Advanced Algorithm Development

"The Missile Defense Agency's (MDA) Decision Architecture (DA) supports algorithm research and development by providing an environment to simulate future threat behavior in response to deployed BMD systems. These excursionary analyses are critical toachieving a tolerable leak rate for deployed national and theater defense systems because emerging long-range ballistic missile powers are not utilizing robust test programs to insure threat system accuracy and reliability, and thus timely intelligence ofpotential enemy capabilities cannot always be expected to inform key architectural decisions. Thus, MDA faces a highly uncertain future countermeasures environment. No amount of "live fire" tests can anticipate and evaluate such a large and uncertaincounter-measure space. For this Phase I effort, Information Extraction and Transport, Inc. (IET) proposes to develop a Learned Mixture Approximation (LMA) algorithm for the DA. The proposed LMA algorithm would approximate the original Bayesian networkwith a small set of tractable approximate networks using the population Markov Chain Monte Carlo algorithm. IET proposes using such an algorithm to support efficient and effective inference for dynamic Bayesian Network generated within the DA environment.The result of this Phase I effort will be Learned Mixture Approximation (LMA) algorithm to support the optimization of dynamic BN generation in the Missile Defense Agency's (MDA) Decision Archite