427 Matching Results

Search Results

Advanced search parameters have been applied.

Accelerated Gibbs Sampling for Infinite Sparse Factor Analysis

Description: The Indian Buffet Process (IBP) gives a probabilistic model of sparse binary matrices with an unbounded number of columns. This construct can be used, for example, to model a fixed numer of observed data points (rows) associated with an unknown number of latent features (columns). Markov Chain Monte Carlo (MCMC) methods are often used for IBP inference, and in this technical note, we provide a detailed review of the derivations of collapsed and accelerated Gibbs samplers for the linear-Gaussian infinite latent feature model. We also discuss and explain update equations for hyperparameter resampling in a 'full Bayesian' treatment and present a novel slice sampler capable of extending the accelerated Gibbs sampler to the case of infinite sparse factor analysis by allowing the use of real-valued latent features.
Date: September 12, 2011
Creator: Andrzejewski, D M
Partner: UNT Libraries Government Documents Department

On the selection of dimension reduction techniques for scientific applications

Description: Many dimension reduction methods have been proposed to discover the intrinsic, lower dimensional structure of a high-dimensional dataset. However, determining critical features in datasets that consist of a large number of features is still a challenge. In this paper, through a series of carefully designed experiments on real-world datasets, we investigate the performance of different dimension reduction techniques, ranging from feature subset selection to methods that transform the features into a lower dimensional space. We also discuss methods that calculate the intrinsic dimensionality of a dataset in order to understand the reduced dimension. Using several evaluation strategies, we show how these different methods can provide useful insights into the data. These comparisons enable us to provide guidance to a user on the selection of a technique for their dataset.
Date: February 17, 2012
Creator: Fan, Y J & Kamath, C
Partner: UNT Libraries Government Documents Department

LR: Compact connectivity representation for triangle meshes

Description: We propose LR (Laced Ring) - a simple data structure for representing the connectivity of manifold triangle meshes. LR provides the option to store on average either 1.08 references per triangle or 26.2 bits per triangle. Its construction, from an input mesh that supports constant-time adjacency queries, has linear space and time complexity, and involves ordering most vertices along a nearly-Hamiltonian cycle. LR is best suited for applications that process meshes with fixed connectivity, as any changes to the connectivity require the data structure to be rebuilt. We provide an implementation of the set of standard random-access, constant-time operators for traversing a mesh, and show that LR often saves both space and traversal time over competing representations.
Date: January 28, 2011
Creator: Gurung, T; Luffel, M; Lindstrom, P & Rossignac, J
Partner: UNT Libraries Government Documents Department

The case of the missing supercomputer performance : achieving optimal performance on the 8, 192 processors of ASCI Q

Description: In this paper we describe how we improved the effective performance of ASCI Q, the world's second-fastest supercomputer, to meet our expectations. Using an arsenal of performance-analysis techniques including analytical models, custom microbenchmarks, full applications, and simulators, we succeeded in observing a serious-but previously undetectable-performance problem. We identified the source of the problem, eliminated the problem, and 'closed the loop' by demonstrating improved application performance. We present our methodology and provide insight into performance analysis that is immediately applicable to other large-scale cluster-based supercomputers.
Date: January 1, 2003
Creator: Petrini, F. (Fabrizio); Kerbyson, D. J. (Darren J.) & Pakin, S. D. (Scott D.)
Partner: UNT Libraries Government Documents Department

Normal-reflection image

Description: Common-angle wave-equation migration using the double-square-root is generally less accurate than the common-shot migration because the wavefield continuation equation for thc former involves additional approximations compared to that for the latter. We present a common-angle wave-equation migration that has the same accuracy as common-shot wave-equation migration. An image obtained from common-angle migration is a four- to five-dimensional output volume for 3D cases. We propose a normal-reflection imaging condition for common-angle migration to produce a 3D output volume for 3D migration. The image is closely related to the normal-reflection coefficients at interfaces. This imaging condition will allow amplitude-preserving migration to generate an image with clear physical meaning.
Date: January 1, 2003
Creator: Huang, L. (Lian-Jie) & Fehler, Michael C.
Partner: UNT Libraries Government Documents Department

Sequential dynamical systems with threshold functions.

Description: A sequential dynamical system (SDS) (see [BH+01] and the references therein) consists of an undirected graph G(V,E) where each node {nu} {epsilon} V is associated with a Boolean state (s{sub {nu}}) and a symmetric Boolean function f{sub {nu}} (called the local transition function at {nu}). The inputs to f{sub {nu}} are s{sub {nu}} and the states of all the nodes adjacent to {nu}. In each step of the SDS, the nodes update their state values using their local transition functions in the order specified by a given permutation {pi} of the nodes. A configuration of the SDS is an n-tuple (b{sub 1}, b{sub 2}...,b{sub n}) where n = |V| and b{sub i} {epsilon} {l_brace}0,1{r_brace} is the state value of node {nu}{sub i}. The system starts in a specified initial configuration and each step of the SDS produces a (possibly new) configuration.
Date: January 1, 2001
Creator: Barrett, C. L. (Christopher L.); Hunt, H. B.; Marathe, M. V. (Madhav V.); Ravi, S. S.; Rosenkrantz, D. J. (Daniel J.) & Stearns, R. E. (Richard E.)
Partner: UNT Libraries Government Documents Department

The Web as an API.

