Title :
INFOTREE: a contextual tree approach to supervised pattern recognition
Author :
Van Dyck, D. ; Garcia, J. Batista
Author_Institution :
Dept. of Phys., Antwerp Univ., Belgium
Abstract :
A new method for pattern recognition is proposed. The model uses information theory to extract in a systematic and hierarchic way the relevant information content in the input pattern so as to avoid preprocessing. An example of the power of the algorithm is demonstrated by its use in the recognition of heavily noisy binary printed letters
Keywords :
feature extraction; information theory; learning (artificial intelligence); neural nets; noise; pattern recognition; trees (mathematics); INFOTREE; algorithm; artificial neural networks; contextual tree; feature vectors; information content; information theory; input pattern; noisy binary printed letters; supervised pattern recognition; Artificial neural networks; Data mining; Data preprocessing; Decision trees; Information theory; Management training; Mutual information; Pattern recognition; Physics; Power system modeling;
Conference_Titel :
Image Processing, 1997. Proceedings., International Conference on
Conference_Location :
Santa Barbara, CA
Print_ISBN :
0-8186-8183-7
DOI :
10.1109/ICIP.1997.638785
