All Seminars
| Title: Sensor Web: Research Challenges and Opportunities |
|---|
| Seminar: Computer Science |
| Speaker: WenZhan Song of Georgia State University |
| Contact: Vaidy Sunderam, vss@emory.edu |
| Date: 2012-11-30 at 3:00PM |
| Venue: MSC W301 |
| Download Flyer |
| Abstract: In this talk, we will discuss several research challenges and opportunities of sensor web in environment monitoring, smart grid and smart environment. Several years ago, we designed and deployed the first space in-situ sensor web in Mount St. Helens collaborating with USGS and JPL in a NASA ESTO project. We are currently advancing this research agenda to create a new paradigm, VolcanoSRI (Volcano Seimic Realtime Imaging), for imaging the 4D volcano tomography in a large-scale sensor network, joint with UNC and MSU in a NSF CDI project. A future effort aims to integrate seismic tomography, InSAR and deformation model to make the fictional holographic projector known as Virgil in the film "Supervolcano" a reality. We are also collaborating with Cornell and UC Berkerley to investigate several key aspects of a computation and information foundation of the smart grids in a NSF CPS project. We are studying distributed demand and response algorithms and designing an open and scalable experimental platform for smart grid, known as SmartGridLab, that integrates a hardware testbed with a software emulator, allowing software virtual nodes to interact with physical nodes in the testbed. We also discusses several research opportunities on smart environments, with the goal of enabling smart healthcare and ambient intelligence. |
| Title: Gossip-based distributed matrix computations |
|---|
| Seminar: Scientific Computing Seminar |
| Speaker: Hana Strakova of University of Vienna and GA Tech |
| Contact: TBA |
| Date: 2012-11-28 at 12:50PM |
| Venue: W306 |
| Download Flyer |
| Abstract: Truly distributed matrix computations with randomized communication schedules, such as gossip-based algorithms, can offer many attractive properties. Due to their randomized communication restricted only to direct neighbors they are very flexible with respect to the underlying hardware infrastructure. They can operate on arbitrary topologies and they can be made resilient against dynamic changes in the network, against message loss or node failures, and against asynchrony between compute nodes. Moreover, their overall cost can be reduced by accuracy-communication trade-offs. Such properties are attractive especially for loosely-coupled distributed systems with unreliable communication links, such as sensor or P2P networks. However, due to the growth in the number of nodes for future extreme-scale HPC systems and the anticipated decrease in reliability, some properties of gossip-based distributed algorithms may become important also for future HPC systems. We are investigating distributed algorithms for various prototypical matrix computation problems which utilize gossip-based aggregation algorithms for performing reduction operations in a distributed manner. Questions addressed relate to the (communication) cost paid for the increased flexibility and robustness, to convergence properties, to numerical accuracy achieved, as well as to the benefits of accuracy-communication cost trade-offs. |
| Title: A characterization of cusp forms by means of the growth of their Fourier coefficients |
|---|
| Seminar: Algebra and Number Theory |
| Speaker: Winfried Kohnen of Heidelberg University |
| Contact: David Zureick-Brown, dzb@mathcs.emory.edu |
| Date: 2012-11-28 at 3:00PM |
| Venue: W306 |
| Download Flyer |
| Abstract: We will characterize elliptic cusp forms and also Siegel cusp forms of degree two by means of the growth of their Fourier coefficents. |
| Title: Introduction to GPU Computing: Basics of OpenCL |
|---|
| Seminar SIAM Student Chapter: Numerical Analysis and Scientific Computing |
| Speaker: Veronica Mejia Bustamante of Emory University |
| Contact: Veronica Mejia Bustamante, vmejia@emory.edu |
| Date: 2012-11-27 at 4:00PM |
| Venue: W306 |
| Download Flyer |
| Abstract: The Emory SIAM Student Chapter is pleased to present an introductory seminar to GPU computing. The seminar is open to both graduate and undergraduate students interested in learning more about this growing field of scientific computing applications. This seminar will provide an introduction to GPU programming in both the hardware and software level. We will discuss the hardware setup of a GPU and the threads execution model. We will also cover the basics of the OpenCL API including the programming model and memory hierarchy and discuss easy techniques to begin writing your first GPU program in OpenCL. This will be an informal session, no previous GPU computing experience is required. Great food will be provided during the seminar! |
