing.
Feedback after 2 weeks (approx).
-
Assignment 2
(updated 18 Nov), deadline extended to
Tue 6 Dec 4pm, manual submission to the ITO.
Here is the associated tarball of software
a211.tar
(updated 18 Nov), and here is the
updated fa.m function
(updated 18 Nov).
Feedback after 2 weeks (approx).
Tutorial groups
Tutorials will start in week 3.
Tutorial groups listing from ITO. This may only be visible
from within .ed.ac.uk. In case of problems with the
assignment to specific groups contact the ISS via the
support form .
Course Rep
Xiao Liu s1100818
Software
We will be using the program JavaBayes, you can find more details
here. We will also be using some MATLAB code.
Office hour
Friday 11-12, starting in week 2.
It is probably best to catch me
after the lecture. I will then walk over to my office in IF 2.27.
No office hour on Fri 25 Nov or Fri 2 Dec, but there will be
one 10-11 on Tues 29 Nov in IF 2.27.
Week-by-Week listing
Will build up here ...
Week 1
Lectures
Introduction
slides
slides4up,
Belief networks
slides
slides4up
Self-check maths sheet Check if you can do the
questions on
tut0.pdf
Tutorials No tutorials in week 1 (nor week 2)
Week 2
Lectures Belief networks continued.
Tutorials No tutorials in week 2, but see sheet below for
the tutorial in week 3
Handout
Worked example for Holmes/Watson
network
JavaBayes example
sprinkler.bif (Bayes net for the
Holmes-Watson-Rain-Sprinkler problem)
Week 3
Lectures
Elimination algorithm
slides,
slides4up;
Junction tree algorithm
slides
slides4up
Handout Worked example of
inference in a junction tree
Handout
Worked example for c->b->a network
Tutorials tut1.pdf
answer sheet
ans1.pdf,
wikipedia
on Monty Hall problem
Week 4
Lectures
Gaussian distribution
slides,
slides4up,
Maximum likelihood estimation
slides,
slides4up
Handout on
Inference
with Gaussian Random Variables
Tutorials tut2.pdf,
answer sheet
ans2.pdf
Week 5
Lectures
Bayesian parameter estimation
slides
slides4up,
Decision theory
slides,
slides4up,
Mixture models
slides,
slides4up
Handout
Working for Gaussian
classifier
Matlab code cointoss.m,
matlab code to illustrate posterior distribution under Beta prior
Web resource The
technical
report by David Heckerman entitled "A tutorial on Learning Bayesian
Networks". (
original
link)
Tutorials tut3.pdf,
answer sheet
ans3.pdf
Week 6
Lectures
Mixture models continued,
Factor analysis and beyond
slides,
slides4up
(updated 4 Nov),
NO LECTURE ON FRIDAY 28 OCT
Handout on
Factor Analysis and Beyond
Handout
Working for EM for
Mixture of Gaussians
Handout Working for PCA
solution as principal eigenvector
Tutorials tut4.pdf, matlab code
plotquad.m,
answer sheet
ans4.pdf
Week 7
Lectures Factor analysis and beyond ctd,
Hidden Markov models
slides
slides4up
Tutorials tut5.pdf,
answer sheet
ans5.pdf
Web resource
Te-Won
Lee's Blind source separation demo
Web resource Short explanation
of blind source separation from Helsinki
Optional reading
Paper on
GTM
by C. M. Bishop, M. Svensen and C. K. I. Williams (
original
link)
Week 8
Lectures
Hidden Markov Models continued,
Time Series Modelling and Kalman Filters
slides,
slides4up
(updated 11 Nov)
Handout
Working for alpha and beta
recursions for HMMs
Web resource L Rabiner tutorial on HMMs from Proceedings of the
IEEE 77(2) 1989 is available from the IEL electonic library
here.
Optional reading
A Gentle Tutorial on the EM Algorithm and its Application to Parameter
Estimation for Gaussian Mixture and Hidden Markov Models by Jeff A. Bilmes
pdf
(
original
link)
Web resource
Harmonising chorales in the style of Johann Sebastian Bach, work
by Moray Allan using a HMM (MSc, School of Informatics, Edinburgh,
2002). See also
HMM Bach demo.
Handout on
Time Series Modelling
Web resource Movie
clips from Prof Andrew Blake's group illustrating tracking with non-linear
Kalman filters
Tutorials tut6_1112.pdf,
answer sheet
ans6_1112.pdf
Week 9
Lectures
Bayesian Model Selection
slides,
slides4up,
Undirected graphical models
slides,
slides4up
Tutorials tut7_1112.pdf,
answer sheet
ans7_1112.pdf
Handouts
Working for local Markov
property,
Working for Boltzmann machine
conditional distribution,
Working for derivative of a
log-linear model
Web resource
chapter
28 from David MacKay's book Information Theory, Inference and
Learning Algorithms. See sections 28.1, 28.2 on model comparison
Web resource
Information
about GrabCut from MSR Cambridge
Week 10
Lectures Last lecture on Tues 22 Nov.
Coding and Information Theory
slides,
slides4up
(not examinable)
Tutorials No tutorials in week 10
This page is maintained by Chris
Williams
