MANUAL.HTML

Copyright (c) by Leandro Nunes de Castro.

This manual describes how to run the Matlab® tutorial presentation on Artificial Immune Systems developed by Leandro N. de Castro and Fernando J. Von Zuben. The program files can be downloaded from the following FTP address: ftp://ftp.dca.fee.unicamp.br/pub/docs/vonzuben/lnunes/demo.zip

The tour is self-guided and can be performed in any order.

To run the presentation, first uncompress the zipped archive and store it in an appropriate directory. Run the Matlab®, enter the selected directory, and type “tutorial” in the Matlab prompt.

>> tutorial
The following information will appear:
*-------------------------------------------------*
ARTIFICIAL IMMUNE SYSTEMS - TUTORIAL PRESENTATION
      Leandro Nunes de Castro
          April/May, 2000
*-------------------------------------------------*

Available Demos:
1) SAND
2) ABNET
3) CLONALG
4) AINET
Type the desired demo number (or CTRL^C to Interrupt):

The demos describe a “Simulated Annealing Model to Increase Diversity in the Antibody Repertoire”, “An Antibody Network”, “The Clonal Selection Algorithm”, and “An Artificial Immune Network Model”, respectively. The first three algorithms are fully described in the Technical Report RT-DCA 01/99 and respective publications, and the last algorithm is described only in the papers. The files can be obtained from this CD or be downloaded from the author’s homepage: https://www.dca.fee.unicamp.br/~lnunes/immune.html.

Before running the demos, we strongly encourage the users to read through the tech report RT-DCA 01/99 and respective papers, in order to achieve a better comprehension of the tour we are about to begin.

1. If you choose demo number (1), by typing 1 in the command line:

Type the desired demo number (or CTRL^C to Interrupt): 1
The following will appear:
** PART I - SAND (Simulated ANnealing Approach to Increase Diversity) **
Available Demo Tasks:
1) REAL-VALUED VECTORS
2) 3-D BINARY POINTS
Type the desired demo number (1) or (2):
These two options illustrate the SAND performance in maximizing the diversity of a real-valued 2-D problem and a binary set of vectors (L = 3), respectively.

2. By choosing demo number 2,

Type the desired demo number (or CTRL^C to Interrupt): 2
The following information will appear:
** PART II - ABNET (An AntiBody NETwork) **

Antigens (columnwise) to be Recognized: [10,10]
     1     0     0     1     1     0     1     1     0     0
     1     1     0     1     1     0     0     1     0     0
     1     1     1     1     1     0     0     0     0     0
     1     1     1     1     1     0     0     0     0     0
     1     1     1     1     1     0     0     0     0     0
     1     1     1     1     1     0     0     0     0     0
     1     1     1     1     1     0     0     0     0     0
     1     1     1     1     1     0     0     0     0     0
     1     1     1     1     0     0     0     0     0     1
     1     1     1     0     0     0     0     0     1     1

Choose the Affinity Threshold [0,L]:

The affinity threshold (epslon) might be chosen within the [0-10] range.

3. Demo number (3),

Type the desired demo number (or CTRL^C to Interrupt): 3
Three other demos arise:
** PART III - CLONALG (The CLONal Selection ALGorithm) **
Available Demo Tasks:
1) GA (GLOBAL SEARCH)
2) CLONALG (MULTI-MODAL OPTIMIZATION)
3) CHARACTER (KNOWLEDGE ACQUISITION)
Type the desired demo number (1), (2) or (3):
The GA is a standard Genetic Algorithm with bi-classist selection. CLONALG is the Clonal Selection Algorithm, and CHARACTER is the application of the CLONALG for a pattern recognition problem. The function to be maximized contains four global maxima, and options (1) and (2) aim at comparing the performance of a standard genetic algorithm and the proposed immune algorithm (CLONALG).

4. Demo number (4),

Type the desired demo number (or CTRL^C to Interrupt): 4
Three other demos arise:
** PART IV - AINET (Artificial Immune NETwork) **
Available Demo Tasks:
1) LINEARLY SEPARABLE CLASSES
2) 2-SPIRALS
3) CHAINLINK
Type the desired demo number (1), (2) or (3):
The suppression threshold is the only parameter whose value was not assigned by default, though suggested values were given.

IMPORTANT NOTE: The Matlab® programming language does not generate EXE files, thus the source code of all algorithms are available for analysis and copy, but the authors and this site MUST be quoted in every use.