Logo

CIRG - Research  -  Artificial Immune Systems 



[Bullet] Home
About
News
[Bullet]

- NN
- DM
- SI
- EC
- MAS
- AIS
- IA
- Bioinf
- Games
- Opt
- FA
- Industry

Publications
People
Resources
Links
Contact Us

OVERVIEW

This research focus area develops new training methods for artificial immune systems. Genetic algorithms and particle swarm optimization techniques are used to optimize the creation of artificial lymphocytes. Approaches are also investigated to optimize the training process of existing learning algorithms. Network-based immune algorithms to cluster of non-stationary data are developed.

ACTIVE MEMBERS

List the current members actively doing research in this focus area. [ Show ]

ALUMNI MEMBERS

M Poggiolini

M.Sc Completed in 2009

A Graaff

PhD Started in 2005
M.Sc Completed in 2003

GROUP PUBLICATIONS

List publications of this research focus area. [ Show ]

MEMBER PROFILE



 Name:

 Mario Poggiolini

Portrait photo

 E-mail:

 mario.poggiolini@accenture.com

 Group(s):

 Artificial Immune Systems

 

 Degree specific information: M.Sc

 Title:

 Generalization Characteristics of Artificial Immune Systems

 Abstract:

The terms generalization and overfitting are well researched within the neural network learning paradigm. As such neural network writers can employ a wide variety of tools to ensure that their networks maximize generalization whilst minimizing overfitting. These terms have never fully been explored within the traditional artificial immune system learning paradigm. This study intends to address this by providing a clear unambiguous definition of generalization and overfitting in terms of artificial immune systems and the shape space model introduced by Perelson. Using these definitions a metric is suggested to measure the performance of a population of detectors. Various detector generating algorithms and matching rules are scrutinized in terms of their generalization and overfitting abilities and number of new algorithms and improvements are suggested to overcome the constraints imposed by the old algorithms and matching rules.

 Supervisor / Co-Supervisor:

 AP Engelbrecht

 Thesis:

 Not available for download yet.




You are visitor #21746
Contact webmaster
Back to top

QualNet Network Simulator University Program Valid XHTML 1.0! Valid CSS!


Computational Intelligence Research Group
University of Pretoria
Copyright © 2017