Paper Info
Title
The Influence of Age Assignments on the Performance of Immune Algorithms.
Abstract
How long a B cell remains, evolves and matures inside a population plays a crucial role on the capability for an immune algorithm to jump out from local optima, and find the global optimum. Assigning the right age to each clone (or offspring, in general) means to find the proper balancing between the exploration and exploitation. In this research work we present an experimental study conducted on an immune algorithm, based on the clonal selection principle, and performed on eleven different age assignments, with the main aim to verify if at least one, or two, of the top 4 in the previous efficiency ranking produced on the one-max problem, still appear among the top 4 in the new efficiency ranking obtained on a different complex problem. Thus, the NK landscape model has been considered as the test problem, which is a mathematical model formulated for the study of tunably rugged fitness landscape. From the many experiments performed is possible to assert that in the elitism variant of the immune algorithm, two of the best age assignments previously discovered, still continue to appear among the top 3 of the new rankings produced; whilst they become three in the no elitism version. Further, in the first variant none of the 4 top previous ones ranks ever in the first position, unlike on the no elitism variant, where the previous best one continues to appear in 1st position more than the others. Finally, this study confirms that the idea to assign the same age of the parent to the cloned B cell is not a good strategy since it continues to be as the worst also in the new efficiency ranking.
Year
DOI
Venue
2018
10.1007/978-3-319-97982-3_2
ADVANCES IN COMPUTATIONAL INTELLIGENCE SYSTEMS (UKCI)
Field
DocType
Volume
Population,Fitness landscape,Ranking,Local optimum,Computer science,Algorithm,Global optimum,Clonal selection,Elitism
Conference
840
ISSN
Citations 
PageRank 
2194-5357
0
0.34
References 
Authors
6
4
Name
Order
Citations
PageRank
Alessandro Vitale100.34
Antonino Di Stefano200.34
vincenzo cutello355357.63
Mario Pavone421219.41