Mapeamento da dinâmica de laser semicondutor por meio de sistema híbrido neuro-evolutivo

Abstract

This paper proposes the modelling of the dynamics of optical laser based on quantum dots by extreme learning ma-chine in its learning conventional configuration and with optimization of its hidden layer in a neuro-evolutive hybrid system trained by a genetic algorithm (GA). The identification, which uses nonlinear autoregressive model with exogenous inputs (NARX), is op-posed to classical methods, mostly based on physical-experimental and mathematical study of system. The models found present validation through residues analysis and the laser database was obtained at a certain temperature in excited and ground states; the system responses is given in optical power as a function of the input current. The GA addition to the conventional model allowed the hidden nodes reduction in the network, at the highest operational cost.