Overlapping Domain Cover for Scalable and Accurate Regression Kernel Machines

Mohamed Elhoseiny and Ahmed Elgammal

Abstract

In this paper, we present the Overlapping Domain Cover (ODC) notion for kernel machines, as a set of overlapping subsets of the data that covers the entire training set and optimized to be spatially cohesive as possible. We propose an efficient ODC framework, which is applicable to various regression models and in particular reduces the complexity of Twin Gaussian Processes (TGP) regression from cubic to quadratic. We also theoretically justified the idea behind our method. We validated and analyzed our method on three human pose estimation datasets and interesting findings are discussed.

Session

Learning and Recognition

Files

PDF iconExtended Abstract (PDF, 231K)
PDF iconPaper (PDF, 421K)
ZIP iconSupplemental Materials (ZIP, 305K)

DOI

10.5244/C.29.94
https://dx.doi.org/10.5244/C.29.94

Citation

Mohamed Elhoseiny and Ahmed Elgammal. Overlapping Domain Cover for Scalable and Accurate Regression Kernel Machines. In Xianghua Xie, Mark W. Jones, and Gary K. L. Tam, editors, Proceedings of the British Machine Vision Conference (BMVC), pages 94.1-94.12. BMVA Press, September 2015.

Bibtex

@inproceedings{BMVC2015_94,
	title={Overlapping Domain Cover for  Scalable  and Accurate Regression Kernel Machines},
	author={Mohamed Elhoseiny and Ahmed Elgammal},
	year={2015},
	month={September},
	pages={94.1-94.12},
	articleno={94},
	numpages={12},
	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
	publisher={BMVA Press},
	editor={Xianghua Xie, Mark W. Jones, and Gary K. L. Tam},
	doi={10.5244/C.29.94},
	isbn={1-901725-53-7},
	url={https://dx.doi.org/10.5244/C.29.94}
}