Register Now for April 12-13 Workshop on Improving Data Management for Bioinformatics
Registration is now open for a workshop on “Improving Data Mobility & Management for International Bioinformatics” to be held April 12-13 at Berkeley Lab. The workshop is the latest in a series called CrossConnects, run by by ESnet. This workshop is also cosponsored by Indiana University. The workshop aims to bring together leaders in the bioinformatics, computing, and networking communities to discuss the resources, partners, and tools needed to support high performance data transfers, distributed data analysis and global collaboration in precision medicine, precision agriculture and their relevant ties to human and plant microbiomic and metagenomic research.»Learn more about CrossConnects.
Save the Date: Feb. 25 Bon Voyage for Greg Bell
On Feb. 25, please join us in wishing well our friend and colleague Greg Bell. Director of the Scientific Networking Division and ESnet, Greg is off on a new adventure as CEO of the start-up Broala. Details to come.
Wehner Quoted in New Scientist About Links Between Floods and Climate Change
New Scientist recently turned to CRD’s Michael Wehner as an expert on climate change in an article about the links between flooding in southern England and climate change. »Read “South England’s 2014 floods made more likely by climate change”.
Wang Hall at Sunset
Lab photographer Roy Kaltschmidt recently captured some beautiful images of Wang Hall at sunset during a gorgeous warm snap. You can see them on »Berkeley Lab’s photo archive site.
DOE Communications Summit to Feature Speakers from CS, JG
On Wednesday, Feb. 10, DOE’s Office of Science is convening a communications summit, bringing together more than 80 communicators from national labs, universities and DOE. One session will focus on “Telling the Unique Story of the User Facilities” and will feature short presentations by Computing Sciences Communications Manager Jon Bashor and JGI Public Affairs Manager David Gilbert. In the afternoon, Gilbert will lead a breakout session to discuss how to identify and amplify stories from the user facilities. The meeting will be held at DOE headquarters in Washington, D.C.
This Week’s CS Seminars
Improving Density Functional Theory for Warm Dense Matter
Mon., Feb. 8, 4–5pm, Bldg 50F Rm 1647
Aurora Pribram-Jones, Lawrence Livermore National Laboratory
Warm dense matter is a high-energy state of matter with characteristics of both solids and plasmas. It is found within planetary interiors, created during shock experiments, and observed along the path to ignition of inertial confinement fusion. The effects of these environments’ high temperatures and pressures demand a mixed quantum-classical treatment. Due to this complicated behavior, simulation of warm dense matter is notoriously challenging for both condensed matter and traditional plasma methods. One of the most successful methods for modeling warm dense matter to date uses density functional theory to describe the electrons within a material and classical molecular dynamics to describe its ions. We know, however, that this treatment ignores an important piece of the electronic energy’s explicit temperature dependence. In this talk, ensemble and other temperature effects on static and time-dependent electronic structure are examined through the lens of mathematical density functional theory. In addition, a new method uniquely suited to warm dense matter simulation will be presented: finite-temperature potential functional theory. Highly accurate, systematically improvable, and computationally efficient, it bridges the theoretical gap between condensed matter and plasma treatments and skirts the computational bottleneck of high-temperature density functional theory.
CS Alvarez Fellowship Seminar
Accelerating PDE-Constrained Optimization Problems using Adaptive Reduced-Order Models
Tues., Feb. 9, 10–11am, Bldg. 50B, Room 4205
Matthew J. Zahr, Luis Alvarez Fellowship Candidate, Stanford University
Optimization problems constrained by partial differential equations are ubiquitous in modern science and engineering. They play a central role in optimal design and control of multiphysics systems, as well as nondestructive evaluation and detection, and inverse problems. Methods to solve these optimization problems rely on, potentially many, numerical solutions of the underlying equations. For complicated physical interactions taking place on complex domains, these solutions will be computationally-expensive in terms of both time and resources to obtain, rendering the optimization procedure difficult or intractable. I will introduce a globally convergent, non-quadratic trust-region method to accelerate the solution of PDE-constrained optimization problems by adaptively reducing the dimensionality of the underlying computational physics discretization. In this approach, the method of snapshots and Proper Orthogonal Decomposition (POD) are used to build a reduced-order model whose fidelity is progressively enriched while converging to the optimal solution. This ensures the reduced-order model is trained exactly along the optimization trajectory and effort is not wasted by training in other regions of the parameter space. A novel minimum-residual framework for computing surrogate sensitivities of the reduced-order model is introduced that equips the trust-region method with desirable properties. The proposed method is shown to solve canonical aerodynamic shape optimization problems several times faster than accepted methods. This work has been extended to address the specific challenges posed by topology optimization, where high-dimensional parameter spaces are inevitable.
Applied Mathematics Seminar
Diffusion Forecast: A nonparametric modeling approach
Weds., Feb. 10, 3:30–4:30pm, 939 Evans Hall, UC Berkeley
John Harlim, Penn State University
I will discuss a nonparametric modeling approach for forecasting stochastic dynamical systems on smooth manifolds embedded in Euclidean space. In the limit of large data, this approach converges to a Galerkin projection of the semigroup solution of the backward Kolmogorov equation of the underlying dynamics on a basis adapted to the invariant measure. This approach allows one to evolve the probability distribution of non-trivial dynamical systems with an equation-free modeling. If time permitted, I will also discuss a semi-parametric modeling framework to compensate for model error by learning an auxiliary dynamical model for the unknown parameters.
Critical Points of Gaussian-Distributed Scalar Fields on Simplicial Grids
Thurs., Feb. 11, 12-1pm, Wang Hall 59-3042
Tom Liebmann, Image and Signal Processing Group, Department of Computer Science, Leipzig University
Simulations and measurements often result in scalar fields with uncertainty due to errors or output sensitivity estimates. This makes common methods and structures to access topological features, e.g. the contour tree, not applicable directly. Furthermore, known techniques addressing topology of uncertain scalar fields are either only capable of approximating the data or miss out on important properties like correlation.
In this talk I will give a brief introduction into our approach to look into the topological structure of Gaussian-distributed scalar fields. Besides describing some of the work we have done in the last year, I want to use the opportunity to introduce myself and talk about the goals I have for my time at the Laboratory.