Accelerating the XGBoost algorithm using GPU computing

Simple item page
dc.contributor.authorMitchell, Roryen_NZ
dc.contributor.authorFrank, Eibeen_NZ
dc.date.accessioned2017-07-26T02:24:26Z
dc.date.available2017en_NZ
dc.date.available2017-07-26T02:24:26Z
dc.date.issued2017en_NZ
dc.description.abstractWe present a CUDA-based implementation of a decision tree construction algorithm within the gradient boosting library XGBoost. The tree construction algorithm is executed entirely on the graphics processing unit (GPU) and shows high performance with a variety of datasets and settings, including sparse input matrices. Individual boosting iterations are parallelised, combining two approaches. An interleaved approach is used for shallow trees, switching to a more conventional radix sort-based approach for larger depths. We show speedups of between 3x and 6x using a Titan X compared to a 4 core i7 CPU, and 1.2x using a Titan X compared to 2x Xeon CPUs (24 cores). We show that it is possible to process the Higgs dataset (10 million instances, 28 features) entirely within GPU memory. The algorithm is made available as a plug-in within the XGBoost library and fully supports all XGBoost features including classification, regression and ranking tasks.
dc.format.mimetypeapplication/pdf
dc.identifier.citationMitchell, R., & Frank, E. (2017). Accelerating the XGBoost algorithm using GPU computing. PeerJ Computer Science, 3. https://doi.org/10.7717/peerj-cs.127en
dc.identifier.doi10.7717/peerj-cs.127en_NZ
dc.identifier.urihttps://hdl.handle.net/10289/11227
dc.language.isoen
dc.relation.isPartOfPeerJ Computer Scienceen_NZ
dc.rightsThis article is published under Creative Commons CC-BY 4.0.
dc.subjectcomputer scienceen_NZ
dc.subjectsupervised machine learningen_NZ
dc.subjectgradient boostingen_NZ
dc.subjectGPU computingen_NZ
dc.subjectMachine learning
dc.titleAccelerating the XGBoost algorithm using GPU computingen_NZ
dc.typeJournal Article
pubs.elements-id200260
pubs.notesQA:10.7717/peerj-cs.127en_NZ
pubs.organisational-group/Waikato
pubs.organisational-group/Waikato/2018 PBRF
pubs.organisational-group/Waikato/FCMS
pubs.organisational-group/Waikato/FCMS/2018 PBRF - FCMS
pubs.organisational-group/Waikato/FCMS/Computer Science
pubs.user.infoFrank, Eibe (eibe@waikato.ac.nz)
pubs.volume3en_NZ
uow.identifier.article-noe127
uow.verification.statusverified
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
peerj-cs-127.pdf
Size:
3.1 MB
Format:
Adobe Portable Document Format
Description:
Published version
Thumbnail Image
Name: peerj-cs-127.pdf
Size: 3.1 MB
Kind: Adobe PDF
License bundle
Now showing 1 - 1 of 1
No Thumbnail Available
Name:
Research Commons Deposit Agreement 2017.pdf
Size:
188.11 KB
Format:
Adobe Portable Document Format
Description:
Name: Research Commons Deposit Agreement 2017.pdf
Size: 188.11 KB
Kind: Adobe PDF
Collections
Computing and Mathematical Sciences Papers