From echeverr at stanford.edu Tue Jan 11 21:08:05 2011 From: echeverr at stanford.edu (David Echeverria Ciaurri) Date: Tue Jan 11 21:09:20 2011 Subject: [BANANA] Stanford SMART FIELDS seminar next Wednesday (January 19th) Message-ID: <4D2D3735.4040004@stanford.edu> STANFORD SMART FIELDS SEMINAR Speaker : Paul M.J. Van den Hof (Delft University of Technology) Title : Parameter Estimation and Identifiability in Large-Scale Reservoir Models Date : January 19th, 12:15pm - 1:15pm Location : Green Earth Sciences Bldg., Room 365 367 Panama Street, Stanford University Next Wednesday (January 19th) we have another Smart Fields seminar. The speaker will be Paul M.J. Van den Hof (p.m.j.vandenhof@tudelft.nl) from Delft University of Technology. He will talk on 'Parameter Estimation and Identifiability in Large-Scale Reservoir Models'. Please find more details about his presentation below. It will be at 12:15pm in Room 365, Green Earth Sciences building. You are all invited to attend the talk. You can read more about Stanford Smart Fields at http://smartfields.stanford.edu/ and see previous seminar presentations at http://smartfields.stanford.edu/resources.presentations.php. With best regards, David Echeverria Ciaurri --------------------------------------------------------- Parameter estimation and identifiability in large-scale reservoir models Paul M.J. Van den Hof Delft Center for Systems and Control Delft University of Technology, The Netherlands In reservoir engineering large-scale reservoir models are used for model-based operations as monitoring, control and (life-cycle) optimization. The estimation of accurate grid-block parameters seems important for providing reliable predictions of production data over the life cycle of the reservoir. Estimating a large number of parameters from production data however leads to problems of identifiability, induced by the limited information content of the data. The identifiability of parameters is considered and analyzed, and its relation with controlability and observability of the underlying model structures is shown. Additionally the question of how model structures can be approximated so as to achieve local identifiability, while retaining the interpretation of the physical parameters is addressed. The information content of production data plays an important role, and relations are explored with gradient-type optimization algorithms as well as with the use of prior geological information in a Bayesian parameter estimation setting. Paul M.J. Van den Hof received the Ph.D. degree from the Department of Electrical Engineering, Eindhoven University of Technology, The Netherlands in 1989. Since 1999 he is a full professor at Delft University of Technology, since 2004 in the Delft Center for Systems and Control, of which he was the director from 2004-2009. Since 2005 he also serves as scientific director of the national graduate school DISC (Dutch Institute of Systems and Control). His research work has concentrated on system identification, identification for control, and model-based optimization and control, including aspects of closed-loop experiments, uncertainty modeling and parametrization issues. He is involved in applications in industrial process control systems, including petroleum reservoir systems, and mechatronic systems. He has acted as General Chair for the 13th IFAC Symposium on System Identification, that was held in Rotterdam, the Netherlands in 2003. He has been a member of the editorial board of Automatica (1992-2005), the IFAC Council (1999-2005), and the Board of Governors of IEEE's Control System Society (2003-2005). He is an IFAC Fellow and a Fellow of IEEE. From saunders at stanford.edu Tue Jan 18 10:31:21 2011 From: saunders at stanford.edu (Michael Saunders) Date: Tue Jan 18 10:33:13 2011 Subject: [BANANA] LA/Opt seminar Thursday Jan 20 (Beresford Parlett) Message-ID: Linear Algebra and Optimization Seminar (CME 510) iCME, Stanford University http://icme.stanford.edu/seminars/seminarInfo.php?seminar_id=2 4:15pm Thursday Jan 20, 2011 Y2E2 101 http://campus-map.stanford.edu/index.cfm?ID=04-070 Professor Beresford Parlett CS Faculty Emeritus, UC Berkeley parlett@math.berkeley.edu http://www.eecs.berkeley.edu/Pubs/Faculty/parlett.html Localization in eigenvectors of tridiagonal matrices In applications to Quantum Chemistry and to Structural Engineering the tridiagonal matrices that are generated in intermediate stages have the property that some of the important eigenvectors have small support. In other words, the interesting part of the eigenvector is determined by a small part of the matrix. In particular, the vectors from an invariant subspace associated with a cluster of close