An Idiotypic Solution Sieve for Selecting the Best Performing Solutions in Real-World Distributed Intelligence

Jerne's Idiotypic Network theory features autonomous network formation, adaptation, learning and self-stabilization, all of which find extensive applications in computational realm. Researchers have used this model in a myriad of applications, however, the use of this model in real networked environments has hardly been addressed. This paper describes an Idiotypic Sieve to filter out the optimal solutions from a set of available solutions for a set of heterogeneous problems that could occur asynchronously or concurrently across a real network. The Idiotypic Sieve described herein, is conceived by emulating an Idiotypic network wherein antibodies (solutions) within a real physical network asynchronously interact with one another and also with the antigens (problems) in a distributed and decentralized manner and stimulate and suppress one another consequently changing their respective global populations across the network. The antibodies (solutions) are provided the much required mobility across the network by a set of mobile agents that autonomously patrol and migrate to nodes that are invaded by the antigens (problems). Emulation results carried out on a real network portrayed in this paper, show the effectiveness of the Idiotypic Sieve in generating and controlling the populations of both optimal and generic solutions to the heterogeneous set of problems.