Show simple item record

dc.contributor.advisorBiros, George
dc.creatorZhou, Brady Beida
dc.date.accessioned2018-08-06T15:27:55Z
dc.date.available2018-08-06T15:27:55Z
dc.date.created2018-05
dc.date.submittedMay 2018
dc.identifierdoi:10.15781/T22N5021H
dc.identifier.urihttp://hdl.handle.net/2152/65949
dc.description.abstractIn this report, we present an efficient library for computing k-nearest neighbors (kNN) on datasets with sparse features (that is, most of the features per database entry are zero). Our work uses advances in parallel computing and optimized GPU routines. This GPU implementation utilizes highly parallel routines that exploit the sparsity property and in cases of extreme sparsity, we are able to achieve over 100x speedup in time compared to other state-of-the-art approaches designed for more general (dense) features.
dc.format.mimetypeapplication/pdf
dc.language.isoen
dc.subjectK-nearest neighbors
dc.subjectGPU
dc.subjectCUDA
dc.subjectSparse
dc.subjectSparse features
dc.subjectSparsity property
dc.subjectDatasets with sparse features
dc.subjectParallel computing
dc.subjectOptimized GPU routines
dc.titleGPU accelerated k-nearest neighbor kernel for sparse feature datasets
dc.typeThesis
dc.date.updated2018-08-06T15:27:55Z
dc.description.departmentComputational Science, Engineering, and Mathematics
thesis.degree.departmentComputational Science, Engineering, and Mathematics
thesis.degree.disciplineComputational Science, Engineering, and Mathematics
thesis.degree.grantorThe University of Texas at Austin
thesis.degree.levelMasters
thesis.degree.nameMaster of Science in Computational Science, Engineering, and Mathematics
dc.type.materialtext


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record