Title :
Multi-Valued Neuron with a periodic activation function - learning with negative examples
Author :
Lupea, V.M. ; Cernazanu-Glavan, Cosmin
Author_Institution :
Dept. of Comput. Sci., Politeh. Univ., Timisoara, Romania
Abstract :
The efficiency of Multi-Value Neuron and Multi-value Neuron with a periodic activation function are well presented in literature. Using these types of neurons, highly non-linear problems can be solved using a single neuron and not a multi-layer Neural Network (NN). On the other hand, using negative example during learning was shown to increase the overall efficiency. This paper brings these concepts together.
Keywords :
learning (artificial intelligence); neural nets; transfer functions; learning; multivalued neuron; negative example; nonlinear problems; periodic activation function; single neuron; Artificial neural networks; Computational intelligence; Encoding; Iris; Neurons; Testing; Training; Multi-Valued Neuron (MVN); classifier; l-periodic; negative example; periodic activation function;
Conference_Titel :
Applied Computational Intelligence and Informatics (SACI), 2014 IEEE 9th International Symposium on
Conference_Location :
Timisoara
DOI :
10.1109/SACI.2014.6840067
