Design of interdigital filters for millimeter-wave applications using artificial neural networks

Abstract

This work presents a purpose to accelerate the design process of mmWaves filters for 5G applications. To aim this objective, this work makes use of a technique of an artificial neural network called the multilayer perceptron. The model has 4 outputs and 6 inputs, where 2 outputs have a strong correlation and the other two have a weak correlation. By using 1000 samples, the model achieves an accuracy value of 0.927 for the outputs with strong correlation, but despite the good result, the predicted model still needs improvement, since |S11| had a not-so-nice performance.