InTheLoop | 04.08.2013
ESnet’s Work in SDN, Energy Efficiency Highlighted at Global Networking Conference
ESnet’s contribution to the global networking community continues to gain momentum, most recently going up a notch or two with a series of talks, vendor presentations, and a demo at OFC/NFOEC 2013, the Optical Fiber Communication Conference and Exposition and the National Fiber Optic Engineers Conference, held March 17–21 in Los Angeles. Mike Bennett and Inder Monga were among those giving presentations. Read more.
NERSC Intern Wins Award for Computing Achievement
Stephanie Cabanela, a student intern in NERSC’s Operations Technology Group, was honored with a National Center for Women and Information Technology (NCWIT) Aspirations in Computing award on Saturday, March 16, in a ceremony in San Jose, CA. The award honors young women at the high school level for their computing-related achievements and interests. A senior at Lowell High School in San Francisco, Cabanela was one of hundreds in the Bay Area to compete for this award. Read more.
CS Mentoring Program Seeks Mentors and Protégés
If you are interested in finding a mentor to help you here at Berkeley Lab or in signing up to be a mentor, this year’s mentoring program will be beginning soon. The CS Mentoring/Networking Program is now accepting applications, with a deadline of April 18. For the Mentor Enrollment Form, go here; for the Protégé Enrollment Form, go here. If you have any questions about the program, please contact Marcia Ocon Leimer.
Horst Simon Named SIAM Fellow
Berkeley Lab Deputy Director Horst Simon has been named a Fellow of the Society for Industrial and Applied Mathematics (SIAM). The 2013 class of fellows were nominated for their exemplary research as well as outstanding service to the community. Through their contributions, SIAM Fellows help advance the fields of applied mathematics and computational science.
Ann Almgren and Weiqun Zhang Co-Author Paper at APS April Meeting
Ann Almgren and Weiqun Zhang of the Center for Computational Sciences and Engineering (CCSE) in CRD are co-authors of a paper to be presented at the American Physical Society April Meeting, April 13–16, in Denver, Colorado. The April Meeting gathers particle physicists, nuclear physicists, and astrophysicists to share new results and insights.
The paper, titled “The Most Powerful Stellar Explosions,” presents the results of 3D simulations of thermonuclear supernovae using CCSE’s CASTRO code. Ke-Jung Chen of the University of Minnesota is the principal author, with contributions from Alexander Heger of Monash University and Stan Woosley of UC Santa Cruz.
Jay Srinivasan Giving Talk at MoabCon 2013 Conference
Jay Srinivasan will give a talk on “Customer Experience with TORQUE / Moab at NERSC” at the MoabCon 2013 conference, which is being held April 8–11 in Park City, Utah. MoabCon 2013 will bring together Adaptive Computing customers, partners and industry experts for four days of talks, technical sessions, and networking events. The conference will provide attendees with opportunities to learn the latest HPC and cloud computing best practices and get an early look at the Adaptive Computing product roadmap.
Robert Saye Named 2013 Luis W. Alvarez Fellow
As the 2013 Luis Alvarez Fellow in Computing Sciences, Robert Saye will be developing numerical methods and computational tools for studying a wide range of problems involving multiple evolving interfaces. His methods will also have applications in multi-region shape optimization and image segmentation. Read more.
This Week’s Computing Sciences Seminars
Applications and Theory of a Continuum-Mechanics-Based Immersed Boundary Method
Monday, April 8, 2:00–3:00 pm, 50F-1647
Dharshi Devendran, University of Chicago
Fluid-structure interaction problems, in which the dynamics of a deformable structure is coupled to the dynamics of a fluid, are prevalent in biology. For example, the heart can be modeled as an elastic boundary that interacts with the blood circulating through it. The immersed boundary (IB) method is a popular method for simulating fluid-structure interaction problems. The traditional IB method discretizes the elastic structure using a network of springs, which makes it difficult to use material models from continuum mechanics within the IB framework. In this talk, I present a new IB method that uses continuum mechanics to discretize the elastic structure, with a finite-element-like discretization. Unlike other versions of the IB method that use finite elements, this method is easy to implement. Consequently, it can be used to set up new models within the IB framework quickly. The method is applied to a three-dimensional fluid-structure interaction problem. I will also present some (numerical) convergence analysis of the method.
