Reconstruction of dielectric cylinder by multi-output least square support vector machine

To save the computation time and improve the accuracy of reconstruction results by support vector machine (SVM), a multi-output least square SVM (multi-output LS-SVM) algorithm is used to reconstruct the relative permittivity profile of a 2-D cylinder in free space. Firstly, the back-scattered electromagnetic fields at a number of observation positions are calculated by the method of moment to generate the training and testing data. Then the multi-output LS-SVM is trained to reconstruct both the relative dielectric permittivity ε_r and the electric conductivity σ of the object from simulated data added with 30dB Gaussian noise. Numerical results show that the approach is accurate and efficient.