Progress In Electromagnetics Research, PIER 26, 67–87, 2000 AN INVERSE SCATTERING APPROACH BASED ON A NEURAL NETWORK TECHNIQUE FOR THE DETECTION OF DIELECTRIC CYLINDERS BURIED IN A LOSSY HALF- SPACE E. Bermani and S. Caorsi Department of Electronics University of Pavia Via Ferrat 1, I-27100, Pavia, Italy M. Raffetto Department of Biophysical and Electronic Engineering University of Genoa Via Opera Pia 11A, I-16145, Genoa, Italy 1. Introduction 2. The Neural Network Approach 2.1 The Neural Network 2.1.1 The architecture and the training algorithm 2.2 The Electromagnetic Problem 3. Numerical Results 3.1 Validation of the Neural Network Technique 3.2 Robustness of the Network 3.2.1 Variations in the dielectric properties of the cylinder 3.2.2 Variations in the dielectric properties of the scenario 3.2.3 Robustness to noisy data 4. Conclusions References