Title :
Multi-valued neuron with a periodic activation function — Influence of l-periodic parameter over the learning process
Author_Institution :
Dept. of Comput. Sci., Politeh. Univ., Timisoara, Romania
Abstract :
Multi-Value Neuron with a periodic activation function (MVN-P) represents an evolution of the Multi-Value Neuron (MVN) by introducing the concept of l-periodic occurrence of the possible output sectors. Increased efficiency was obtained based on this new approach and it was observed that the l-periodicity parameter has an important impact over the learning process. This paper presents a study of this impact in terms of learning efficiency (error rate and learning epochs).
Keywords :
error statistics; learning (artificial intelligence); transfer functions; MVN-P; error rate; l-periodic occurrence; l-periodicity parameter; learning efficiency; learning epochs; learning process; multivalue neuron; periodic activation function; Biological neural networks; Computational intelligence; Conferences; Distance measurement; Materials; Neurons; Zirconium; Multi-Valued Neuron (MVN); l-periodic; periodic activation function;
Conference_Titel :
Computational Intelligence and Informatics (CINTI), 2013 IEEE 14th International Symposium on
Conference_Location :
Budapest
Print_ISBN :
978-1-4799-0194-4
DOI :
10.1109/CINTI.2013.6705189
