Emerging Alerts with a General Learning Engine (EAGLE)

Machine learning (ML) and intelligence are usually done with the primary objective of maximizing the fidelity of the quality of the machine, providing a decision that will be as good as or better than that of a human subject matter expert (SME). The main drawback of this approach is that it is performed in a vacuum or without consideration of the amount of time or volume of representative data needed to train the machine effectively. The primary purpose of our proposal, termed the EAGLE (Emerging Alerts with a General Learning Engine) system, is to research the application of Conditional Generative Adversarial Networks (cGANs) with a limited training set, while using conditional information (i.e., common knowledge) output from rules that identify events of interest (EOIs). This alternative approach is expected to provide another mechanism for automatically training machines with a much smaller initial training dataset, while providing similar quality to pure machine decision-making. The benefit is that ML models for sparse events (such as anomalous spacecraft behavior) avoid bias that would have been present without the identification of EOIs by the rules.