A Modified Two-Step ANN Ensemble Approach to Improve Generalization and its Application in Fractal Antenna Design

Abstract

In this work, a fractal antenna with rectangular base shape has been designed for 5.2[Formula: see text]GHz resonant frequency. Recently, ANN modeling of fractal antenna parameters have shown to be an effective approach for its design optimization. But the selection of ANN ensemble members and their output aggregation are two important issues for efficient operation of ANN ensembles. A novel two-step method based on output sensitivity of ANN and genetic algorithm optimization of ensemble members has been proposed in this paper for addressing the above two issues. The proposed approach, which is a modified form of existing technique, has been first validated by applying it over six benchmark functions and compared with existing approach for improved performance. Further, the proposed approach has been used to develop an ANN ensemble for design optimization of fractal antenna at 5.2[Formula: see text]GHz resonant frequency. The selected antenna geometry has been experimentally verified and has shown significant miniaturization of 48.34% as compared to conventional rectangular micro-strip antenna.