Monday, November 25, 2013 - 4:00pm to 4:50pm
KEC 1001

Speaker Information

Mark Crowley
Postdoctoral Scholar
School of EECS
Oregon State University


There are many impediments to bringing about a more environmentally sustainable society but one that is often overlooked are the significant and interesting computational challenges that need to be solved in sustainability domains.  The field of Computational Sustainability attempts to connect research in the computational sciences to hard sustainability problems such as invasive species management, wildlife reserve design and sustainable forest management.  I will give an overview of some of these domains with a focus on spatial decision making problems where actions must be chosen across a large physical landscape.  These problems can be modelled as using Markov decision processes but the scale, complex dynamics and spatial correlations make these problems hard to solve using traditional algorithms. I will describe research I have been involved with obtaining approximate solutions to spatially factored MDPs using policy gradients and sample based planning on invasive species, forest harvest and wildfire prevention problems. I'll also outline future challenges in spatial planning which provide opportunities for AI and Machine Learning research to make an impact on making sustainable world more computationally feasible.

Speaker Bio

Mark Crowley is a postdoctoral scholar at Oregon State University and received his PhD in Computer Science from the University of British Columbia in 2012. His primary research project explores optimization algorithms appropriate for problems with spatial structure, particularly in sustainability domains. His research interests span decision making under uncertainty, machine learning, causal modelling, visualization of spatial data, inference in probabilistic graphical models and game theory.