Exploiting immunological metaphors in the development of serial, parallel and distributed learning algorithms

This thesis examines the use of immunological metaphors in building serial,parallel, and distributed learning algorithms. It offers a basic study in thedevelopment of biologically-inspired algorithms which merge inspiration frombiology with known, standard computing technology to examine robust methodsof computing. This thesis begins by detailing key interactions found within theimmune system that provide inspiration for the development of a learning system.It then exploits the use of more processing power for the development of fasteralgorithms. This leads to the exploration of distributed computing resources forthe examination of more biologically plausible systems.This thesis offers the following main contributions. The components of theimmune system that exhibit the capacity for learning are detailed. A frameworkfor discussing learning algorithms is proposed. Three properties of every learningalgorithm—memory, adaptation, and decision-making—are identified for thisframework, and traditional learning algorithms are placed in the context ofthis framework. An investigation into the use of immunological componentsfor learning is provided. This leads to an understanding of these componentsin terms of the learning framework. A simplification of the Artificial ImmuneRecognition System (AIRS) immune-inspired learning algorithm is provided byemploying affinity-dependent somatic hypermutation. A parallel version of theClonal Selection Algorithm (CLONALG) immune learning algorithm is developed.It is shown that basic parallel computing techniques can provide computationalbenefits for this algorithm. Exploring this technology further, a parallel versionof AIRS is offered. It is shown that applying these same parallel computingtechniques to AIRS, while less scalable than when applied to CLONALG, stillprovides computational gains. A distributed approach to AIRS is offered, and itis argued that this approach provides a more biologically appealing model.Biological immune systems exhibit complex cellular interactions. Themechanisms of these interactions, while often poorly understood, hint at anextremely powerful information processing/problem solving system. This thesisdemonstrates how the use of immunological principles coupled with standardcomputing technology can lead to the development of robust, biologically-inspiredlearning algorithms.