InTheLoop | 04.06.2015
Big Neuron to Open New Vistas in Brain Research
Over the years, dozens of imaging paradigms and algorithms have been created for visualizing the 3D structure of neurons—leading to a variety of disparate datasets in the field. To solve the mysteries of the brain, neuroscientists need to cross-compare these datasets. That’s why many of the field’s brightest minds and most advanced supercomputing facilities, including the Department of Energy's NERSC at Berkeley Lab, are participating in BigNeuron. This community effort will define and advance the state of the art of single neuron reconstruction and analysis and create a common platform for analyzing 3D neuronal structure. »Read more.
Almgren, Ng Named SIAM Fellows
Ann Almgren and Esmond Ng of Lawrence Berkeley National Laboratory’s Computational Research Division (CRD) are among the 2015 class of 31 mathematicians named as Fellows of SIAM, the Society for Industrial and Applied Mathematics.
Almgren is being recognized for contributions to the development of numerical methods for fluid dynamics and applying them to large-scale scientific and engineering problems. A staff scientist with the Center for Computational Sciences and Engineering (CCSE), she is also acting group lead of CCSE and the Scalable Solvers Group in CRD. Almgren’s areas of research encompass asymptotic analysis, numerical analysis and high-performance computing. She serves on the editorial board for SIAM Review.
Ng is being recognized for contributions to the development, analysis, and application of sparse matrix algorithms for solving large-scale scientific and engineering problems. Ng heads the Applied Mathematics Department in CRD. He has served as Vice Chair of the SIAM Activity Group on Supercomputing as well as on the editorial boards of the SIAM Journal on Scientific Computing and the SIAM Journal on Matrix Analysis and Applications. His areas of research span sparse matrix computation, numerical linear algebra, computational complexity, parallel computing and mathematical software. »Read more.
MyNERSC: Easier Access to Data, Jobs, Wait Times
MyNERSC, a web-based portal that provides NERSC users with real-time information on their jobs, disk usage, allocations and queue wait times, is garnering rave reviews following recent upgrades that have broadened its functionality and streamlined its ease of use. New enhancements include an individualized dashboard for all users containing repo and disk usage information, job and system status, completed job information and global chat. There is also an advanced queue-time predictor based on historical data, which for the first time can exclude any time jobs spend on hold, allowing users to optimize their job scripts to minimize queue latency. In addition, users can now monitor the backlog of NERSC systems and track repos’ usage (theirs and others) over time up to years. They can also monitor over time file system and compute benchmark performance at NERSC using the same data NERSC uses to monitor performance on its HPC systems. In addition, MyNERSC lets users to see a rich amount of detail regarding queued, running and completed jobs. This includes tracking the rank in the queue over time and a variety of performance data from Darshan, ALTD, Cray Proc Stats and LMT about completed jobs. »Read more.
World's Largest Database of Elastic Properties to Accelerate Materials Research
Scientists at Berkeley Lab have published the world’s largest set of data on the complete elastic properties of inorganic compounds using the infrastructure of the Materials Project hosted at the Department of Energy's National Energy Research Scientific Computing Center (NERSC). This database increases by an order of magnitude the number of compounds for which such data exists.
This new data set is expected to be a boon to scientists working on developing new materials where mechanical properties are important, such as for hard coatings, or stiff materials for cars and airplanes. While there is previously published experimental data for a few hundred inorganic compounds, the Materials Project now stores the complete elastic properties for 1,181 inorganic compounds, with dozens more added every week. The data are published and shared with other scientists via the Materials Project web site. This Science Gateway allows scientists to virtually construct possible new materials and calculate their predicted properties before moving to experimentation, speeding the search for new and novel materials for a range of applications. »Read more.
»Learn more: Materials Project at Berkeley Lab.
This Week's CS Seminars
Applied Math Seminar: Coarsening of Particle Systems
Tuesday, April 7, 11:00am - 12:00pm, 939 Evans Hall, UC Berkeley campus
David Levermore, University of Maryland
Each particle in a simulation of a system of particles usually represents a huge number of real particles. We present a framework for constructing the dynamics for the so-called coarsened system of simulated particles. We build an approximate solution to the Liouville equation for the original system from the solution of an equation for the phase-space density of a smaller system. We do this with a Markov approximation in a Mori-Zwanzig formalism based on a reference density. We then identify the evolution equation for the reduced phase-space density as the forward Kolmogorov equation of a Markov process. The original system governed by deterministic dynamics is then simulated with the coarsened system governed by this Markov process. Both Monte Carlo (MC) and molecular dynamics (MD) simulations can be view from this framework. More generally, the reduced dynamics can have elements of both MC and MD.
