Principal component analysis in application for filter tuning algorithm

This elaboration presents the improvement of an algorithm based on artificial neural network (ANN) for microwave filter tuning. In the applied algorithm, which is based on direct mapping of the detuned filter characteristics to the tuning element error, the sets of ANN learning vectors containing scattering filter characteristics and corresponding tuning element deviations are used. In the concept presented here filter characteristics are converted to “principal component” representation before ANN training. Such representation allows us to truncate less effective data components and thus significantly reduce the number of neurons in the ANN input layer. Experimental results of ANN training have shown that, when the presented approach is used, the ANN input vector dimension can be reduced even 32 times without losing ANN generalization ability.