Recovery of Simultaneously Structured Models with Limited Information
Monday, April 8, 2:30–3:30 pm, 521 Cory Hall, UC Berkeley
Maryam Fazel, University of Washington
Finding models with a low-dimensional structure, given a number of linear observations much smaller than the ambient dimension, has been well-studied in recent years. Examples of such models are sparse vectors, low-rank matrices, and the sum of sparse and low-rank matrices, among others. In many signal processing and machine learning applications, the desired model has multiple structures simultaneously. Examples include recovering a sparse signal from phaseless measurements (sparse phase retrieval), and learning models with several structural priors in machine learning tasks. Often convex penalties that promote individual structures are known, and require a minimal number of generic measurements (e.g., ℓ1 norm for sparsity, nuclear norm for matrix rank), so it is reasonable to minimize a combination of such norms to recover a simultaneously structured model. We show that, surprisingly, if we use multiobjective optimization with the individual norms, we can do no better (order-wise) in terms of required measurements than an algorithm that exploits only one of the structures. This result holds in a general setting and suggests that to fully exploit the multiple structures, we need an entirely new convex relaxation, not one that is a function of relaxations used for each structure. Time permitting, we also consider denoising for simultaneously structured signals, and provide bounds on the minimax denoising risk.
Using the NERSC Web Publications Database
Tuesday, April 9, 12:00–1:20 pm, OSF 943-238
Margie Wylie, LBNL Computing Sciences Communications Group
Structured Backward Errors for the Tridiagonal Eigenproblem: Scientific Computing and Matrix Computations Seminar
Wednesday, April 10, 12:10–1:00 pm, 380 Soda Hall, UC Berkeley
Beresford N. Parlett, UC Berkeley
We seek a bound on the relative perturbations to the parameters defining a tridiagonal matrix so that an approximate eigentriple (lambda, x, y*) becomes exact. We have found how to compute these perturbations rapidly, in O(n) operations for an n by n matrix. Now it is reasonable to compute this bound along with the triple itself and a relative condition number as standard procedure. This is joint work with F. Dopico and C. Ferreira.
Programmable Traffic Measurements and Graceful/Resilient Network Management
Wednesday, April 10, 1:00–2:30 pm, 50B-4205
Chen-Nee Chuah, University of California, Davis
A key challenge for network wide traffic monitoring is deciding both “How to configure?” and “Where to place?” individual monitors. Previous research has investigated ways to optimize both these decisions when monitoring objectives and traffic characteristics are known a priori. These approaches tend to be suboptimal in the face of dynamically altering monitoring objectives or traffic characteristics.
In the first part of the talk, we present a programmable network measurement framework that addresses the challenges mentioned above by (a) enabling application-aware and reconfigurable data collection at individual monitors, (b) performing online, close-loop traffic analysis to guide subsequent measurement configurations, hence directing computation and storage resources where most needed to achieve the desired accuracy, (c) intelligently routing/re-routing important traffic sub-populations on-the-fly across desired monitors that have been deployed through Measurement-aware Routing, and (d) solving large inference problems in distributed local monitors.
Network measurement is only one of multiple operations that jointly bring about the services provided by the network. The second part of the talk will highlight the importance of “synergistic networking,” which seeks jointly optimized design and/or scheduling of multiple network operations to achieve resilience and graceful degradation in the presence of failures and attacks. We introduce graceful network state migration framework, which synergistically schedules premeditated network operations to minimize potential service/performance disruption.
EECS Colloquium: Modeling and Analysis of High-Throughput RNA Structure Mapping Experiments
Wednesday, April 10, 4:00–5:00 pm, 306 Soda Hall (HP Auditorium), UC Berkeley
Sharon Aviran, Center for Computational Biology, Molecular and Cell Biology and Mathematics, UC Berkeley
New regulatory roles continue to emerge for both natural and engineered RNAs, many of which have specific structures essential to their function. This highlights a growing need to develop technologies that enable rapid and accurate characterization of RNA structure. Yet, available techniques that are reliable are also vastly limited, while the accuracy of popular computational methods is generally poor. These limitations pose a major barrier to determination of structure from sequence.
To address this need, we have developed a high-throughput structure characterization assay, called SHAPE-Seq, which simultaneously measures structural information at nucleotide-resolution for hundreds of distinct RNAs. SHAPE-Seq combines a novel chemistry with next-generation sequencing of its products. Following sequencing, we extract the structural information using a fully automated algorithmic pipeline that we developed. In this talk, I will focus on SHAPE-Seq’s analysis methodology, which relies on a novel probabilistic model of a SHAPE-Seq experiment, adjoined by maximum-likelihood parameter estimation. I will demonstrate the accuracy, simplicity, and efficiency of our approach, and will then present an algorithm that uses such structure mapping data to inform computational RNA secondary structure prediction.
Link of the Week: How to Cook Up a Math Poem in n Easy Steps
“A mathematical poem attempts to distill a mathematical concept and present it in a literary or visually compelling way,” writes Caleb Emmons in “How to Cook Up a Math Poem in n Easy Steps,” published in a recent issue of the Journal of Humanistic Mathematics. The paper presents an outline of his own personal method of composing such poetry, using a poem based on pi and Morse code as an example. The author acknowledges that interpreting the digits of pi as an alphanumeric code is well established; his first citation is David Bailey’s Pi Search, http://pi.nersc.gov/.