Computational physics (Artifical and natural intelligence)
Bert Kappen

Last updated May 2010

Course information

Please contact the lecturers if you want to take this course.

This course will present various advanced computational topics in neural networks and/or machine learning, depending on the year and the interest of the students. Most of these topics require insights from statistical mechanics and stochastic processes.

Course material

Introduction to biophysics reader chapter 2 for a brief intro about neurons and the brain
Computational physics reader and sheets
David Mackay: Information theory, inference and learning algorithms (online version) can be used as background material on MCMC methods.

Weekly schedule

Date Topic Material Exercises
1 May 18 Introduction neurons and brain
Networks of binary neurons; Markov process; ergodicity
Reader Introduction Biophysics chapter 2
Reader Computational physics chapter 2
Section 2.5
2 May 25 Discussion of exercises
Boltzmann-Gibbs distribution
Mean field theory; Linear response theory
Reader Computational physics chapter 3 Section 3.7 1-3
3 June 1 No class
4 June 8 Discussion exercises ch 3
Linear response theory
Reader Computational physics chapter 3
Computer exercise 3.7 4 on mean field theory
initial matlab code
5 June 15 Discussion exercises ch 3.7 4
Inverse Ising models for brain modeling
Schneidman et al., Nature 2006
Roudi, Tyrcha and Hertz 2009
6 June 22 Bayesian reasoning
Simple expert system
MCMC methods
MacKay chapter 2, 29 Computer exercise combinatorial optimization

Solutions to exercises

Chapter 2
Chapter 3 2b.m 2c.m


Each computer exercise is summarized in a report. These reports and the results of the other exercises make up the final grade.