Description: As programmers we have worked with many Application Development Interface API development kits. They are well suited for interaction with a particular system. A vast source of information can be made accessible by using the http protocol through the web as an API. This setup has many advantages including the vast knowledge available on setting web servers and services. Also, these tools are available on most hardware and operating system combinations. In this paper I will cover the various types of systems that can be developed this way, their advantages and some drawbacks of this approach. Index Terms--Application Programmer Interface, Distributed applications, Hyper Text Transfer Protocol, Web.
Date: January 1, 2001
Creator: Roman, J. H. (Jorge H.)
Partner: UNT Libraries Government Documents Department

Topological Cacti: Visualizing Contour-based Statistics

Description: Contours, the connected components of level sets, play an important role in understanding the global structure of a scalar field. In particular their nestingbehavior and topology-often represented in form of a contour tree-have been used extensively for visualization and analysis. However, traditional contour trees onlyencode structural properties like number of contours or the nesting of contours, but little quantitative information such as volume or other statistics. Here we use thesegmentation implied by a contour tree to compute a large number of per-contour (interval) based statistics of both the function defining the contour tree as well asother co-located functions. We introduce a new visual metaphor for contour trees, called topological cacti, that extends the traditional toporrery display of acontour tree to display additional quantitative information as width of the cactus trunk and length of its spikes. We apply the new technique to scalar fields ofvarying dimension and different measures to demonstrate the effectiveness of the approach.
Date: May 26, 2011
Creator: Weber, Gunther H.; Bremer, Peer-Timo & Pascucci, Valerio
Partner: UNT Libraries Government Documents Department

Topological Galleries: A High Level User Interface for Topology Controlled Volume Rendering

Description: Existing topological interfaces to volume rendering are limited by their reliance on sophisticated knowledge of topology by the user. We extend previous work by describing topological galleries, an interface for novice users that is based on the design galleries approach. We report three contributions: an interface based on hierarchical thumbnail galleries to display the containment relationships between topologically identifiable features, the use of the pruning hierarchy instead of branch decomposition for contour tree simplification, and drag-and-drop transfer function assignment for individual components. Initial results suggest that this approach suffers from limitations due to rapid drop-off of feature size in the pruning hierarchy. We explore these limitations by providing statistics of feature size as function of depth in the pruning hierarchy of the contour tree.
Date: June 30, 2011
Creator: MacCarthy, Brian; Carr, Hamish & Weber, Gunther H.
Partner: UNT Libraries Government Documents Department

What Scientific Applications can Benefit from Hardware Transactional Memory?

Description: Achieving efficient and correct synchronization of multiple threads is a difficult and error-prone task at small scale and, as we march towards extreme scale computing, will be even more challenging when the resulting application is supposed to utilize millions of cores efficiently. Transactional Memory (TM) is a promising technique to ease the burden on the programmer, but only recently has become available on commercial hardware in the new Blue Gene/Q system and hence the real benefit for realistic applications has not been studied, yet. This paper presents the first performance results of TM embedded into OpenMP on a prototype system of BG/Q and characterizes code properties that will likely lead to benefits when augmented with TM primitives. We first, study the influence of thread count, environment variables and memory layout on TM performance and identify code properties that will yield performance gains with TM. Second, we evaluate the combination of OpenMP with multiple synchronization primitives on top of MPI to determine suitable task to thread ratios per node. Finally, we condense our findings into a set of best practices. These are applied to a Monte Carlo Benchmark and a Smoothed Particle Hydrodynamics method. In both cases an optimized TM version, executed with 64 threads on one node, outperforms a simple TM implementation. MCB with optimized TM yields a speedup of 27.45 over baseline.
Date: June 4, 2012
Creator: Schindewolf, M; Bihari, B; Gyllenhaal, J; Schulz, M; Wang, A & Karl, W
Partner: UNT Libraries Government Documents Department

