High Performance Computing Tools with Epidemiological Applications

Dr. William A. Maniatty
Clinical Assistant Professor
Department of Computer Science
Rensselaer Polytechnic Institute


Abstract


This talk focuses on the development of algorithms and software tools
supporting spatially explicit simulation of multi-species ecosystems. Many
interesting phenomena are difficult to explore via computer simulation
because limited computational resources tend to prevent the simulation's
execution within an acceptable time frame. The performance of the
simulation was enhanced by careful extraction of parallelism from the
underlying application. The development of parallel tools to model
epidemics and population dynamics resulted in both computational and
biological contributions.

The study of spatial and temporal aspects of multi-species biological
systems is central to ecology. Epidemics are significant to agriculture,
public health and ecology and this motivated their selection as an
application area. Knowing where and when these phenomena occur can be
useful for applications like pest control and disease control. Various
models of related phenomena and high performance computing techniques for
simulation were implemented. Past and current research issues include:
innovative modeling techniques, design of parallel algorithms, simulation
of epidemics (including the ecologically significant vector-borne case)
and performance analysis. The integrated parallel simulation system was
ported to several high performance computers, including: a MasPar MP-1, an
IBM SP2, a network of workstations and an SGI Origin 2000. Current work on
parallel simulation of epidemics, using genetic algorithms to model
evolutionary aspects of the spread of disease will be presented.