Memory-efficient implementation of a graphics processor-based cluster detection algorithm for large spatial databases

Thapa, Rajeev J.; Trefftz, Christian; Wolffe, Greg

doi:10.1109/eit.2010.5612134

Cited by 13 publications

(7 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It should be noted that unlike the datasets tested previously, these are synthetic datasets and have constant spatial density throughout, with the points ordered by cluster. Table IV shows the execution times of the various implementations with the parameters set to Eps = 1.5 and MinP ts = 4 as used in [Thapa et al 2010]. The number of PEs was set to 50 as this extracts the maximum amount of parallelism available in the data and allows for a clock speed of 393 MHz.…”

Section: Resultsmentioning

confidence: 99%

“…Additionally, as occurs with the standard DBSCAN implementations, with larger Eps the algorithm takes longer to compute. Furthermore, the proposed system was tested using the datasets that were used in [Thapa et al 2010], which the authors have kindly provided. It should be noted that unlike the datasets tested previously, these are synthetic datasets and have constant spatial density throughout, with the points ordered by cluster.…”

Section: Resultsmentioning

confidence: 99%

“…To analyse the performance of the design when clustering datasets with an increasing number of dimensions, synthetic datasets similar to the ones used in [Thapa et al 2010] were generated. These were generated such that they have constant spatial density throughout and organised to form eight clusters with the same Eps and MinP ts parameters.…”

Section: Resultsmentioning

confidence: 99%

“…Thapa et al [Thapa et al 2010] propose a Graphics Processing Unit (GPU) implementation of the DBSCAN algorithm that takes advantage of the large amounts of memory and processor cores available on modern GPUs. Two different approaches are explored in attempt to accelerate this algorithm through parallelism.…”

Section: Related Workmentioning

confidence: 99%

“…Clustering has many applications including data mining, statistical data analysis, pattern recognition and image analysis [Martin Ester et al 1996;Daszykowski et al 2001;Thapa et al 2010]. Various clustering algorithms have been developed so far, usually targeting a specific domain of applications by defining the notion of a cluster accordingly.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Arc 2014

Scicluna

Bouganis

2015

ACM Trans. Reconfigurable Technol. Syst.

View full text Add to dashboard Cite

Clustering large numbers of data points is a very computationally demanding task that often needs to be accelerated in order to be useful in practical applications. This work focuses on the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, which is one of the state-of-the-art clustering algorithms, targeting its acceleration using an FPGA device. The paper presents an, optimised, scalable and parameterisable architecture that takes advantage of the internal memory structure of modern FPGAs in order to deliver a high performance clustering system. Post-synthesis simulation results show that the developed system can obtain mean speed-ups of 31x in real-world tests and 202x in synthetic tests when compared to state-of-the-art software counterparts running on a quad-core 3.4 GHz Intel i7-2600k. Additionally, this implementation is also capable of clustering data with any number of dimensions without impacting the performance.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations