Resumen

MORALES LAGUADO, Lina; ESPITIA CUCHANGO, Helbert  y  SORIANO MENDEZ, José. PROPOSAL OF A NEURO-DBR SYSTEM AND ITS APPLICATION TO THE PREDICTION OF LORENZ TIME SERIES. Cienc. Ing. Neogranad. [online]. 2010, vol.20, n.2, pp.31-51. ISSN 0124-8170.

This paper proposes the Lorenz time series prediction using a new method known as neural-DBR system and comparing this with a typical fuzzy-neural design. the neural-DBR technique is a union of neural networks and the defuzzification methodology based on Boolean relations (DBR). The DBR theory aims to facilitate the implementation of a fuzzy inference and improve processing time of fuzzy systems, getting also a good performance. Neural-DBR systems try to take advantage of the complementarity between both techniques, using their benefits and avoid the unfavorable ones of each. Firstly, the paper presents the neural-DBR training algorithm proposed for identification of nonlinear systems. Later, the identifier for Lorenz equations, using a neural-DBR system and comparing it with a typical fuzzy-neural design through the root mean square error (RMSE) and the correlation coefficient (IC) as performance indices. The results of the system proposed show the reduction in training time and computation calculation. Boolean logic is accepted as a useful tool for automata and digital systems design. An alternative to improve automation systems is using fuzzy sets instead of Boolean logic. This is to obtain a continuous description for the actuator. By such a change and implementing the methodology of design automation systems, the fuzzy inference systems based on Boolean relations may appear.

Palabras clave : Lorenz equations; DBR; back-propagation algorithm; Neuro-DBR systems; nonlinear systems.

        · resumen en Español     · texto en Español     · Español ( pdf )