eigenvalues are sometimes determined from the same small submatrix. Our goal is to determine such a submatrix in advance so that the cost of calculation is greatly reduced. We also want to discover quickly when there is no localization. The talk will present both theory and practice. Forthcoming: Thu Jan 27 Edmond Chiu Aero-Astro, Stanford edmondc@stanford.edu Thu Feb 03 Rob Tibshirani HRP and Statistics, Stanford http://www-stat.stanford.edu/~tibs Thu Mar 10 Jim Lambers Math, Southern Mississippi http://www.math.usm.edu/lambers/ -------------- next part -------------- An HTML attachment was scrubbed... URL: http://csmr.ca.sandia.gov/pipermail/banana/attachments/20110118/7684f701/attachment.html From egng at lbl.gov Fri Jan 14 15:27:18 2011 From: egng at lbl.gov (Rachel Lance) Date: Tue Jan 18 18:24:46 2011 Subject: [BANANA] Berkeley Lab - Computing Sciences Seminar - Wednesday, 1/19/2011, 11:00am Message-ID: <4D30DBD6.5070209@lbl.gov> Berkeley Lab - Computing Sciences Seminar Date: Wednesday, January 19, 2011 Time: 11:00am - 12:00pm Location: Bldg. 50B, Room 2222 Speaker: Xuefei Yuan Columbia University Title: "Numerical simulation of magnetic reconnection in dynamically adaptive grids via NKS? Abstract: Numerical simulations of the four-field extended MagnetoHydroDynamics (MHD) equations with hyper-resistivity terms present a difficult challenge because of demanding spatial resolution requirements. An $r$-type, variational grid generator based on an equidistribution principle defines a new grid at each time step by solving a single Monge-Amp\`{e}re (MA) equation, adaptive to the corresponding current density. This time-dependent sequence of grid coordinates defines a `grid velocity' that enters the equation set thereby avoiding the need for a separate rezone step. The MHD equations are transformed from Cartesian coordinates so that the solution-defined curvilinear coordinates replace Cartesian coordinates as the independent variables. The application of an implicit scheme to the time-dependent problem prevents the time step size from being restricted by a Courant-Friedrichs-Lewy (CFL) condition, but only by accuracy. The Newton-Krylov-Schwarz (NKS) algorithm is used to solve above systems at each time iteration in parallel. Convergence studies show that curvilinear solutions converge faster than Cartesian solutions, and accuracy studies show that curvilinear solutions can achieve the same accuracy as Cartesian solutions with fewer grid points. Host of Seminar: Alice Koniges Computing Sciences Seminars Web Site: http://crd.lbl.gov/SCG/CSSeminars. For additional information, such as site access or directions to the conference room, please contact CSSeminars-Help@hpcrd.lbl.gov. Web Contact: CSSeminars-Help@hpcrd.lbl.gov -------------- next part -------------- An HTML attachment was scrubbed... URL: http://csmr.ca.sandia.gov/pipermail/banana/attachments/20110118/200e295e/attachment.html From egng at lbl.gov Fri Jan 14 15:49:06 2011 From: egng at lbl.gov (Rachel Lance) Date: Tue Jan 18 18:25:21 2011 Subject: [BANANA] Berkeley Lab - Computing Sciences Seminar - Friday, 1/21/2011, 10:00am Message-ID: <4D30E0F2.1060100@lbl.gov> Berkeley Lab - Computing Sciences Seminar Date: Friday, January 21, 2011 Time: 10:00am - 11:00am Location: Bldg. 50B, Room 4205 Speaker: Lin Lin Princeton University Title: Mathematical studies of electronic structure theory Abstract: Electronic structure theory describes the distribution of electrons in molecules and in solids. Among different formalisms in electronic structure theory, the Kohn-Sham density functional theory (KSDFT) is by far the most widely used and practical approach that achieves the best compromise between accuracy and efficiency. The computational complexity of a standard implementation of KSDFT is $O(N3)$, where N is the number of electrons. The cubic scaling with respect to N limits the application of KSDFT to systems with at most tens of thousands of atoms. Reducing the computational complexity of KSDFT requires an in-depth study of the mathematical structure of the KSDFT. In this talk I will present novel techniques that allow us to reduce the $O(N3)$ scaling to $O(N1.5)$ for two-dimensional surface systems and $O(N2)$ scaling for three-dimensional bulk systems. These techniques are quite general from a numerical analysis point of view. They have applications beyond the study of electronic structure theory. Host of Seminar: Ann Almgren Computing Sciences Seminars Web Site: http://crd.lbl.gov/SCG/CSSeminars. For additional information, such as site access or directions to the conference room, please contact CSSeminars-Help@hpcrd.lbl.gov. Web Contact: CSSeminars-Help@hpcrd.lbl.gov -------------- next part -------------- An HTML attachment was scrubbed... URL: http://csmr.ca.sandia.gov/pipermail/banana/attachments/20110118/61db0b6c/attachment-0001.html From egng at lbl.gov Fri Jan 14 15:38:16 2011 From: egng at lbl.gov (Rachel Lance) Date: Tue Jan 18 18:25:22 2011 Subject: [BANANA] Berkeley Lab - Computing Sciences Seminar - Thursday, 1/20/2011, 10:00am Message-ID: <4D30DE68.3070103@lbl.gov> Berkeley Lab - Computing Sciences Seminar Date: Thursday, January 20, 2011 Time: 10:00am - 11:00am Location: Bldg. 50F, Room 1647 Speaker: Abhinav Bhatele University of Illinois at Urbana-Champaign Title: Automating Topology Aware Mapping for Supercomputers Abstract: Petascale machines with hundreds of thousands of cores are being built. These machines have varying interconnect topologies and large network diameters. Computation is cheap and communication on the network is becoming the bottleneck for scaling of parallel applications. Network contention, specifically, is becoming an increasingly important factor affecting overall performance. Most parallel applications have a certain communication topology. Mapping of tasks in a parallel application based on their communication graph, to the physical processors on a machine can potentially lead to performance improvements. Placement of communicating tasks on nearby physical processors can minimize the distance traveled by messages and reduce the chances of contention. Performance improvements through topology aware placement for applications such as NAMD and OpenAtom are used to motivate this work. Building on these ideas, I will present algorithms and techniques for automatic mapping of parallel applications to relieve the application developers of this burden. The hop-bytes metric is proposed for the evaluation of mapping algorithms as a better metric than the previously used maximum dilation metric. The main focus of this work is on developing topology aware mapping algorithms for parallel applications with regular and irregular communication patterns. The automatic mapping framework is a suite of such algorithms with capabilities to choose the best mapping for a problem with a given communication graph. More details on my research available at: http://charm.cs.illinois.edu/~bhatele/phd/ Bio: Abhinav received a B. Tech. degree in Computer Science and Engineering from I.I.T. Kanpur (INDIA) in May 2005 and M. S. and Ph. D. degrees in Computer Science from the University of Illinois at Urbana-Champaign in 2007 and 2010 respectively. He is a postdoctoral research associate working with Professors Kale and Gropp at Illinois. His research is centered around topology aware mapping and load balancing for parallel applications. He is also interested in performance analysis and optimization of parallel applications and studying algorithmic feasibility to exascale. Abhinav is an ACM/IEEE George Michael HPC Fellow (2009). He received the David J. Kuck Outstanding MS Thesis Award in 2009, a Distinguished Paper Award at Euro-Par 2009 for his mapping work and the David J. Kuck Outstanding PhD Thesis Award for 2011. Host of Seminar: Alexander Sim Computing Sciences Seminars Web Site: http://crd.lbl.gov/SCG/CSSeminars. For additional information, such as site access or directions to the conference room, please contact CSSeminars-Help@hpcrd.lbl.gov. Web Contact: CSSeminars-Help@hpcrd.lbl.gov -------------- next part -------------- An HTML attachment was scrubbed... URL: http://csmr.ca.sandia.gov/pipermail/banana/attachments/20110118/de6e0c38/attachment.html From saunders at stanford.edu Thu Jan 20 09:27:40 2011 From: saunders at stanford.edu (Michael Saunders) Date: Thu Jan 20 09:29:30 2011 Subject: [BANANA] [icme-linear-algebra-opt LA/Opt seminar Thursday Jan 20 (Beresford Parlett) In-Reply-To: References: Message-ID: Reminder: LA/Opt seminar this afternoon 4:15pm Special note: Please join us for tea and coffee 3:30pm in iCME kitchen to welcome Prof Parlett Linear Algebra and Optimization Seminar (CME 510) iCME, Stanford University http://icme.stanford.edu/seminars/seminarInfo.php?seminar_id=2 4:15pm Thursday Jan 20, 2011 Y2E2 101 http://campus-map.stanford.edu/index.cfm?ID=04-070 Professor Beresford Parlett CS Faculty Emeritus, UC Berkeley parlett@math.berkeley.edu http://www.eecs.berkeley.edu/Pubs/Faculty/parlett.html Localization in eigenvectors of tridiagonal matrices In applications to Quantum Chemistry and to Structural Engineering the tridiagonal matrices that are generated in intermediate stages have the property that some of the important eigenvectors have small support. In other words, the interesting part of the eigenvector is determined by a small part of the matrix. In particular, the