Modeling of L-Band Erbium-Doped Fiber Amplifiers Using Artificial Neural Networks
This work presents the modeling of L-band EDFAs using artificial neural networks, specifically comparing the performance between Multi-Layer Perceptron (MLP) and Recurrent Neural Network (RNN) architectures. Experimental data encompassing various amplification scenarios were used to train the models under different input configurations. The analysis demonstrates that the RNN model outperforms MLP in terms of convergence and estimation accuracy for the chosen amplifier figures of merit. Overall, a high-precision modeling is achieved, enabling the simulation and optimization of wideband amplification schemes under flexible transmission conditions with low complexity and low computational cost.