Title :
Multi-Valued Neuron with a periodic activation function — New learning strategy
Author :
Lupea, Valentin Mircea
Author_Institution :
Dept. of Comput. Sci., Politeh. Univ. of Timisoara, Timisoara, Romania
fDate :
Aug. 30 2012-Sept. 1 2012
Abstract :
Multi-Value Neurons (MVN) arouse from the necessity to solve those highly non-linear, multi-threshold problems. Various activation functions and learning algorithms were developed in the last years, which in time gave birth to a new type of MVN, called MVN with a periodic activation function (MVN-P). This neuron presented higher functionality than a MVN. The next natural step was trying to integrate this MVN-P into a Neural Network (NN) to increase further more its functionality.
Keywords :
learning (artificial intelligence); neural nets; MVN-P; learning algorithms; learning strategy; multivalue neurons; nonlinear multithreshold problems; periodic activation function; Artificial neural networks; Biological neural networks; Guidelines; Iris; Machine learning; Neurons; Zirconium; Multi-Valued Neuron (MVN); Neural Network (NN); activation function; learning strategy;
Conference_Titel :
Intelligent Computer Communication and Processing (ICCP), 2012 IEEE International Conference on
Conference_Location :
Cluj-Napoca
Print_ISBN :
978-1-4673-2953-8
DOI :
10.1109/ICCP.2012.6356164
