The Colloquium Series of the
            Department of Computer Science,
		  University of Wyoming

			presents

                      Lucas R. Shaw

                  University of Wyoming
                     Masters Defense

  "A Computational Framework for Modeling the Spread of
 Pathogens and Generating Effective Containment Strategies
            in Weakly Connected Island Models"
 
                 Monday, April 23, 2007
                      Room EN5069

                    1:30 - 2:30 p.m.

Abstract:

This thesis provides a novel framework for modeling the spread of pathogens
throughout a population, and for generating policies that minimize the 
impact of those pathogens on the population. As far as we can tell, this
thesis provides the first epidemiological framework that combines all of 
the following three elements: agent-based simulation, mathematical analysis,
and a sophisticated optimization tool.

First, we describe our agent-based simulation for modeling certain types of
complex systems comprised of weakly connected islands. Each island contains
a population of agents that not only interact with each other, but can move
from island to island. Differential equations govern the state of the agents
within each island. A probability matrix models the movement of agents from
island to island. The complete simulation is parallelized by running each
island on a separate processor.

Second, we apply this model to the study of the spread of human viruses
between cities via air travel. Confirmation of our approach is provided
by real-world data that monitored the spread of the flu virus from the 
winter of 2005/2006. We then extend the model to include a supply of 
vaccine, which is to be distributed in an optimal fashion. We provide
mathematical analyses of this model, which allow us to determine whether
a vaccination policy is even feasible.

Third, we apply a sophisticated in-house optimization toolkit, based on
Evolutionary Algorithms (EAs), to search for effective vaccination policies
that minimize the number of sick people. These policies are compared to
plausible benchmark policies. The EA policies are superior. Analysis of 
the EA policies indicates that the vaccine is generally distributed to
(1) the city of origin for the virus, (2) cities that are traveled to most
often, and (3) smaller cities. The latter observation is most surprising 
and counter-intuitive.

Finally, we hand-craft a new benchmark policy that takes into account all
three observations. This new policy is superior to all of our benchmark
policies and the EA-generated policies. We believe this new policy is quite
general, however, if the model changes sufficiently, new policies will
eventually be required. In that case the EA is always available to guide
the human in the creation of the new policies for changing situations.