Design and Implementation of Parallel Batch-mode Neural Network on Parallel Virtual Machine
Adang Suwandi Ahmad, Arief Zulianto, Eto Sanjaya · 1999
Artificial Neural Network (ANN) computation process is a parallel computation process that should run using parallel processor. Budget constraint drives ANN implementation using sequential processor with sequential programming algorithm. Applying specific algorithm to existing sequential-computers offer emulation of ANN computation processes executes in parallel mode. This method can be applied by building a virtual parallel machine that featuring parallel-processing environment. This machine consist of many sequential machine operated in concurrent mode utilize operating system capability to manage inter-process communication and resource computation process, although this will increase complexity of the implementation of ANN learning algorithm process. This paper will describe the adaptation and development of sequential algorithm of feedforward learning into parallel programming algorithm on a virtual parallel machine based on PVM (Parallel Virtual Machine [5]) that was developed by Oak Ridge National Laboratory. PVM combines UNIX software calls to present a collection of high-level subroutines that allow the user to communicate between processes; synchronize processes; spawn, and kill processes on various machines using message passing construct. These routines are all combined in a user library, linked with user source code, before execution. Some modifications are made to adapt PVM into ITB network environment.