| Title: Induced and noninduced Ramsey numbers of $k$-partite, $k$-uniform hypergraphs |
|---|
| Seminar: Combinatorics |
| Speaker: Steve La Fleur of Emory University |
| Contact: Dwight Duffus, dwight@mathcs.emory.edu |
| Date: 2012-11-16 at 4:00PM |
| Venue: MSC W303 |
| Download Flyer |
| Abstract: Given two (hyper)graphs $S$ and $T$, the Ramsey number $r(S,T)$ is the smallest integer $n$ such that, for any two-coloring of the edges of $K_n$ with red and blue, we can find a red copy of $S$ or a blue copy of $T$. Similarly, the induced Ramsey number, $r_{\mathrm{ind}}(S,T)$, is defined to be the smallest integer $N$ such that there exists a (hyper)graph $R$ with the following property: In any two-coloring of the edges of $R$ with red and blue, we can always find a red \emph{induced} copy of $S$ or a blue \emph{induced} copy of $T$. In this talk we will discuss bounds for $r(K^{(k)}_{t,\dots,t}, K_s^{(k)})$ where $K^{(k)}_{t,\dots,t}$ is the complete $k$-partite $k$-graph with partition classes of size $t$. We also present new upper bounds for $r_{\mathrm{ind}}(S, T)$, where $T \subseteq K^{(k)}_{t,\dots,t}$ and $S \subseteq K_s^{(k)}$. |
| Title: An information theoretic approach to Sobolev and isoperimetric inequalities |
|---|
| Colloquium: N/A |
| Speaker: Professor Deane Yang of NYU-Poly |
| Contact: Vladimir Oliker, oliker@mathcs.emory.edu |
| Date: 2012-11-15 at 4:00PM |
| Venue: MSC W301 |
| Download Flyer |
| Abstract: I will describe connections among information theoretic, Sobolev-type, and isoperimetric inequalities and how these ideas can be used to establish new results. |
| Title: Scalable Efficient Methods for Incompressible Fluid-dynamics in Engineering Problems |
|---|
| Defense: Dissertation |
| Speaker: Umberto Villa of Emory University |
| Contact: Umberto Villa, uvilla@emory.edu |
| Date: 2012-11-12 at 4:30PM |
| Venue: MSC W301 |
| Download Flyer |
| Abstract: Accurate and effective methods for the numerical solution of incompressible fluid dynamics is an old but still important challenging problem, as more and more complex problems in engineering biology, ecology, medicine, sport are tackled with computational methods.\\ \\ In this dissertation defense, we investigate efficient solvers for two important models that governs the motion of a fluid, the incompressible Navier-Stokes and the Brinkman equations. The former describes the motion of an incompressible fluid in either an open or closed domain. The latter is used for describing the dynamics in a matrix of an inhomogeneous porous media, alternating bubbles and open channels.\\ \\ For the solution of the unsteady Navier-Stokes Equations, we move from the pressure correction algebraic factorization formerly proposed by Saleri, Veneziani (2005). These schemes feature an intrinsic hierarchical nature, such that an accurate approximation of the pressure Schur complement is obtained by computing intermediate low-order guesses. The difference between the pressure at two successive correction steps provides a natural a-posteriori estimator for time adaptivity with no additional computational cost. \\ \\ For the solution of the Brinkman Equations, we follow the approach presented in Mardal, Winther (2011) to precondition symmetric saddle point problems in a Hilbert settings. More specifically, we first present a novel mixed formulation of the Brinkman problem, with improved stability properties, in which we introduce the flow's vorticity as additional unknown. Based on stability analysis of the problem in the H(curl) - H(div) - L2 norms, we derive a scalable block diagonal preconditioner which is optimal in the constant coefficient case.\\ \\ Algorithms and preconditioners analysed in this thesis have been implemented in a parallel C++ code, using the finite element libraries LifeV and MFEM, and the linear algebra libraries Trilinos and HYPRE. We emphasize the performance of the proposed algorithms in solving problems of practical interest, involving complex geometries and realistic flow conditions. Numerical experiments in 2D and 3D confirm the effectiveness of our approach showing good efficiency and parallel scalability properties of the solvers proposed. |
| Title: Iterative Methods and Partitioning Techniques for Large Sparse Problems in Network Analysis |
|---|
| Defense: Dissertation |
| Speaker: Verena Kuhlemann of Emory University |
| Contact: Verena Kuhlemann, vkuhlem@emory.edu |
