Predictive Graph Convolutional Networks

Increased availability of graph-structured military data and recent technical advances in neural network design and training methods has led to an opportunity to advance the state of the art while simultaneously establishing and improving the ability:(1) to monitor a platform or force, (2) to predict capabilities and limitations of the force, and (3) to suggest opportunities and vulnerabilities. To that end, Metron and Northeastern University have developed in Phase I and propose to extend in Phase II a predictive graph convolutional network capability called TIDYNGS that provides Timely Inference on DYNamic Graph Sequences.