Applied Math Seminar: The Mori-Zwanzig formalism for the reduction of complex dynamics models
Wednesday, April 8, 2:30pm - 3:30pm, 939 Evans Hall, UC Berkeley campus
Xiantao Li, Pennsylvania State University
Mathematical models of complex physical processes often involve large number of degrees of freedom as well as events occurring on different time scales. Therefore, direct simulations based on these models face tremendous challenge. This focus of this talk is on the Mori-Zwanzig (MZ) projection formalism, which has re-emerged recently as a power tool for reducing the dimension of a complex dynamical system. The goal is to mathematically derive a reduced model with much fewer variables, while still able to capture the essential properties of the system. In many cases, this formalism also eliminates fast modes and makes it possible to explore events over longer time scales. The motivation for this work is from molecular dynamics models of material science problems, where only a small fraction of the atomic degrees of freedom are directly responsible for the defect formation and migration. But the methodology has been applied to macromolecular systems as well.
The models that are directly derived from the MZ projection are typically too abstract to be practically implemented. We will first discuss cases where the model can be simplified to generalized Langevin equations (GLE). Furthermore, we introduce systematic numerical approximations to the GLE, in which the fluctuation-dissipation theorem (FDT) is automatically satisfied. More importantly, these approximations lead to a hierarchy of reduced models with increasing accuracy, which would also be useful for an adaptive model refinement (AMR).
Examples, including the nonlinear Schrodinger equation, molecular dynamics models of materials defects, nanoscale heat conduction and molecular models of proteins, will be presented to illustrate the applications of the methods.
BIDS Data Science Lecture: Enabling Collaboration Workflows for Datasets Using Dat
Friday, April 10, 1:00pm - 2:30pm, 190 Doe Library, UC Berkeley campus
Max Ogden, US Open Data
Dat is an open source tool, funded by the Sloan Foundation in the United States as part of their Science Tools research funding, that seeks to enable collaboration workflows for datasets with an emphasis on approaches that are automated and reproducible.
The core Dat tool is a streaming dataset versioning and replication system developed with a heavy Unix philosophy designed to encourage extreme modularity and enable many third-party applications to be built on top.
Our project website is https://github.com/maxogden/dat.
In this talk, we will introduce what we are working on and seek feedback from the audience.
Upcoming Seminars and Events
April 16 CS Seminar
Electromagnetic scattering by gratings and random rough surfaces: implementation of high performance algorithms for solving eigenvalues problems
11:00am - 12:00pm, Bldg. 50F, Room 1647
Cihui Pan, Universite de Versailles Saint- Quentin-en-Yvelines (UVSQ)
In this talk, we consider the physical and numerical aspects of the electromagnetic scattering by grating and rough surfaces which has a wild range of applications in optics, material sciences, communications, oceanography and remote sensing. The curvilinear coordinate method (C-method) is an exact method for this physical problem. The C-method is based on Maxwell’s equations under covariant form written in a non-orthogonal coordinate system and this method leads to an eigenvalue problem of the scattering matrix. All the eigenvalues and eigenvectors of this complex, nonsymmetric and dense scattering matrix are required and this leads our research on the global eigensolvers.
For gratings, we developed a new version of C-method which leads to a differential system with initial conditions instead of the eigenvalue problem. This new version of C-method can be used to study multilayer gratings with homogeneous medium or inhomogeneous medium.
For rough surfaces, we present the parallel QR algorithm that is specifically designed for the C-method, where we take advantage of Rayleigh expansion and define the “early shifts”. The multi-window bulge chasing and aggressive early deflation are also used. The implementation shows a significant speed up. To improve the scalability, we conduct research on the “Spectral projection method as a global eigensolver” where we combine the multiple implicitly restarted Arnoldi method with nest subspaces (MIRAMns) and the contour integral based projection method (SSM) to form a global eigensolver. This global eigensolver has a parallel nature and good scalability. We show this global eigensolver has a multilevel parallel nature and a very good potential scalability. We present some experiments to validate this approach.
Link of the Week: Asking for Advice Makes a Good Impression
In a recent Scientific American article, Harvard professors Alison Wood Brooks and Francesca Gino write about their studies that found asking for advice isn't detrimental, as people often fear. In fact the opposite is true:
Though extremely common, fears about appearing incompetent by asking for help or information are sorely misplaced. Here is why: when you ask for advice, people do not think less of you; they think you are smarter. They reason, “I'm brilliant (of course), so this guy's smart for asking for my advice.” And by asking someone to share his or her wisdom, a person strokes the adviser's ego and can gain valuable insights. Indeed, seeking guidance from others encourages information exchange and meaningful connection between us and our friends and colleagues.