Arts & Sciences Events
This calendar presented by
Arts & Sciences
[PAST EVENT] Mathematics Colloquium
February 17, 2012
2pm - 3pm
Title: Optimally Fragmenting Graphs Against Stochastically-located Threats: Containing Wildfire, Invasive Species, and Epidemics
Abstract: Increasing frequency of catastrophically-damaging wildfire events has stimulated interest among foresters in the effective use of preventative fuel reductions like dead-brush removal and small-scale controlled burns. How can data produced by forest scientists be used to determine the right level of investment in these preventative measures? How can we optimally place these treatments on the landscape to best complement real-time firefighting capabilities? Exploring these questions motivates a natural new family of budgeted stochastic optimization problems that fragment (or cut) a landscape graph to isolate a stochastically occurring ignition point. These models represent novel extensions of well-studied ideas in the theoretical computer science literature and capture crucial features of other contemporary environmental applications (e.g., containing the spread of invasive species). I will explain a hierarchy of efficient and provably-good algorithmic results for these models, and discuss some future directions.
Abstract: Increasing frequency of catastrophically-damaging wildfire events has stimulated interest among foresters in the effective use of preventative fuel reductions like dead-brush removal and small-scale controlled burns. How can data produced by forest scientists be used to determine the right level of investment in these preventative measures? How can we optimally place these treatments on the landscape to best complement real-time firefighting capabilities? Exploring these questions motivates a natural new family of budgeted stochastic optimization problems that fragment (or cut) a landscape graph to isolate a stochastically occurring ignition point. These models represent novel extensions of well-studied ideas in the theoretical computer science literature and capture crucial features of other contemporary environmental applications (e.g., containing the spread of invasive species). I will explain a hierarchy of efficient and provably-good algorithmic results for these models, and discuss some future directions.
Contact
[[rrkinc, Rex Kincaid]]