People and machines are being inundated with ever-increasing amounts of data that include sensor readings, video, text and audio. It would be useful to reduce these large data sets by identifying elements that are salient, within the context of a mission or goal. Current approaches to automating the identification of salient information cannot adapt to different data types, domains and goals. To address these issues, IAI and its collaborator, University of Maryland (UMD), have been awarded a new contract entitled, “An Adaptive, Biologically-Inspired Framework for Identifying Salience in Data.” A framework that can automatically adapt to a given context will be developed, drawing from human perceptual processing and recent artificial neural network models that can control internal information flow. The framework will leverage UMD’s expertise in gated neural network architectures, which are highly general and adaptive representations. These can be automatically generated through learning and evolutionary computation, by providing either sample labels from human experts and/or performance feedback. This methodology can adapt to different data modalities, domains, contexts and goals through a process of offline training, including supervised learning, reinforcement learning and even unsupervised learning, along with evolutionary computation. Distributed computation will be used to handle the computationally intensive learning techniques. IAI’s approaches to processing video data and IAI’s in-house infrastructure for distributed computation and capabilities for large-scale data processing, particularly within the context of machine learning, will be leveraged in this effort. The decision support systems developed using these trained networks can be deployed to support analysts.