Generative, Adaptive Multi-Fidelity Simulation System (GAMFSS)

Seed Innovations (Seed) applies our combined experience in artificial intelligence, machine learning, data mining, modeling and simulation, scientific machine learning, genetic algorithms to accomplish the research for the Missile Defense Agency (MDA) Small Business Innovation Research (SBIR) topic: Generative, Adaptive Multi-Fidelity Simulation System (GAMFSS). Through the research on this project, Seed designs and develops a proof-of-concept GAMFSS system whose purpose is to allow the modeler to use advanced artificial intelligence / machine learning (AI/ML) techniques to develop, train, and test data-fitted models for use in simulations quickly and effectively. In addition to standard data mining and data fitting techniques, GAMFSS uses scientific machine learning (SciML) techniques to characterize models by data-fitting model characterization parameters prior to training. GAMFSS will further enhance the data driven model development, training, and testing processes by using a non-generational genetic algorithm to efficiently explore the model characterization and training spaces. In the development of the proof-of-concept GAMFSS system, Seed applies modern software techniques including data mining, artificial intelligence, machine learning, and modeling and simulation in support of quick and effective development and validation of complex data-driven models for simulation to help ensure their use in subsequent data analyses are performed effectively and ultimately help the warfighter protect the US against attack. When deployed to meet MDAs simulation data analysis challenge, GAMFSS enables development of validated data driven models to produce complex simulation results thereby substantially reducing the time required for data analytics and increasing data analyst efficiency. Approved for Public Release | 24-MDA-11906 (16 Sep 24)