Target tracking in distributed networks faces the challenge in coping with large volumes of distributed data which requires efficient methods for real time applications with minimal communication overhead. the complex...
详细信息
Target tracking in distributed networks faces the challenge in coping with large volumes of distributed data which requires efficient methods for real time applications with minimal communication overhead. the complexity considered in this paper is when each sensor in a distributed network observes a large number of measurements which are all required to be processed at each time step. the particle filter has been widely used for localisation and tracking in distributed networks with a small number of measurements [1]. this paper goes beyond the current state-of-the-art and presents a novel particle filter approach, combined withthe expectation propagation framework, that is capable of dealing withthe challenges presented by a large volume of measurements in a distributed network. In the proposed algorithm, the measurements are processed in parallel at each sensor node in the network and the communication overhead is minimised substantially. We show results with large improvements in communication overhead, with a negligible loss in tracking performance, compared withthe standard centralised particle filter.
Analytical queries are crucial for many emerging Semantic Web applications such as clinical-trial recruiting in Life Sciences that incorporate patient and drug profile data. Such queries compare aggregates over multip...
详细信息
We introduce the method of the Maximum Entropy (MaxEnt) model for fusing local decisions in a distributed multiple sensor system. the fusion center receives local binary decisions in the usual parallel architecture. N...
详细信息
We introduce the method of the Maximum Entropy (MaxEnt) model for fusing local decisions in a distributed multiple sensor system. the fusion center receives local binary decisions in the usual parallel architecture. No assumptions are made about knowing any local decision rules. Our approach is based on the concept of machine learning, wherein the MaxEnt parametric model is used for supervised classification and prediction serving as the central (global) decision rule. therefore, the system is able to learn the detection performance of the sensors as a function of time without prior knowledge of the actual probabilities of local decisions, only requiring an initial set of random training data. thus it is demonstrated that the system is adaptive and can learn contextual changes of the sensors. Furthermore, we provide simulation results comparing the MaxEnt fusion center performance with published results using boththe Bayesian formulation and Neyman-Pearson criterion and with MaxEnt achieving the best, realistic detection performance demonstrating the effectiveness of the method.
Tascell is a task parallel language that supports distributed memory environments. the conventional implementation of Tascell realizes inter-node communication with TCP/IP communication via Tascell servers. this imple...
详细信息
ISBN:
(纸本)9781509028269
Tascell is a task parallel language that supports distributed memory environments. the conventional implementation of Tascell realizes inter-node communication with TCP/IP communication via Tascell servers. this implementation is suitable for dynamic addition of computation nodes and wide-area distributed environments. On the other hand, in supercomputer environments, TCP/IP may not be available for inter-node communication and there may be no appropriate places for deploying Tascell servers. In this study, we have developed a server-less implementation of Tascell that realizes inter-node communication with MPI communication in order to evaluate its performance on massively parallelsystems. It performs well on four Xeon Phi coprocessors (with 456 workers) and the K computer, for instance, our 19-queens solver achieves a 4615-fold speedup relative to a serial implementation with 7168 workers on the K computer. Our server-less implementation realizes deadlock freedom, although it only requires the two-sided communication paradigm and the MPI_thREAD_FUNNELED support level. On Xeon Phi coprocessors, we compare our implementation with other implementations that employ TCP/IP or the MPI_thREAD_MULTIPLE support level.
Efficient resource management is an important requirement for many process-oriented applications. Typically, work items are assigned to resources through their work lists. there are many reasons for reordering work it...
详细信息
ISBN:
(纸本)9781467393317
Efficient resource management is an important requirement for many process-oriented applications. Typically, work items are assigned to resources through their work lists. there are many reasons for reordering work items in a resource's work list. For process scheduling, for example, swapping process instances constitutes a mean to keep due times. At the same time, reducing the throughput time of the global process is typically not the primary goal. For process optimization, in turn, the implications of reordering work items on the overall temporal performance of the process might be crucial. In this paper, we investigate how reordering work items affects performance parameters that are typically associated with a first-in-first-out processing mechanism at resources. the analysis is conducted for single process tasks and for typical control flow patterns such as sequence as well as parallel and alternative branchings. It is shown that the implications on the global throughput time are less than expected, while the effects on instance-based parameters strongly depend on the control-flow pattern in which the reordering mechanism is implemented. the results are supported by means of a simulation.
Airport drop-off service provided by airlines is a chauffeur-driven service(i.e. Uber and Di Di) as an emerging travel choice for travelers. More and more passenger enjoy the drop-off service. In practice, we find an ...
