Agile Machine Learning in Dynamic Environments for Complex Event Processing

Under this STTR Phase I project, we propose a framework for complex event detection that addresses the limitations of the state-of-the-art by effectively combining deep learning models with neuro-symbolic reasoning. In our approach, we apply a joint deep neural network (DNN) and statistical relational learning (SRL) based framework that combines semantic reasoning and relational learning with data-driven methods for sensor exploitation, trajectory-based feature computation, activity feature extraction, and foundation data to overcome the challenges for complex event detection in dynamic environments. The Phase I effort will focus on developing a simulation framework to generate multi-sensor activity data that will serve as a test-bed for the proposed complex event detection system,  developing an end-to-end system that can ingest and process streams from multiple sensors and perform complex event detection, developing a joint data-driven and neuro-symbolic framework for reasoning for developing the joint training the architecture using a variety of datasets that include simulated data, internally collected real-world data, as well as public datasets, and investigating the performance and trade-offs of various technology components.