README_gif.txt
J Pillow.
Oct 30, 2007 (version 1).

DESCRIPTION OF THE CODE CONTAINED IN THE ARCHIVE: code_gif_v1.tgz

Performs simulation and max-likelihood estimation of generalized
integrate-and-fire model.

INSTALLATION
============

(1) Unnpack the archive using a compression utility such as WinZip
    (or, from the command line in linux: 'tar -xvzf FILENAME.tgz').

(2) Launch matlab and cd into the directory containing the code
    (e.g. '/code_gif/').

(3) Run the initialization script 'initialize_mexcode' from the
    command line (this will compile the C files in the sub-directory
    'mexcode' into mex using the default compiler settings).  compiles the
    C files into mex).


USE
===

Open the script 'testGIFcode.m'.  This script contains code for
generating the four figures shown in DemonstrationFigs.pdf (which
contain additional captions).  

The script is divided into four blocks that illustrate:
(a) Bases used to define filters; A default setting of model params
(b) simulation of the model using a short stimulus.
(c) graphical illustration of likelihood calculation for a single ISI,
    performed using Density Propogation (Fokker-Planck), and validated
    with Monte-Carlo sampling.
(d) maximum likelihood estimation of GIF parameters from simulated
    data.

Note: MLgif.m, the main function for performing ML estimation,
requires the matlab optimization toolbox, since it makes a call to
"fmincon".  If you don't have this, the code can be altered to use
"fminsearch", though this will run more slowly.