Predictive Graph Convolutional Networks

The US Navy’s mission to maintain, train and equip combat-ready Naval forces requires that decision makers have situational awareness of the capabilities, limitations, vulnerabilities/opportunities for adversarial and allied forces. An incomplete or inaccurate understanding of the current landscape and associated trends could lead to suboptimal mission readiness and outcomes. Analysts need tools that can comprehend the large and complex amount of information they are presented with, and predict risks and trends from these data. Areté Associates proposes to use multi-channel, recurrent graph convolutional networks to make full use of all information available, and to create a system that is extensible to new data sources and types. This approach will provide useful information to improve situational awareness and readiness, while remaining extensible to accommodate the continual development of available data sources.