itblogo

Lecture, Summer 2004
Computational Neuroscience IV: Analysis of Neural Systems
(Analyse neuronaler Systeme)

 hulogo

Lectures and Tutorials: Dr. Richard Kempter, Prof. Andreas V.M. Herz, and Dr. Martin Stemmler
Programming Course: Tiziano Zito

Time and Location:
Lectures: Mondays from 12:15 to 2:00 pm in the seminar room 1322 of the ITB, Invalidenstr. 43
Tutorials: Tuesdays from 8:30 to 10:00 am in the seminar room 1322 of the ITB, Invalidenstr. 43
Programming Course: Fridays from 5:00-7:00 pm in the Computer 'CIP' Pool (room 106/107), Invalidenstr. 42

Target Group: Students of Biology, Biophysics, Physics, Mathematics, and Computer Science.

Requirements: Basic knowledge in Neurobiology; ''Mathematics for Biologists'' I and II (III recommended), ''Data Analysis and Stochastic Processes'' recommended; basic knowledge in programming.

Lectures

With the advent of new functional brain imaging methods (EEG, MRI, optical methods) and multi-electrode devices it is now possible to simultaneously record from many different sites of living brains. Vast amounts of data can be collected, the analysis of which poses a challenge. It has thus become important to extract the essential information from the data to allow an easier interpretation of their properties, and to test relevant scientific hypotheses. The lecture series ''CNS IV'' will provide an overview over currently used statistical methods for the analysis of high-dimensional data. A framework for the analysis of neural systems will be given, using probability theory, estimation theory, information theory, and optimization. In the first part of the lecture series, we introduce Independent Component Analysis (ICA), which is a method for finding ''factors'' (or ''sources'' or ''components'') that underlie sets of random variables. ICA has also applications in denoising data, data compression and mining, feature extraction, and pattern recognition A short summary can be found in the article by Hyväarinen and Oja, 2000. The second half of the lectures will focus on the analysis of spike data.

Topics: analysis of high-dimensional data (neural network/EEG/MEG/imaging/audio), description of network states (e.g. the hippocampus), cocktail-party problem, principal and independent component analysis, probability theory and higher order statistics, information theory, estimation theory, optimization methods, curse of dimensionality, Bayesian modeling, Expectation Maximization (Baum-Welch) algorithm, time series prediction.

Guest Lectures:
24 May 2004, Dr. Laurenz Wiskott (HU Berlin), Slow Feature Analysis.
14 June 2004, PD Dr. Sonja Grün (FU Berlin), Unitary Event Analysis.

Background material:
L. Wiskott. Principal Component Analysis. ITB, 2004. (download manuscript)
A. Hyvärinen, J. Karhunen, and E. Oja. Independent Component Analysis. Wiley, New York, 2001. (table of contents, first chapter)
A. Hyvärinen, E. Oja. Independent component analysis: algorithms and applications. Neural Networks 13:411-430, 2000.
E. N. Brown, R. E. Kass, and P. P. Mitra. Multiple neural spike train data analysis: state-of-the-art and future challenges. Nature Neurosci. 7:456-461, 2004.
D.J.C. Mackay. Information Theory, Inference, and Learning Algorithms. Cambridge University Press, Cambridge, 2003. (download the book)
A. Webb. Statistical Pattern Recognition. Second Edition, Wiley, 2002.

Tutorials

The tutorials accompany the lectures. To obtain a Course Certificate (''Schein''), about 50% of the problems must be solved and handed in. Regular attendance of the weekly tutorials as well as occasional presentations of solutions during the tutorials are also required.

Exercises 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11

Programming Course

The programming course accompanies the lectures. In the first half of the semester, computer assignments will be distributed weekly, and the given problems can be solved in the Computer Pool. Help will be provided during the regular course hours by Tiziano Zito. At the beginning of the second half of the semester, a set of more advanced problems will be distributed. A group of two or three students can sign up for a specific project, and they should work together on a solution for several weeks. Help will again be provided during the regular course hours.

The web page on the Programming Course contains all the details on hardware, software, assignments, and solutions.

To obtain a Course Certificate (''Schein''), an oral presentation of the results of the project together with a one-page written summary of the results is required. The presentations will take place on Monday, July 12, 2004, instead of the lecture.

Last update: July-6-2004

Valid HTML 4.01!