Title :
Push-and-pull for piecewise linear machine training
Author :
Knoll, Travis Dean ; Lo, James Ting-Ho
Author_Institution :
Dept. of Math. & Stat., Maryland Univ., Baltimore, MD, USA
Abstract :
A piecewise linear machine (PLM) is capable of pattern classification. A simple and robust training method for the PLM called push-and-pull training is presented. This method avoids the difficulties encountered by the learning vector quantization (LVQ) methods of T. Kohonen. The machine was capable of finding optimal solution efficiency for problems with either separated or overlapping category regions
Keywords :
learning (artificial intelligence); pattern recognition; pattern classification; piecewise linear machine training; push-and-pull training; training method; Aggregates; Mathematics; Mirrors; Piecewise linear techniques; Probability density function; Prototypes; Robustness; Statistics; Vector quantization;
Conference_Titel :
Neural Networks, 1992. IJCNN., International Joint Conference on
Conference_Location :
Baltimore, MD
Print_ISBN :
0-7803-0559-0
DOI :
10.1109/IJCNN.1992.227113
