USE OF ARTIFICIAL NEURAL NETWORKS FOR MULTILAYER WAVEGUIDE ANALYSIS

The need to optimize processes has led to the construction of methodscapable of performing tasks with more consistency and accuracy. The present workproposes the optimization of a multilayer waveguide through computational algorithms.Neural Networks were developed with different configurations and trained with differentmethods: Cross Validation and Standard Training. The architecture consists of 7 inputsparameters and one output, with 3 hidden layers. The neural network with the bestperformance presented a Mean Square Error of 1.070×10-11, regression 0.999996 and16.25 seconds of processing, highlighting the Tansig activation function, evaluated asthe best function among those used.