vectors from an invariant subspace associated with a cluster of close eigenvalues are sometimes determined from the same small submatrix. Our goal is to determine such a submatrix in advance so that the cost of calculation is greatly reduced. We also want to discover quickly when there is no localization. The talk will present both theory and practice. Forthcoming: Thu Jan 27 Edmond Chiu Aero-Astro, Stanford edmondc@stanford.edu Thu Feb 03 Rob Tibshirani HRP and Statistics, Stanford http://www-stat.stanford.edu/~tibs Thu Mar 10 Jim Lambers Math, Southern Mississippi http://www.math.usm.edu/lambers/ _______________________________________________ icme-linear-algebra-optimization mailing list icme-linear-algebra-optimization@lists.stanford.edu https://mailman.stanford.edu/mailman/listinfo/icme-linear-algebra-optimization -------------- next part -------------- An HTML attachment was scrubbed... URL: http://csmr.ca.sandia.gov/pipermail/banana/attachments/20110120/b2c47213/attachment.html From mgu at math.berkeley.edu Thu Jan 20 23:14:07 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Thu Jan 20 23:18:28 2011 Subject: [BANANA] LAPACK seminar on Jan. 26, 2011 Message-ID: <201101210714.p0L7E7V8024652@panda.math.berkeley.edu> Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm We kick off our semester with a joint Scientific Computing/Applied Math Seminar featuring Prof. Rokhlin of Yale University. Note that for Jan. 26 only, we will meet from 4:00-5:00PM in 939 Evans Hall. Have a great spring semester!! Date: Jan. 26 Speaker: Vladimir Rokhlin, Yale University Date: Feb. 2 Speaker: Dr. David Gleich Title: Fast Katz and Commuters - Quadrature Rules and Sparse Linear Solvers for Link Prediction Heuristics From mgu at math.berkeley.edu Mon Jan 24 21:04:53 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Mon Jan 24 21:09:02 2011 Subject: [BANANA] Reminder: LAPACK seminar on Jan. 26, 2011 In-Reply-To: <201101210714.p0L7E7V8024652@panda.math.berkeley.edu> References: <201101210714.p0L7E7V8024652@panda.math.berkeley.edu> Message-ID: Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm We kick off our semester with a joint Scientific Computing/Applied Math Seminar featuring Prof. Rokhlin of Yale University. Note that for Jan. 26 only, we will meet from 4:00-5:00PM in 939 Evans Hall. Date: Jan. 26 Speaker: Vladimir Rokhlin, Yale University Title: Accurate randomized algorithms of numerical analysis Abstract: See attached. Date: Feb. 2 Speaker: Dr. David Gleich Title: Fast Katz and Commuters - Quadrature Rules and Sparse Linear Solvers for Link Prediction Heuristics -------------- next part -------------- Title: Accurate Randomized Algorithms of Numerical Analysis Vladimir Rokhlin, Yale University Abstact: Randomized algorithms are ubiqutous in computer science and computer engineering. Many problems that are intractable when viewed deterministic ally canbeeffectively solved withprobabilistictechniques. Perhapsthemost important aspect of most randomized procedures in current use is the fact thattheyproducethe correct result with(practically speaking) 100% reliabili ty, and with(essentially) machineprecision. Historically, randomized techniques have been less popular in numerical analysis. Most of them trade accuracy for speed, and in many numerical environments one does not want to add yet another source of inaccuracy to the calculation that is already sufficiently inaccurate. One could say that in numerical analysis probabilistic methods are an approach of last resort. I will discuss several probabilistic algorithms of numerical linear algebra that are never less accurate than their deterministic counterparts, and in fact tend to produce better accuracy. In many situations, the new schemes have lowerCPU time requirements than existing methods,both asymptotically and intermsofactual timings. Iwillillustratetheapproach with several numerical examples, and discuss possible extensions. From saunders at stanford.edu Tue Jan 25 09:54:11 2011 From: saunders at stanford.edu (Michael Saunders) Date: Tue Jan 25 09:56:09 2011 Subject: [BANANA] LA/Opt seminar Thursday Jan 27 (Edmond Chiu) Message-ID: Linear Algebra and Optimization Seminar (CME 510) iCME, Stanford University http://icme.stanford.edu/seminars/seminars.php 4:15pm Thursday Jan 27, 2011 Y2E2 101 http://campus-map.stanford.edu/index.cfm?ID=04-070 Edmond Chiu Dept of Aeronautics and Astronautics, Stanford University edmondc@stanford.edu A Conservative Meshless Method through Quadratic Programming The seminar will feature a novel conservative meshless method we have recently developed for numerically solving general conservation laws. The method is based on approximations of function