An irrelevant variability normalization approach to discriminative training of multi-prototype based classifiers and its applications for online handwritten Chinese character recognition
作者：
Highlights：
• Jointly discriminative training of feature transforms and classifier parameters.
• Rprop optimization for SSM-MCE based objective function is adopted.
• Both piecewise linear transform and weighted sum of linear transforms are compared.
• The IVN-based recognizer can be made both compact and efficient by using fast-match.
• Experiments show the effectiveness of our approach over the state-of-the-art.
摘要
Highlights•Jointly discriminative training of feature transforms and classifier parameters.•Rprop optimization for SSM-MCE based objective function is adopted.•Both piecewise linear transform and weighted sum of linear transforms are compared.•The IVN-based recognizer can be made both compact and efficient by using fast-match.•Experiments show the effectiveness of our approach over the state-of-the-art.
论文关键词：Irrelevant variability normalization,Sample separation margin,Minimum classification error,Rprop,Discriminative training,Online handwritten Chinese character recognition
论文评审过程：Received 4 January 2014, Revised 19 May 2014, Accepted 17 June 2014, Available online 23 June 2014.
论文官网地址：https://doi.org/10.1016/j.patcog.2014.06.014
