Buffer k-d Trees: Processing Massive Nearest Neighbor Queries on GPUs

Fabian Gieseke, Justin Heinermann, Cosmin Oancea, Christian Igel
; Proceedings of the 31st International Conference on Machine Learning, PMLR 32(1):172-180, 2014.

Abstract

We present a new approach for combining k-d trees and graphics processing units for nearest neighbor search. It is well known that a direct combination of these tools leads to a non-satisfying performance due to conditional computations and suboptimal memory accesses. To alleviate these problems, we propose a variant of the classical k-d tree data structure, called buffer k-d tree, which can be used to reorganize the search. Our experiments show that we can take advantage of both the hierarchical subdivision induced by k-d trees and the huge computational resources provided by today’s many-core devices. We demonstrate the potential of our approach in astronomy, where hundreds of million nearest neighbor queries have to be processed.

Cite this Paper


BibTeX
@InProceedings{pmlr-v32-gieseke14, title = {Buffer k-d Trees: Processing Massive Nearest Neighbor Queries on GPUs}, author = {Fabian Gieseke and Justin Heinermann and Cosmin Oancea and Christian Igel}, booktitle = {Proceedings of the 31st International Conference on Machine Learning}, pages = {172--180}, year = {2014}, editor = {Eric P. Xing and Tony Jebara}, volume = {32}, number = {1}, series = {Proceedings of Machine Learning Research}, address = {Bejing, China}, month = {22--24 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v32/gieseke14.pdf}, url = {http://proceedings.mlr.press/v32/gieseke14.html}, abstract = {We present a new approach for combining k-d trees and graphics processing units for nearest neighbor search. It is well known that a direct combination of these tools leads to a non-satisfying performance due to conditional computations and suboptimal memory accesses. To alleviate these problems, we propose a variant of the classical k-d tree data structure, called buffer k-d tree, which can be used to reorganize the search. Our experiments show that we can take advantage of both the hierarchical subdivision induced by k-d trees and the huge computational resources provided by today’s many-core devices. We demonstrate the potential of our approach in astronomy, where hundreds of million nearest neighbor queries have to be processed.} }
Endnote
%0 Conference Paper %T Buffer k-d Trees: Processing Massive Nearest Neighbor Queries on GPUs %A Fabian Gieseke %A Justin Heinermann %A Cosmin Oancea %A Christian Igel %B Proceedings of the 31st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2014 %E Eric P. Xing %E Tony Jebara %F pmlr-v32-gieseke14 %I PMLR %J Proceedings of Machine Learning Research %P 172--180 %U http://proceedings.mlr.press %V 32 %N 1 %W PMLR %X We present a new approach for combining k-d trees and graphics processing units for nearest neighbor search. It is well known that a direct combination of these tools leads to a non-satisfying performance due to conditional computations and suboptimal memory accesses. To alleviate these problems, we propose a variant of the classical k-d tree data structure, called buffer k-d tree, which can be used to reorganize the search. Our experiments show that we can take advantage of both the hierarchical subdivision induced by k-d trees and the huge computational resources provided by today’s many-core devices. We demonstrate the potential of our approach in astronomy, where hundreds of million nearest neighbor queries have to be processed.
RIS
TY - CPAPER TI - Buffer k-d Trees: Processing Massive Nearest Neighbor Queries on GPUs AU - Fabian Gieseke AU - Justin Heinermann AU - Cosmin Oancea AU - Christian Igel BT - Proceedings of the 31st International Conference on Machine Learning PY - 2014/01/27 DA - 2014/01/27 ED - Eric P. Xing ED - Tony Jebara ID - pmlr-v32-gieseke14 PB - PMLR SP - 172 DP - PMLR EP - 180 L1 - http://proceedings.mlr.press/v32/gieseke14.pdf UR - http://proceedings.mlr.press/v32/gieseke14.html AB - We present a new approach for combining k-d trees and graphics processing units for nearest neighbor search. It is well known that a direct combination of these tools leads to a non-satisfying performance due to conditional computations and suboptimal memory accesses. To alleviate these problems, we propose a variant of the classical k-d tree data structure, called buffer k-d tree, which can be used to reorganize the search. Our experiments show that we can take advantage of both the hierarchical subdivision induced by k-d trees and the huge computational resources provided by today’s many-core devices. We demonstrate the potential of our approach in astronomy, where hundreds of million nearest neighbor queries have to be processed. ER -
APA
Gieseke, F., Heinermann, J., Oancea, C. & Igel, C.. (2014). Buffer k-d Trees: Processing Massive Nearest Neighbor Queries on GPUs. Proceedings of the 31st International Conference on Machine Learning, in PMLR 32(1):172-180

Related Material