1 January 2013 in Procedia Computer Science
Training of Artificial Neural Networks for large data sets is a time consuming task. Various approaches have been proposed to reduce the efforts, many of them by applying parallelization techniques. In this paper we develop and analyze two novel parallel training approaches for Backpropagation neural networks for face recognition. We focus on two specific paralleliza- tion environments, using on the one hand OpenMP on a conventional multithreaded CPU and CUDA on a GPU. Based on our findings we give guidelines for the efficient parallelization of Backpropagation neural networks on multicore and GPU architectures.Additionally, we present a traversal method finding the best combination of learning rate and momentum term by varying the number of hidden neurons supporting the parallelization efforts
1 October 2016 in Computer Networks
1 January 2011 in Procedia Computer Science
In distributed systems, where several deployments of a specific service exist, it is a crucial task to select and combine concrete deployments to build an executable workflow. Non-functional properties such as performance and availability are taken into account in such selection processes that are designed to reach certain objectives while meeting constraints. In this paper, a concrete data-intensive application scenario from a High-Energy Physics experiment comprising a deployment selection challenge is introduced. A generic model for distributed systems is presented based on which a formal model representing the individual components of the system is derived. The optimization problem is approached both from the angle of the user and the angle of the system provider. Moreover the dynamic aspects of the underlying system are taken into account. This results in a dynamic multi-objective optimization problem for which an explicit memory-based genetic algorithm is proposed
20 August 2008 in Concurrency and Computation: Practice and Experience