Investigating the use of the Supervised Descent Method for Electromagnetic Imaging in PEC Enclosed Chambers

We consider the use of the Supervised Descent Method (SDM) for 2D transverse magnetic imaging inside perfect electric conductor (PEC) enclosed electromagnetic imaging problems. When the electromagnetic imaging region is surrounded by PEC, and the background permittivity is lossless, classic optimization-based inversion algorithms such as Contrast Source Inversion sometimes provide degraded results compared to those obtained from unbounded domains. The SDM, by learning average search directions based on a large synthetic data set, may be able to avoid certain local minima that normally trap other optimization methods. Through the use of synthetic examples, we show that SDM can provide improved performance within an enclosed chamber.