Experimental study of performance of pattern classifiers and the size of design samples
References (7)
- et al.
Dimensionality and sample size consideration in pattern recognition practice
- et al.
Estimation of a covariance matrix under Stein's loss
Ann. Statist.
(1985) Introduction to Statistical Pattern Recognition
(1990)
Cited by (8)
Learning algorithms may perform worse with increasing training set size: Algorithm-data incompatibility
2014, Computational Statistics and Data AnalysisCitation Excerpt :One problem that concerns practitioners in all of the applications of learning is the size of the training dataset. Some theoretical results, simulations, and experience in the majority of applications indicate that the performance of a learning method improves as the training sample size increases (Yousef et al., 2004; Chan, 2003; Chan et al., 1998, 1999, 1997; Raudys, 1997; Raudys and Pikelis, 1980; Raudys and Jain, 1991; Takeshita and Toriwaki, 1995; Dobbin and Simon, 2007). Qualitatively speaking, performance is measured as the capability of the learning method to predict an unknown output from a new but known input.
Results in statistical discriminant analysis: A review of the former Soviet Union literature
2004, Journal of Multivariate AnalysisSample size issues in the choice between the best classifier and fusion by trainable combiners
2014, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)Likelihood-ratio-based verification in high-dimensional spaces
2014, IEEE Transactions on Pattern Analysis and Machine IntelligenceVerification under increasing dimensionality
2010, Proceedings - International Conference on Pattern Recognition
- 1
A part of this research was supported by Grant-in-Aid for Scientific Research of the Ministry of Education, Science and Culture (05558035) and the Hori Information Science Promotion Foundation.