Using Deep Learning and Reinforcement Learning to Tame Spatially Spreading Processes
Mark Crowley is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Waterloo and a core member of WICI. His research seeks dependable and transparent ways to augment human decision making in complex domains in the presence of spatial structure, large scale streaming data, or uncertainty. Dr. Crowley often works in collaboration with researchers in computational sustainability, forest management, ecology and medical imaging. Such domains offer unique challenges for traditional algorithms for decision making, prediction and anomaly detection. His work focuses on the fields of Reinforcement Learning, Deep Learning and Ensemble Methods.
Recent advances in Artificial Intelligence and Machine Learning (AI/ML) allow us to learn predictive models and control policies for larger, more complex systems than ever before. However, some important real world domains such as forest wildfire spread, flooding and medical imaging present a particular challenge. They contain spatially spreading processes (SSP) where some local features change over time based on proximity in space. This talk will present new approaches to learning for SSPs using Deep Learning and Reinforcement Learning such as learning a model of wildfire spread from satellite images as if the wildfire were an agent making decisions about where to move next. This approach could lead to learning models which are more interpretable and aid domain experts in creating rich, agent-based models based on data.