Grid e-Services for Multi-Layer SOM Neural Network Simulation

In this paper it is presented a framework environment for neural network simulation through the development of e-Services running on the EGEE Grid infrastructure. Parallel computing and distribution of jobs can be synchronized by the e-Service through a web interface. This interface provides an easy to use environment and can be requested over the internet. This simulation tool is based on a neural network agent approach that implements the e-Service using software agents and services offered by the Grid middleware. As an example, it is hereby presented an e-Service based on an implementation of the Self-Organizing Map (SOM) neural network algorithm which indicates that significant improvements in performance can be gained. In particular, users of SOM neural networks can take advantage of these widely available computer networks to obtain the computing power that they require.