Speakers, titles and abstracts:
- Lubomin Banas (Heriot-Watt University)
A convergent finite element approximation of a phase field model for
incompressible fluid flows.
Abstract:
We present a finite element approximation for a system of
Cahn-Hilliard and Navier-Stokes equations
which describes the flow of two immiscible and incompressible fluids.
We show stability bounds for the approximation and prove convergence,
and hence existence
of a solution to the system of coupled equations.
We propose an efficient implementation of the method and present some
numerical experiments.
- Coralia Cartis (University of Edinburgh)
Adaptive regularization methods for nonlinear optimization
Abstract:
A new class of methods for nonlinear nonconvex optimization problems will be
presented that approximately globally minimizes a quadratic model of the
objective regularized by a cubic term, extending to practical large-scale
problems earlier approaches by Nesterov (2007) and Griewank (1982).
Preliminary numerical experiments show our method to perform better than a
basic trust-region implementation, while our convergence and complexity
results show it to be at least as reliable to the latter approaches.
Extensions to problems with simple constraints will also be presented.
Lastly, solving nonlinear least-squares problems and systems of equations
will be discussed, and a lower-order regularization method will be analysed
in this context, namely where a first-order model of the least-squares is
regularized by a quadratic term. An overestimation property of functions
with Lipschitz-continuous Hessians, and of systems with Lipschitz-continuous
Jacobian underlies and justifies the entire work to be presented. This is
joint work with Nick Gould (RAL), Philippe Toint (Namur, Belgium), and for
the latter part of my talk, also with Stefania Bellavia and Benedetta Morini
(Florence, Italy).
- Jack Dongarra (University of Manchester, University of Tennessee at Knoxville and Oak Ridge National Laboratory)
Five important concepts to consider when using computational high
performance systems at scale
Abstract:
In this talk we examine how high performance computing has changed
over the last 10-year and look toward the future in terms of trends.
These changes have had and will continue to have a major impact on
our software.? Some of the software and algorithm challenges have
already been encountered, such as management of communication and
memory hierarchies through a combination of compile--time and
run--time techniques, but the increased scale of computation, depth
of memory hierarchies, range of latencies, and increased run--time
environment variability will make these problems much harder.?
We will look at five areas of research that will have an importance
impact in the development of software.
We will focus on following themes:
* Redesign of software to fit multicore architectures
* Automatically tuned application software
* Exploiting mixed precision for performance
* The importance of fault tolerance
* Communication avoiding algorithms
- Ozgur S. Ergul (University of Strathclyde)
Efficient and accurate solutions of large-scale electromagnetics
problems using the multilevel fast multipole algorithm
Abstract:
Solutions of scattering and radiation problems are extremely
important to analyze electromagnetic interactions of devices with each other
and with their environments including living and nonliving objects. A vast
variety of areas, such as antennas, radars, metamaterials, remote sensing,
electronic packaging, optical imaging, nanotechnology, medical imaging, and
wireless communications, involve scattering or radiation of electromagnetic
waves. Mathematical formulations of physical applications lead to set of
integral equations, which can be converted into matrix equations and solved
numerically on computers. Unfortunately, accurate simulations of real-life
problems usually require solutions of very large matrix equations, which
cannot be achieved easily without using special acceleration algorithms.
The multilevel fast multipole algorithm (MLFMA) is a powerful method for the
fast and efficient solution of electromagnetics problems discretized with
large numbers of unknowns. This method reduces the complexity of
matrix-vector multiplications required by iterative solvers, and it enables
the solution of large-scale problems that cannot be investigated by using
traditional techniques. Nevertheless, efficiency and accuracy of solutions
via MLFMA still depend on many parameters, such as the underlying
integral-equation formulation, discretization, iterative solver,
preconditioning, computing platform, parallelization, and many other details
of the numerical implementation. This talk will provide an overview of our
efforts to develop sophisticated implementations of MLFMA for rigorous
solutions of real-life scattering and radiation problems involving
three-dimensional complicated objects discretized with millions of unknowns.
- Doron Levy (University of Maryland, College Park)
Can mathematics cure leukemia? (Lessons on mathematics in medical sciences)
Abstract:
Leukemia is a cancer of the blood that is characterized by an abnormal production of white blood cells. The treatment of Chronic Myelogenous Leukemia (CML) was revolutionized over the past decade with the introduction of new molecular-targeted drugs such as Imatinib. While these drugs keep leukemia in most patients in remission, they do not cure the disease. In this talk we will show how mathematical modeling, analysis, and scientific computation, combined with new experimental data, suggest the feasibility of a low-risk, clinical approach to enhancing the effect of the drug therapy, possibly leading to a durable cure of the disease. We will use this scientific problem to demonstrate challenges that face the dialogue between applied mathematics and medical sciences.
- Magnus Svard (University of Edinburgh)
Shock capturing for high-order central difference schemes
Abstract:
During the last decade, there has been a considerable effort to derive linearly stable high-order finite difference schemes for flow problems. The key difficulty has been to design boundary closures and derive stability proofs by energy estimates for the initial-boundary value problem. The schemes have been shown effective for smooth flows in several benchmark cases. However, linear stability proofs does not guarantee stability nor convergence in the presence of discontinuous solutions, such as shocks.
In this work, we derive suitable diffusion terms for high-order central schemes and prove entropy stability without compromising the linear stability proofs for the initial-boundary value problem. Furthermore, we propose a limiter function that localizes the diffusion terms near discontinuities. For systems of conservation laws, we prove entropy stability of the limiter scheme and for a scalar conservation law, we also prove convergence.
Numerical computations demonstrate the properties of the proposed scheme and comparisons with the WENO schemes are also presented.
Finally, the extension of entropy stability to the initial-boundary value problem will be discussed.
|
