Abstract-Single threaded applications cannot catch up with parallel systems when it comes to scalability mainly because of the clock frequency limitation of modern CPUs and economical reasons. Using shared memory or distributed memory architectures parallel systems can provide tremendous speed up compared to single threaded systems. CUDA, a shared memory parallel software, is considered to be a powerful language because of its easy thread management aspect and support for GPUs. Gauss blurring is a well-known image processing technique which reduces image noise and detail. Because of the high computation requirement of this technique, single threaded applications exhibit poor performance. In this paper we show that orders of magnitude speed up can be achieved by carrying out this operation on CUDA architecture with the help of high parallelism.
Localization is an essential task in Wireless Sensor Networks (WSN) for various use cases such as target tracking and object monitoring. Anchor nodes play a critical role in this task since they can find their location via GPS signals or manual setup mechanisms and help other nodes in the network determine their locations. Therefore, the optimal placement of anchor nodes in a WSN is of particular interest for reducing the energy consumption while yielding better accuracy at finding locations of the nodes. In this paper, we propose a novel approach for finding the optimal number of anchor nodes and an optimal placement strategy of them in a large-scale WSN, based on the output of Grey Wolf Optimization (GWO) and Particle Swarm Optimization (PSO) methods. As an initial step in this approach, the virtual localization process is executed over a virtual coordinate system in order to optimize the efficiency of the localization process. GWO and PSO methods are compared with a coverage-based analytical method and machine learning approaches such as Support Vector Machine (SVM) regression and Multiple Regression. The simulations we run with different numbers of nodes in a WSN and different communication ranges of nodes demonstrate that the proposed approaches are superior for minimizing the localization errors while reducing the number of anchor nodes.
Sentiment analysis techniques are widely used for extracting feelings of users in different domains such as social media content, surveys, and user reviews. This is mostly performed by using classical text classification techniques. One of the major challenges in this field is having a large and sparse feature space that stems from sparse representation of texts. The high dimensionality of the feature space creates a serious problem in terms of time and performance for sentiment analysis. This is particularly important when selected classifier requires intense calculations as in k-NN. To cope with this problem, we used sentiment analysis techniques for Turkish Twitter feeds using the NVIDIA's CUDA technology. We employed our CUDA-based distance kernel implementation for k-NN which is a widely used lazy classifier in this field. We conducted our experiments on four machines with different computing capacities in terms of GPU and CPU configuration to analyze the impact on speed-up.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.