Chromatic dispersion of microstructured fiber using neural network

In this paper a neural Network model of chromatic dispersion of a photonic crystal fiber with triangle-lattice and hexagonal geometry has been designed trained and simulated. The training data are carried out using the multipole method. The three layer's hexagonal PCFs studied in this paper have a silica core, obtained by introducing a defect. To train the proposed neural network we have used four structures with the same period different diameter and in each computation we have derived the refractive index of PCF. The obtained mean square error reaches 1e-5 for 60.000 epochs and the proposed neural network permit to the designer to compute directly the refractive index and the chromatic dispersion without return back to the multipole method.