Network Analysis Code Readme

Author: Yen-Nan Lin

Contents

About

The code was written and tested in MATLAB R2012a. Goal of this package was to analyse the difference between neural networks and theoretical networks in horizontal and vertical propagations. Because some functions used random number generator (most of the functions are module-related), the scripts might generate slightly differnet result in different computers. More details were shown in the manuscript.

This package include several toolboxes to analyze networks and plot figures. The toolboxes are under in the Toolbox folder.

Network analysis toolbox

Plotting toolboxes

Run

The codes could be classified into 2 groups: scripts and functions. All goXXXXXXX.m files should be scripts. Below is a process of standard analysis. The step of plotting are not included in the standard analysis process.

  1. load neural network
  2. analyse neural network
  3. create theoretical networks based on the loaded neural network
  4. analyse theoretical networks
  5. compute the average of realisation results
  6. plot the result

Here is a standard analysis example. In the matlab, and command

goAnalysis

Then, run specific plotting script based on which figure you want to plot.

If you want to plot result of figure 9, you should run another analysis script. However, this scripts may take several hours to compute.

goAnalysisRecur

Script of plotting figures

More information

The full documents of functions and scripts are shown in doc/index.html.

The dependencies of functions are shown in doc/NA/graph.html.