详细信息
ISBN:
(纸本)9781509012572
Airport drop-off service provided by airlines is a chauffeur-driven service(i.e. Uber and Di Di) as an emerging travel choice for travelers. More and more passenger enjoy the drop-off service. In practice, we find an interesting question: if a passenger has ever choice the drop-off service, whether they are willing to recommend this service to other traveler? Although the acknowledgment that social learning is related to travel decision is promoted, quantitative analysis about how social learning shape and impact the decision of passengers is still limited. We study and estimate a diffusion probability between different passengers by proposing a CCM(Co-travel Link Cascade Model) based on a modified EM iterative algorithm. then, we segment passengers into three types(Influenced, Unchecked and Immune). the three types of passengers are predicated by approaches of IC-like model, Random Forest model and probabilistic model, respectively. In addition, we also design a parallel implementation of our proposed algorithm in the Apache Spark distributed data processing environment. Experimental results on a real aviation data set demonstrate that CCM can efficiently infer the decision of travelers.
Constraint Satisfaction Problems (CSPs) occur now in different domains. Several methods are used to solve them. In particular, Particle Swarm Optimization (PSO) allows to solve efficiently CSPs by significantly reduci...
详细信息
Constraint Satisfaction Problems (CSPs) occur now in different domains. Several methods are used to solve them. In particular, Particle Swarm Optimization (PSO) allows to solve efficiently CSPs by significantly reducing the calculation time to explore the search space of solutions. However, this metaheuristic is excessively costing when facing large instances. In this paper we address the Maximal Constraint Satisfaction Problems (Max-CSPs). We introduce a new resolution approach that allows solving efficiently the Max-CSPs even with large instances. Our purpose is to implement a PSO based method by using the GPU architecture as a parallelcomputing framework. In particular, we focus on the implementation of two parallel novel approaches. the first one is a parallel GPU-PSO for Max-CSPs (GPU-PSO) and the second one is a GPU distributed PSO for Max-CSPs (GPU-DPSO). Our experimental results show the efficiency of the two proposed approaches and their ability to exploit GPU architecture. (C) 2015 the Authors. Published by Elsevier B.V.
the near-Earth space surveillance requires accurate and fast orbit predictions for the multitude of objects that are present around our planet. this prediction is performed by means of orbit propagators which offer so...
详细信息
ISBN:
(纸本)9781479984817
the near-Earth space surveillance requires accurate and fast orbit predictions for the multitude of objects that are present around our planet. this prediction is performed by means of orbit propagators which offer solutions to the satellite equations of motion. Based on the method of computation those propagators are split into three categories: analytical, numerical and semi-analytical, each with its own advantages and disadvantages. Today a large number of propagators are implemented in the form of software libraries, some proprietary and some freeware or even open-source. In this paper we have taken a semi-analytical propagator from an open-source Java library and used it in a distributed environment in order to compute the position of a satellite during a large time interval.
the reproducibility of an in-silico experiment is a great challenge because of the parallel and distributed environment and the complexity of the scientific workflows. In order to solve such problems on one hand prove...
详细信息
ISBN:
(纸本)9781467379397
the reproducibility of an in-silico experiment is a great challenge because of the parallel and distributed environment and the complexity of the scientific workflows. In order to solve such problems on one hand provenance data has to be captured about the dataflow, the ancestry of the results and the environment of the execution, on the other hand description data has to be collected from the scientist and stored about the essential details, the types and samples of input/output data, and the operation of the experiment. the ultimate goal of our work is to propose a minimal dataset for recording and reporting scientific workflow based experiment, which will facilitate the reproducibility of such experiments, the public repositories and enable to share and reuse the scientific results. One part of the dataset can be filled in manually by the scientist, certain part can be filled in automatically by the system and other part can be filled in from provenance data.
the actual trend in automotive is to reduce as much as possible pollution caused by motor vehicle emissions. this could be achieved by using electrical propulsion systems which usually contains: electrical machine, th...
详细信息
ISBN:
(纸本)9781479984817
the actual trend in automotive is to reduce as much as possible pollution caused by motor vehicle emissions. this could be achieved by using electrical propulsion systems which usually contains: electrical machine, three phase voltage source inverter (VSI), voltage supply system. the main problem of the full electric vehicles represents the autonomy due to low energy capacity of batteries. this issue could be solved by using a hydrogen fuel cell in parallel with a battery pack to extend the range of the electrical vehicles. this paper presents a voltage supply system which connects to a DC bus a battery, a fuel cell, and a supercapacitor using DCDC converters. the focus of this work is on keeping controlling the voltage level of DC bus and on energy distribution between elements of voltage supply system. the proposed algorithm is validated using a close loop model which contains simulation of: fuel cell, battery pack, supercapacitor, DCDC converter, distributed control strategy of voltage supply system, DC Link capacitor, VSI, electrical machine, vehicle dynamics, and driver. the model is developed in Matlab/Simulink.
暂无评论