DocumentCode :
3614946
Title :
Sample set reduction for nearest neighbor classifiers under different speed requirements
Author :
S. Grabowski;A. Jozwik
Author_Institution :
Comput. Eng. Dept., Tech. Univ. of Lodz, Poland
fYear :
2003
fDate :
6/25/1905 12:00:00 AM
Firstpage :
465
Lastpage :
468
Abstract :
We compare several sample set reduction algorithms for the 1-NN rule with two criteria in mind: classification accuracy and classification speed. The main conclusion is that under aggressive reduction requirements, our scheme with local reduced set selection performs better than conventional algorithms. The results also cast doubt upon the widely used consistency criterion for reduced set generation, especially in noisy domains.
Keywords :
"Nearest neighbor searches","Noise reduction","Proposals","Testing","Noise generators","Cellular neural networks","Genetic mutations","Filtering algorithms"
Publisher :
ieee
Conference_Titel :
CAD Systems in Microelectronics, 2003. CADSM 2003. Proceedings of the 7th International Conference. The Experience of Designing and Application of
Print_ISBN :
966-553-278-2
Type :
conf
DOI :
10.1109/CADSM.2003.1255123
Filename :
1255123
Link To Document :
بازگشت