MITHaystack / VariantDBSCAN

Variant Density-Based Spatial Clustering of Applications with Noise (VariantDBSCAN)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

VariantDBSCAN

Maximizing Clustering Throughput

Project lead: Mike Gowanlock

Relevant papers:

  • [1] Gowanlock, M., Blair, D. M. & Pankratius, V. (2016) Exploiting Variant-Based Parallelism for Data Mining of Space Weather Phenomena. In Proc. of the 30th IEEE International Parallel & Distributed Processing Symposium (IPDPS 2016). pp. 760-769 DOI: 10.1109/IPDPS.2016.10 [http://dx.doi.org/10.1109/IPDPS.2016.10 ]

  • [2] Gowanlock, M., Blair, D. M., Pankratius, V. Optimizing Parallel Clustering Throughput in Shared Memory. IEEE Transactions on Parallel and Distributed Systems DOI: 10.1109/TPDS.2017.2675421 [http://dx.doi.org/10.1109/TPDS.2017.2675421]

alt text

Figure: Relative performance gains utilizing all of the optimizations over the sequential implementation on a space weather TEC dataset in [1, 2]. Values over the black line indicate a performance improvement. The red line indicates the performance gain from index optimizations only. See the papers above.

We acknowledge support from NSF ACI-1442997 and NASA AIST14-NNX15AG84G.

About

Variant Density-Based Spatial Clustering of Applications with Noise (VariantDBSCAN)


Languages

Language:Cuda 60.6%Language:C++ 33.6%Language:Python 4.2%Language:C 0.8%Language:Makefile 0.7%Language:Shell 0.0%