Feature Based Machine Leaning for Multiple Target Detection and Debris Mitigation
In this research effort, Vadum will demonstrate the feasibility of a machine learning approach to address the problem of debris mitigation and improve multiple target discrimination. This algorithm is a very fast, highly accurate multi-class approach based upon the concepts of bagging (bootstrap aggregation), boosting and random subspace projection. This algorithm will allow for de-emphasis (probabilistic soft decisions) or suppression (hard decisions) of uninteresting scatterers, while maintaining ballistic missile target tracks within the BMDS (Ballistic Missile Defense System) threat environment. The approach inherently manages large data sets, high dimensionality, missing features and sample outliers while being cautious of over-fitting. The approach has been applied in the research areas of: malware/phishing/spam detection, ovarian cancer detection, protein interaction prediction, real-time human pose recognition and general feature selection. This proposal presents the novel application of this approach to ballistic target detection and debris mitigation.
Small Business Information at Submission:
601 Hutton St STE 109 Raleigh, NC -
Number of Employees: