Parallel clustering system using the methodologies of evolutionary computations
Reginald L. Walker · 2002
Several versions of the parallel clustering system were studied to improve performance of its initial implementation. The current versions were restricted to 1024 Web pages which, in turn, were used to create adaptive probe sets that were distributed to each indexer node. The probe sets were used to compute fitness measures associated with each indexer node used to create sub-species for the purpose of applying the new and traditional GA/GP operators. Speedup resulted from fitness-enhancing mechanisms that provided information results from previous fitness measurements of previous generations, such as the non-genetic transmission of cultural information. The clustering results are being used in the Tocorime Apicu project to develop a bioinformatic approach to the design and validation of an integrated, experimental search engine. This model provides a foundation for an evolutionary expansion of this computational model as World Wide Web (WWW) documents continue to grow. The clustering results were generated using message passing interface (MPI) on a network of SUN workstations.