Thèse

Emergence of Oriented Circuits driven by Synaptic Pruning associated with Spike-Timing-Dependent Plasticity (STDP)

Emergence de circuits neuromimétiques orientés sous l'effet de l'épissage
associé à la plasticité synaptique à modulation temporelle relative (STDP)

by Javier Iglesias, MSc.

This page is the web appendix of my PhD thesis dissertation, submitted to the Universities of Lausanne in Switzerland and Grenoble 1 (Joseph Fourier) in France in the fall of year 2005.

I was co-directed by the Professor Marco Tomassini (Information Systems Institute) and the Professor Alessandro E.P. Villa (Grenoble - Institute of Neuroscience).

This joint transdisciplinar work led to the two titles of Doctorat ès Sciences of the University of Lausanne and Doctorat neurosciences-neurobiologie of the University Joseph Fourier.

Besides the printable documents

  1. Abstract in english ;
    36kb, PDF format

  2. Résumé en français ; and
    40kb, PDF format

  3. the PhD. dissertation document itself ;
    4.2Mb, PDF format

you will find here some material that could not fit into the printed medium of the PhD. dissertation, like: multimedia content, labile links to software source code and executables, sample files for the different file formats, Appendix C package to start you up, plus a place for the sequels and an erratum.

You might also want to check my resume

1. Multimedia content

Animations are provided in the GIF89a format that should be readable with any browser.

1.1. Stimuli

Here are a sample of the different stimuli used during the simulations. Intensity is color-coded with 'black'=0% and 'white'=100% maximum input. Time has been slowed down by a factor of 100 approximately.

  1. vertical bars stimuli
    100x100 units network, 10 bars, 10ms duration simple stimulus

  2. spatio-temporal pattern
    100x100 units network, 10x40 units groups, 10ms duration
    randomly chosen over 100% of the network surface: stimulus A and stimulus B

  3. spatio-temporal pattern
    100x100 units network, 10x40 units groups, 10ms duration
    randomly chosen over 25% of the network surface: stimulus C and stimulus D

2. Software

Many pieces of software have been developed to help answering the scientific questions. Some of them make sense outside of their original mission. They are available under the GNU Public License (GPL).

Most software pieces are being developped in a collaborative manner on sourceINforge.unil.ch. You should also check the multiple dependencies for these programs and the sample files.

Check the page dedicated to software.

2.1. Java applications

Java applications can be installed and launched directly from the browser, if a Java Virtual machine version 1.4+ is installed on your computer, thanks to the Java Webstart technology.

launch now! Data Toolbox
a pluggeable Java tool for simple multivariate time series data manipulation.
Launch now!
download package "Data_Toolbox"

launch now! RasterViewer
a raster-plot graphics viewer.
Launch now!

launch now! XY-Viewer
a generic scientific graphics viewer.
Launch now!
download package "XY-Viewer"

launch now! YaTiSeWoBe a full featured Java graphical framework for multivariate time series data manipulation.
Launch now!
download package "YaTiSeWoBe"

2.2. C programs

C programs must be downloaded, compiled and installed on your computer for them to run. This task is easier than you might think, provided you have a GNU compatible platform: GNU/Linux, Apple MacOS X, or Microsoft Windows with Cygwin installed.

  1. feign: a spiking neuron simulator
    download
    requires: XML library 2 (libxml2 and libxslt) and GNU Scientific Library (GSL)
    optional: MySQL library (mysqlclient)
  2. manip: a versatile time series data filter and editor
    download package "Analyses"
    requires: SQLite library (libsqlite3)

2.3. Other programs

  1. forge: a simulation organizer

3. Sample files

The dissertation mentions several file formats, some defined in the course of the work, some reused from external work. Sample files for those formats are presented here. They can be used with the softwares presented in the previous section.

  1. .fnet: feign networks
    This network state represents a 4x3 units network, connected as a complete synfire chain on length l=4 and width w=3.

  2. .xpdl: feign experimental protocols
    This protocol describes a simulation experiment that starts with an initialization of the network and a recovery pause, followed by a learning phase during which a specific stimulation is presented ten times in a row. The real experiment starts after the learning step. For the rest of the simulation duration, 5% of the stimulus presentations will be constituted of 3 variants of the learned stimulus in equivalent proportions. The other 95% of the times, the learned stimulus will be presented.

  3. .xyv: XY-Viewer native data format
    This document Shows two versions of the same Interspike Interval Histogram, the first is raw, while the second as been smoothed using a gaussian convolution of bin=5.

  4. .graphml graph representations
    This graph represents the same 4x3 units network, connected as a complete synfire chain on length l=4 and width w=3 of the .fnet example, but only the connections are represented, along with a custom extension for the strength (activation attribute of edge).

  5. .sdf time series
    The time series file format we are using is the Spike Data Format defined by Moshe Abeles. This data file can be opened with DataToolbox, RasterViewer, and YaTiSeWoBe.

  6. .sng time series
    This image is the ASCII representation of a gray scaled interrogation mark. This sample file could be used as a stimulus to a neural network simulated with feign.

4. Appendix C package

Appendix C of the dissertation leads you into running the software yourself. The following package contains configuration files and partial results to get you started.

5. Sequels

Check the publications

6. Erratum

Nothing, for now.