Parallelization of the nearest-neighbour search and the cross-validation error evaluation for the kernel weighted k-nn algorithm applied to large data dets in matlab

Rubio, Gines; Guillen, Alberto; Pomares, Hector; Rojas, Ignacio; Paechter, Ben; Glosekotter, Peter; Torres-Ceballos, C. I.

Abstract

The kernel weighted k-nearest neighbours (KWKNN) algorithm is an efficient kernel regression method that achieves competitive results with lower computational complexity than Least-Squares Support Vector Machines and Gaussian Processes. This paper presents the parallel implementation on a cluster platform of the sequential KWKNN implemented in Matlab. This implies both the parallelization of the k nearest-neighbour search and the evaluation of the cross-validation error on a large distributed data set. The results demonstrate the good performances of the implementation.

Citation

Rubio, G., Guillen, A., Pomares, H., Rojas, I., Paechter, B., Glosekotter, P., & Torres-Ceballos, C. I. (2009, June). Parallelization of the nearest-neighbour search and the cross-validation error evaluation for the kernel weighted k-nn algorithm applied to large data dets in matlab. Presented at 2009 International Conference on High Performance Computing & Simulation

• Conference Name: 2009 International Conference on High Performance Computing & Simulation
• Start Date: Jun 21, 2009
• End Date: Jun 24, 2009
• Online Publication Date: Aug 7, 2009
• Publication Date: Aug 7, 2009
• Deposit Date: Aug 1, 2016
• Publisher: Institute of Electrical and Electronics Engineers
• Book Title: HPCS '09. International Conference on High Performance Computing & Simulation, 2009.
• ISBN: 978-1-4244-4906-4
• DOI: https://doi.org/10.1109/hpcsim.2009.5192804
• Keywords: Large Scale Scientific Computing, Libraries and Programming Environments, Languages, Message Passing, Matlab
• Public URL: http://researchrepository.napier.ac.uk/Output/321936