Implementation of Conduction Delay and Collective Communication in a Parallel Spiking Neural Network Simulator.
Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Implementation of conduction delay and collective communication in a parallel spiking neural network simulator
As we know neural networks have a parallel structure and it is well suited for implementations in a parallel environment. The Bayesian Confidence Propagation Neural Network (BCPNN) which has been developed past thirty years is the main subject this thesis. An important issue is the implementation of communications between the processors. The aim of this thesis is to investigate point to point and collective communication methods and check how it works in real time. A second goal is to introduce time delay in point-to-point communication. These schemes have been implemented on Blue Gene Supercomputer using Message Passing Interface (MPI). At the end of thesis, the comparison between the two communication methods and the results of the two different models are shown.
Place, publisher, year, edition, pages
Trita-CSC-E, ISSN 1653-5715 ; 2011:134
IdentifiersURN: urn:nbn:se:kth:diva-130655OAI: oai:DiVA.org:kth-130655DiVA: diva2:654102
Master of Science - Computational and Systems Biology