Reliable and Explainable Data-driven Methods for ELINT Signal Detection/Classification -- 25-012

The objective of this project is to develop data-driven signal processing methods for electronic intelligence (ELINT) signal detection and classification that are reliable and explainable.  We propose to develop a nonparametric data-driven classifier that harnesses the statistical properties of the data through the cumulant generating function of the training data. The cumulant generating function classifier (CGFC) is straightforward to implement using a simple and reliable convex optimization procedure that guarantees a globally optimum solution and requires only a single tuning parameter.  Realistic synthetic datasets of ELINT data will be constructed that have statistical properties which can be controlled for the purposes of analysis.  The CGFC approach will be applied to matched and mismatched training and testing datasets and its performance compared to model-based methods and data-driven machine learning methods.