EECS graduate student and member of the Stochastic Systems Group in the Laboratory for Information and Decisions, LIDS, Emily Fox recently presented her work to the Neural Information Processing Systems, NIPS, conference on the automated construction of computer models to accurately describe complex and seeming erratic system behaviors such as the dancing of honeybees or the activity of stock markets.
Working with her advisor, Edwin S. Webster Professor of Electrical Engineering and Computer Science, Alan Willsky, and co-authors Erik Sudderth and Michael I. Jordan of the University of California at Berkely, Fox explained that by coming up with the simplest model to explain sets of data including patterns, helpful equations to describe them and probabilities to predict future or ongoing behaviors, the new methodology provides a tool that will potentially save time and effort for scientists, economists and researchers of complex systems.
Although the current focus of the work is on the model's descriptive abilities and the accuracy of the extracted models, these very models could also provide real-time estimation, tracking and prediction potentials such as the discovery of likely depositional environments in the field of oil exploration.
The research was funded by the Army Research Office and the Air Force Office of Scientific Research.
Read more in the Dec. 10, 2008 MIT News Office article, "Deciphering honeybee dances and stock market swings, Grad student's model brings order to complex systems through math."