Pattern recognition: statistical, structural and neural approaches
Pattern recognition: statistical, structural and neural approaches
Machine learning, neural and statistical classification
Machine learning, neural and statistical classification
Artificial Neural Networks: A Tutorial
Computer - Special issue: neural computing: companion issue to Spring 1996 IEEE Computational Science & Engineering
Machine Learning
IEEE Transactions on Pattern Analysis and Machine Intelligence
The Random Subspace Method for Constructing Decision Forests
IEEE Transactions on Pattern Analysis and Machine Intelligence
Computer Processing of Line-Drawing Images
ACM Computing Surveys (CSUR)
Bit Plane Decomposition and the Scanning n-tuple Classifier
IWFHR '02 Proceedings of the Eighth International Workshop on Frontiers in Handwriting Recognition (IWFHR'02)
Hi-index | 0.00 |
In this paper we investigate the properties of novel systems for handwritten character recognition which are based on input space transformations to exploit the advantages of multiple classifier structures. These systems provide an effective solution to the problem of utilising the power of n-tuple based classifiers while, simultaneously, addressing successfully the issues of the trade-off between the memory requirements and the accuracy achieved. Utilizing the flexibility offered by multi-classifier schemes we can subsequently exploit this complementarity of different transformations of the original feature space while at the same time decompose it to simpler input spaces, thus reducing the resources requirements of the sn-tuple classifiers used. Our analysis of the observed behaviour based on Mutual Information estimators between the original and the transformed input spaces showed a direct correspondence of the values of this information measure and the accuracy obtained. This suggests Mutual Information as a useful tool for the analysis and design of multi-classifier systems. The paper concludes with a number of comparisons with results on the same data set achieved by a diverse set of classifiers. Our findings clearly demonstrate the significant gains that can be obtained, simultaneously in performance and memory space reduction, by the proposed systems.