SPEAKER: Yuri Nesterov

ABSTRACT:

We consider a new class of huge-scale problems, the problems with sparse subgradients. The most important functions of this type are piece-wise linear. For optimization problems with uniform sparsity of corresponding linear operators, we suggest a very efficient implementation of subgradient iterations, which total cost depends logarithmically in the dimension. This technique is based on a recursive update of the results of matrix/vector products and the values of symmetric functions. It works well, for example, for matrices with few nonzero diagonals and for max-type functions.

We show that the updating technique can be efficiently coupled with the simplest subgradient methods, the unconstrained minimization method by B. Polyak, and the constrained minimization scheme by N. Shor. Similar results can be obtained for a new non- smooth random variant of a coordinate descent scheme. We present also the promising results of preliminary computational experiments.

SPEAKER: Hugh Chipman

ABSTRACT:

Sequential design, or "active learning" can be an effective way to plan a experiment, so as to gain maximal information about a response model. The data-generating mechanism as well as the scientific objective can have important influence on the way in which the design is generated, and the estimated response model. For example, if the objective is to maximize response, we may only be interested in accurate estimates near the maximum. In computer experiments, Gaussian process models are a common approach, and have been used for sequential design and optimization. Instead we use an adaptive nonparametric regression model ("Bayesian Additive Regression Trees", or BART) to deal with nonstationarities and other complex relationships. By providing both point estimates and uncertainty bounds for prediction, BART provides a basis for sequential design criteria to ?nd optima with few function evaluations. Other applications, including sequential design in high-throughput screening for drug discover will also be discussed.

There will be refreshments served beginning at 5 p.m.

CLICK HERE TO REGISTER

Abstract

Crystallography has been one of the mature sciences.  Over the years, the modern science of crystallography that started by experimenting with x-ray diffraction from crystals in 1912, has developed a major paradigm – that all crystals are ordered and periodic.  In-deed, this was the basis for the definition of “crystal” in textbooks of crystallography and x-ray diffraction. Based upon a vast number of experimental data, constantly improving research tools, and deepening theoretical understanding of the structure of crystalline materials no revolution was anticipated in our understanding the atomic order of solids.

However, such revolution did happen with the discovery of the Icosahedral phase, the first quasi-periodic crystal (QC) in 1982, and its announcement in 1984 [1, 2].  QCs are ordered materials, but their atomic order is quasiperiodic rather than periodic, enabling formation of crystal symmetries, such as icosahedral symmetry, which cannot exist in peri-odic materials.  The discovery created deep cracks in this paradigm, but the acceptance by the crystallographers' community of the new class of ordered crystals did not happen in one day.  In fact it took almost a decade for QC order to be accepted by most crystallographers. The official stamp of approval came in a form of a new definition of “Crystal” by the International Union of Crystallographers. The paradigm that all crystals are periodic has thus been changed. It is clear now that although most crystals are ordered and periodic, a good number of them are ordered and quasi-periodic.

While believers and nonbelievers were debating, a large volume of experimental and the-oretical studies was published, a result of a relentless effort of many groups around the world. Quasi-periodic materials have developed into an exciting interdisciplinary science. 

This talk will outline the discovery of QCs and discuss their structure as well as some of their properties and uses.

Bio

After receiving his doctorate from the Technion in Haifa, Israel, Danny Shechtman was an NRC fellow at the Aerospace Research Laboratories of Wright Patterson AFB, Ohio, where for three years he studied the microstructure and physical metallurgy of titanium aluminides. In 1975 he joined the Department of Materials Engineering at Technion, where he is currently a Distinguished Professor. Between 1981 and 2004 he was visited Johns Hopkins University several times on sabbatical as part of a joint program with NBS-NIST. During this period he discovered by TEM the Icosahedral Phase, which opened the new science of quasiperiodic crystals, and performed research on other subjects. As of 2004 he is also a Professor at MSE and Ames Lab at Iowa State University. His current research efforts center on developing strong and ductile magnesium alloys for a variety of applications and deformation mechanisms in B2 intermetallics.

Shechtman is a member of several academies, including the National Academy of Engineering, and he is an honorary member of professional societies around the globe. He has been awarded many prizes, including the Wolf Prize in Physics, the Gregori Aminoff Prize of the Royal Swedish Academy of Sciences, the EMRS award and the 2011 Nobel Prize in Chemistry.

Join us for this hands-on training on effective ways to visually present your data for articles, poster sessions and more.
Space is first come, first served, but priority is given to those who have registered.
Register Now

Please contact Chris Doty (Phone: 404-894-4536) if you need additional information.

Part of the G.L.U.E. ('Graduate Library User Education') Workshop Series

"Design by Transformation - Application to Dense Linear Algebra Libraries"

Abstract

The FLAME project has yielded modern alternatives to LAPACK and related effort.  An attractive feature of this effort is the complete vertical integration of the entire software stack, starting with low level kernels that support the BLAS and finishing with a new distributed memory library, Elemental.  In between are layers that target a single core, multicore, and multiGPU architectures. What this now enables is a new approach where libraries are viewed not as instantiations in code but instead as a repository of algorithms, knowledge about those algorithm, and knowledge about target architectures. Representations in code are then mechanically generated by a tool that performs optimizations for a given architecture by applying high-level transformations much like a human expert would.  We discuss how this has been used to mechanically generate tens of thousands of different distributed memory implementations given a single sequential algorithm.  By attaching cost functions to the component operations, a highly optimized implementation is chosen by the tool that invariably matches or exceeds the performance of implementations by human experts. We call this approach Design by Transformation (DxT).

Bio

Robert van de Geijn is a Professor of Computer Science and member of the Institute for Computational Engineering and Sciences at UT-Austin. He received his Ph.D. in Applied Mathematics from the University of Maryland.  His interests are in linear algebra libraries, scientific computing, parallel computing, and formal derivation of programs. He has written more than a hundred refereed articles and several books.

This work is in collaboration with Bryan Marker, Don Batory, Jack Poulson, and Andy Terrell.

SPEAKER: Professor Weichung Wang

ABSTRACT:

Numerical simulations of three-dimensional photonic crystalshave gathered great importance in the investigations of their physicalproperties and real world applications. A main purpose of thesimulations is to predict particular lattice structures to achievemaximal bandgaps. This target is a challenge as the computations arehighly time-consuming and many tuning parameters are involved. Thesetuning parameters include structure parameters in the photonicscrystal lattices and algorithmic performance parameters in theeigenvalue solvers. To identify the optimal lattice structures and theoptimal algorithm combinations, we use several statistical tools tomodel the unknown response functions of the tuning parameters. Toolsarising in design and analysis of computer experiments will bediscussed to show the details of the schemes. Promising numericalresults will also be presented to justify the proposed schemes.