derivatives using pointwise function values weighted by meshless coefficients satisfying a set of reciprocity and polynomial consistency conditions. These conditions translate to an underdetermined system of linearly dependent constraints. We use QR and singular value decompositions to enforce compatibility of the constraints, and solve a minimum-norm quadratic programming problem to obtain the meshless coefficients. The resulting coefficients minimize an upper bound of a representation of the global discretization error of the meshless derivative. The presentation will include the detailed formulation and the coefficient generation procedure of the method, and also some of the promising results from solving the linear advection equation and the Euler equations for fluid dynamics using the new method. Forthcoming: Thu Feb 03 Rob Tibshirani HRP and Statistics, Stanford http://www-stat.stanford.edu/~tibs Thu Feb 10 Zaiwen Wen IPAM, UCLA http://math.sjtu.edu.cn/faculty/zw2109 Thu Mar 10 Jim Lambers Math, Southern Mississippi http://www.math.usm.edu/lambers/ From saunders at stanford.edu Thu Jan 27 10:32:27 2011 From: saunders at stanford.edu (Michael Saunders) Date: Thu Jan 27 10:34:20 2011 Subject: [BANANA] LA/Opt seminar TODAY (Edmond Chiu) Message-ID: Reminder: LA/Opt seminar this afternoon Linear Algebra and Optimization Seminar (CME 510) iCME, Stanford University http://icme.stanford.edu/seminars/seminars.php 4:15pm Thursday Jan 27, 2011 Y2E2 101 http://campus-map.stanford.edu/index.cfm?ID=04-070 Edmond Chiu Dept of Aeronautics and Astronautics, Stanford University edmondc@stanford.edu A Conservative Meshless Method through Quadratic Programming The seminar will feature a novel conservative meshless method we have recently developed for numerically solving general conservation laws. The method is based on approximations of function derivatives using pointwise function values weighted by meshless coefficients satisfying a set of reciprocity and polynomial consistency conditions. These conditions translate to an underdetermined system of linearly dependent constraints. We use QR and singular value decompositions to enforce compatibility of the constraints, and solve a minimum-norm quadratic programming problem to obtain the meshless coefficients. The resulting coefficients minimize an upper bound of a representation of the global discretization error of the meshless derivative. The presentation will include the detailed formulation and the coefficient generation procedure of the method, and also some of the promising results from solving the linear advection equation and the Euler equations for fluid dynamics using the new method. Forthcoming: Thu Feb 03 Rob Tibshirani HRP and Statistics, Stanford Thu Feb 10 Zaiwen Wen IPAM, UCLA Thu Mar 10 Jim Lambers Math, Southern Mississippi From mgu at math.berkeley.edu Thu Jan 27 16:09:18 2011 From: mgu at math.berkeley.edu (Ming Gu) Date: Thu Jan 27 16:13:33 2011 Subject: [BANANA] LAPACK seminar on Feb. 2, 2011 Message-ID: <201101280009.p0S09IeT026806@panda.math.berkeley.edu> Math 290, Section 29, CS 298, Section 6, Spring 2011 (Matrix Computations and Scientific Computing) We meet WEDNESDAYS 11:10 - noon in Room 380 Soda Hall, Berkeley campus. The coordinators are Profs. J. Demmel (demmel@cs.berkeley.edu) and M. Gu (mgu@math.berkeley.edu). The program will be a mixture of research talks and tutorials. The tutorials will provide a partial sequel to Math 221. For the schedule and other details about the seminar, please see math.berkeley.edu/~mgu/LAPACKSeminar.htm Date: Feb. 2 Speaker:Dr. David Gleich Title: Fast Katz and Commuters - Quadrature Rules and Sparse Linear Solvers for Link Prediction Heuristics Abstract: Motivated by social network data mining problems such as link prediction and collaborative filtering, significant research effort has been devoted to computing topological measures including the Katz score and the commute time. Existing approaches approximate all pairwise relationships simultaneously. We are interested in computing he score for a single pair of nodes; e top-k nodes with the best scores from a given source node. For the pairwise problem, we introduce an iterative algorithm that computes upper and lower bounds for the measures we seek. This algorithm exploits a relationship between the Lanczos process and a quadrature rule. For the top-k problem, we propose an algorithm that only accesses a small portion of the graph, similar to algorithms used in personalized PageRank computing. To test scalability and accuracy, we experiment with three real-world networks and find that our algorithms run in milliseconds to seconds without any preprocessing. Date: Feb. 9 Speaker: Jim Demmel