Application of Supervised Descent Method to Parametric Level-set Approach

Abstract

In this work, we investigate the feasibility of the parametric level-set inversion using supervised descent method (SDM) for 2D microwave data. Specifically, the compactly supported radial basis functions (CSRBFs) are used to describe the domain of investigation. The SDM inversion learns the descent directions of the cost function offline, and updates the models using the learned descents in online prediction. Numerical examples show that the level-set & SDM approach can reconstruct the structures accurately in the data sparse case.