Colloquium Series
Spring 2017
All talks are from 3:454:45 p.m. in the Colloquium room, unless otherwise specified.
Tea and cookies will be served in the Lecture room starting at 3:30 p.m.

Apr26

The strange and varied appearances of extended multiple zeta values, or, how a pure mathematician sank so low as to write a paper with decimal points.Prof. Mike HoffmanUSNATime: 03:45 PM
View Abstract
Stanislaw Ulam famously described himself as "a pure mathematician who had sunk so low that his latest paper actually contained numbers with decimal points." I recently found myself in a similar situation. The extended multiple zeta values, which I defined in 1996, have taken on a life of their own and keep coming back to haunt me. After defining the extended multiple zeta values, I will describe three such "hauntings," with an emphasis on the most recent one.

Mar01

Lagrangian Relaxation of the DefenderAttackerDefender Constrained Shortest Path ProblemCDR Gary LazzaroUSNATime: 03:45 PM
View Abstract
Trilevel Defender AttackerDefender (DAD) optimization models have never been applied to the constrained shortest path problem before. The particular challenge associated the DAD constrained shortest path problem is that an additional side constraint breaks the network structure of a shortest path problem into a more complicated trilevel integer program. We create new solution procedures for the DAD constrained shortest path problem. We merge the attacker model with Lagrangian relaxation of the operator model into a single formulation that can obtain fast heuristic solutions. We combine our heuristic algorithm with traditional methods to obtain provably optimal or nearoptimal solutions. We test our algorithms on medium and large networks, and our results show that our innovations can significantly outperform traditional nested decomposition.

Feb08

TriLevel Optimization Algorithms for Solving DefenderAttackerDefender Network ModelsCDR Gary LazzaroUSNATime: 03:45 PM
View Abstract
The optimal defense and operation of networks against worstcase attack is an important problem for military analysts. We review development of existing solutions for the DefenderAttackerDefender (DAD) trilevel optimization model and investigate new applications and solution procedures. We develop an implicit enumeration algorithm that incorporates addition of new defenses as an alternative solution method for the DAD model. Our testing demonstrates that implicit enumeration can efficiently generate all equivalent optimal or nearoptimal solutions for DAD problems. When budgets for network defense or attack are uncertain, decision makers usually prioritize defenses in nested lists. We quantify the costs of various strategies for nesting of defenses. We design a parametric programming formulation of the DAD model to find nested defenses that have the smallest cost difference from optimal nonnested solutions.

Jan31

How the Rules of Engagement Affect the Emergence of ConsensusProf. Eitan TadmorUniversity of MarylandLocation: Rickover 102Time: 07:00 PM
View Abstract
Opinion dynamics in human crowds and flocking of birds are two prototypical examples for systems which are driven by the "social engagement" of members in such crowds. The "social engagements" are formulated by certain rules which quantify how each member is interacts with its neighbors. These local interactions may lead, over time, to global patterns of the whole crowd, such as the emergence of consensus of opinions, flocking of birds etc. We explore the question how specific rules of interaction lead to the emergence of consensus/flocking.

Jan19

On a class of multiscale problems arising in oceanic and atmospheric processesTime: 03:45 PM
View Abstract
For the last few years I have been working on a set of problems in ocean and atmosphere dynamics. They including modeling seaice's melting and reemergence, laser beam propagation in a turbulent medium, and formation of tornadoes. These problems have the common feature of possessing temporal and spatial scales that range over a dozen orders of magnitude. Their mathematical models almost always end up being turbulent dynamical systems with large positive Lyapunov exponents, and very large dimensional phase spaces. While these problems are exceeding difficult, we are fortunate in the last two decades to have access to field data to inform our models. Machine learning in combination with solving PDEs numerically, and incorporating data assimilation are the techniques explored to get a handle on the level of complexity these problems present. In this talk I will introduce three problems and discuss various approaches we have taken to analyze them.