Long-term Patterns of Life from Sporadic Observations

This SBIR Phase-I project will develop novel approaches for unsupervised life-long learning architecture to extract long-term patterns of life from sporadic observations. The proposed architecture models longterm dependencies, learns associations between features from sensory observations and the contexts in which these features occur, identifies sudden changes in the patterns for anomaly detection and adapts to gradual changes in the patterns. The proposed works builds upon Novateur’s prior work on biological cognitive systems and incorporates computational models and learning algorithms to build a comprehensive patterns of life extraction system with strong parallels to all the capabilities of biological cognitive system.