Efficient Kernel Smoothing Regression using KD-Tree

Efficiency improved multivariant kernel regression using kd-tree

Description

Kernel regression is a power full tool for smoothing, image and signal processing, etc. However, it is computationally expensive when it is extented for multivariant cases. The efficiency can be improved by only using neighbors within the effective range arond a regression point. To improve the efficiency further, the kd-tree tool developed by Steven Michael http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?objectId=7030&objectType=file is used to efficiently identify points within a range. For large data sets, this code can reduce computation time by 3 to 5 times.

Acknowledgements

The author wishes to acknowledge the following in the creation of this submission:
KD Tree Nearest Neighbor and Range Search, Multivariant Kernel Regression and Smoothing

• MATLAB release: MATLAB 7.5 (R2007b)

[V. Poor]

 

[Turkay YILDIZ]