Parallelized neural networks as a service
Altaf Ahmad Huqqani, Erich Schikuta, Erwin Mann · 2014
We present a novel neural network simulation framework, which provides the parallelized execution of artificial neural network by exploiting modern hardware and software environments adhering to the service oriented paradigm within the N2Sky system. The goal of the N2Sky system is to share and exchange neural network resources, neural network specific knowledge, neural network objects and paradigms and, in turn, to deliver a transparent environment to novice and experienced users to do neural network research. The parallelization techniques are deployed transparently for the user reducing significantly the time consuming training of neural networks by selecting appropriate service implementation according to complexity of the problem. N2Sky follows the sky computing paradigm fostering ample resources by using federated clouds.