| Date: 2012-11-09 at 1:00PM |
| Venue: MSC W303 |
| Download Flyer |
| Abstract: The analysis of networks is an important aspect in many fields. Here we consider three different topics: the numerical solution of Markov chains, ranking of genes, and parallel computations with large scale-free graphs.\\ \\ First, additive Schwarz methods are a class of domain decomposition methods that are suitable for the solution of large linear systems in serial as well as in parallel mode. We adapt the Restricted Additive Schwarz (RAS) method to the computation of the stationary probability distribution vector of large, sparse, irreducible stochastic matrices. The convergence properties are analyzed and extensive numerical experiments aimed at assessing the effect of varying the number of subdomains and the choice of overlap are discussed.\\ \\ Next, the ranking of genes plays an important role in the identification of key genes for a specific disease. A modification of the PageRank algorithm that is used to rank web pages is the GeneRank algorithm. We assessed the performance of additive Schwarz methods as well as a Chebyshev iteration for the solution of the GeneRank problem.\\ \\ Finally, many large networks are scale-free. That is, the degree distribution follows a power-law. Currently available graph partitioners are not efficient for such an irregular degree distribution. The lack of a good partitioning leads to excessive inter-processor communication requirements during every matrix-vector multiplication on parallel distributed-memory computers. We present an approach to alleviate this problem based on embedding the original irregular graph into a more regular one by disaggregating (splitting up) vertices in the original graph. Even though the latter graph is larger, we are able to decrease the communication requirements considerably and improve the performance of matrix-vector multiplication. |
| Title: On Folkman-type problems |
|---|
| Seminar: Combinatorics |
| Speaker: Andrzej Dudek of Western Michigan University |
| Contact: Dwight Duffus, dwight@mathcs.emory.edu |
| Date: 2012-11-09 at 4:00PM |
| Venue: MSC W303 |
| Download Flyer |
| Abstract: A classical Ramsey theorem states that in any 2-coloring of the edges of a sufficiently large complete graph, one will always find a monochromatic complete subgraph. In 1970, Folkman extended this result showing that for any graph $G$ there exists a graph $H$ with the same clique number as $G$ such that any 2-coloring of the edges of $H$ yields a monochromatic copy of $G$. In this talk, we present some old and recent developments concerning Folkman-type results. |
| Title: Rotationally Symmetric Planes in Comparison Geometry |
|---|
| Defense: Dissertation |
| Speaker: Eric Choi of Emory University |
| Contact: Eric Choi, echoi7@emory.edu |
| Date: 2012-11-08 at 12:00PM |
| Venue: MSC W502 |
| Download Flyer |
| Abstract: Global Riemannian geometry is concerned with relating geometric data such as curvature, volume, and radius to topological data. Cheeger-Gromoll showed that any noncompact complete manifold $M$ with nonnegative sectional curvature contains a boundaryless, totally convex, compact submanifold $S$, called a soul, such that $M$ is homeomorphic to the normal bundle over $S$. In the first part of our talk, we show that if $M$ is a rotationally symmetric plane $M_m,$ defined by metric $dr^2 + m^2(r) d\theta^2$, then the set of souls is a closed geometric ball centered at the origin, and if furthermore $M_m$ is a von Mangoldt plane, then the radius of this ball can be explicitly determined. We show that the set of critical points of infinity in $M_m$ is equal to this set of souls, and we make some additional observations on the set of critical points of infinity when $M_m$ is von Mangoldt with a point at which the sectional curvature is negative. We close out the first part of the talk with showing that the slope $m^{\prime}(r)$ of $M_m$ near infinity can be prescribed with any number in $(0, 1]$. In the second part of our talk, we extend results in a paper by Kondo-Tanaka in which the authors generalize the Toponogov Comparison Theorem such that an arbitrary noncompact manifold $M$ is compared with a rotationally symmetric plane $M_m$ satisfying certain conditions to establish that $M$ is topologically finite. We substitute one of the conditions for $M_m$ with a weaker condition and show that our method using this weaker condition enables us to draw further conclusions on the topology of $M$. We also completely remove one of the conditions required for the Sector Theorem, one of the principal results in the same paper. |