Title :
Sample Selection Based on Minimum Likelihood Ratio
Author :
Liu, Gang ; Cui, Yu-jing ; Zhang, Hong-Gang ; Guo, Jun
Author_Institution :
Beijing Univ. of Posts & Telecommun., Beijing
Abstract :
Training data have important effect on recognition system performance. This paper proposes an algorithm of sample selection based on minimum likelihood ratio (MLR) which obtaining boundary samples for training. The experiment results show that this method is effective in improving performance of the recognition system.
Keywords :
boundary-value problems; data analysis; pattern recognition; sampling methods; boundary samples; minimum likelihood ratio; pattern recognition system; training data; Clustering methods; Cybernetics; Data engineering; Data mining; Machine learning; Multi-layer neural network; Neural networks; Pattern recognition; System performance; Training data; Boundary samples; Likelihood ratio; MLR; Samples selection;
Conference_Titel :
Machine Learning and Cybernetics, 2007 International Conference on
Print_ISBN :
978-1-4244-0973-0
Electronic_ISBN :
978-1-4244-0973-0
DOI :
10.1109/ICMLC.2007.4370105