The Network Completion Problem: Inferring Missing Nodes and Edges in Networks

Description: Network structures, such as social networks, web graphs and networks from systems biology, play important roles in many areas of science and our everyday lives. In order to study the networks one needs to first collect reliable large scale network data. While the social and information networks have become ubiquitous, the challenge of collecting complete network data still persists. Many times the collected network data is incomplete with nodes and edges missing. Commonly, only a part of the network can be observed and we would like to infer the unobserved part of the network. We address this issue by studying the Network Completion Problem: Given a network with missing nodes and edges, can we complete the missing part? We cast the problem in the Expectation Maximization (EM) framework where we use the observed part of the network to fit a model of network structure, and then we estimate the missing part of the network using the model, re-estimate the parameters and so on. We combine the EM with the Kronecker graphs model and design a scalable Metropolized Gibbs sampling approach that allows for the estimation of the model parameters as well as the inference about missing nodes and edges of the network. Experiments on synthetic and several real-world networks show that our approach can effectively recover the network even when about half of the nodes in the network are missing. Our algorithm outperforms not only classical link-prediction approaches but also the state of the art Stochastic block modeling approach. Furthermore, our algorithm easily scales to networks with tens of thousands of nodes.
Date: November 14, 2011
Creator: Kim, M & Leskovec, J
Partner: UNT Libraries Government Documents Department

FY11 Level-2 Milestone 3953: TLCC2 contract awarded

Description: This report documents completion of FY11 L2 milestone No.3953-TLCC2 contract award. This milestone was scheduled for completion on 3/31/11 and was completed on 4/14/11. There is a separate milestone (3856), due at the end of the fiscal year, concerned with installation of the first LLNL SU and early user access. Efforts related to this tri-lab L2 milestone started early in 2010 with the development of tri-lab requirements for the second ASC capacity system procurement. The SOW was then developed along with necessary RFP paperwork and sent to HQ/DOE for their review prior to being released. There was significant delay in getting this step completed which led to this milestone being put at risk for several months. However, once the RFP was approved and released we were able to get the procurement back on track with aggressive proposal response and review timelines.
Date: September 12, 2011
Creator: Carnes, B
Partner: UNT Libraries Government Documents Department

GPU-based Scalable Volumetric Reconstruction for Multi-view Stereo

Description: We present a new scalable volumetric reconstruction algorithm for multi-view stereo using a graphics processing unit (GPU). It is an effectively parallelized GPU algorithm that simultaneously uses a large number of GPU threads, each of which performs voxel carving, in order to integrate depth maps with images from multiple views. Each depth map, triangulated from pair-wise semi-dense correspondences, represents a view-dependent surface of the scene. This algorithm also provides scalability for large-scale scene reconstruction in a high resolution voxel grid by utilizing streaming and parallel computation. The output is a photo-realistic 3D scene model in a volumetric or point-based representation. We demonstrate the effectiveness and the speed of our algorithm with a synthetic scene and real urban/outdoor scenes. Our method can also be integrated with existing multi-view stereo algorithms such as PMVS2 to fill holes or gaps in textureless regions.
Date: September 21, 2011
Creator: Kim, H; Duchaineau, M & Max, N
Partner: UNT Libraries Government Documents Department

Making tensor factorizations robust to non-gaussian noise.

Description: Tensors are multi-way arrays, and the CANDECOMP/PARAFAC (CP) tensor factorization has found application in many different domains. The CP model is typically fit using a least squares objective function, which is a maximum likelihood estimate under the assumption of independent and identically distributed (i.i.d.) Gaussian noise. We demonstrate that this loss function can be highly sensitive to non-Gaussian noise. Therefore, we propose a loss function based on the 1-norm because it can accommodate both Gaussian and grossly non-Gaussian perturbations. We also present an alternating majorization-minimization (MM) algorithm for fitting a CP model using our proposed loss function (CPAL1) and compare its performance to the workhorse algorithm for fitting CP models, CP alternating least squares (CPALS).
Date: March 1, 2011
Creator: Chi, Eric C. (Rice University, Houston, TX) & Kolda, Tamara Gibson
Partner: UNT Libraries Government Documents Department

Asynchronous Checkpoint Migration with MRNet in the Scalable Checkpoint / Restart Library

Description: Applications running on today's supercomputers tolerate failures by periodically saving their state in checkpoint files on stable storage, such as a parallel file system. Although this approach is simple, the overhead of writing the checkpoints can be prohibitive, especially for large-scale jobs. In this paper, we present initial results of an enhancement to our Scalable Checkpoint/Restart Library (SCR). We employ MRNet, a tree-based overlay network library, to transfer checkpoints from the compute nodes to the parallel file system asynchronously. This enhancement increases application efficiency by removing the need for an application to block while checkpoints are transferred to the parallel file system. We show that the integration of SCR with MRNet can reduce the time spent in I/O operations by as much as 15x. However, our experiments exposed new scalability issues with our initial implementation. We discuss the sources of the scalability problems and our plans to address them.
Date: March 20, 2012
Creator: Mohror, K.; Moody, A. & de Supinski, B. R.
Partner: UNT Libraries Government Documents Department

Software Requirements Specification: Multi-scale Epidemiological and Economic Simulation and Analysis (MESA) Scenario Bank

Description: This document builds on the discussion notes from September 21, 2006. It provides a summary of the ideas relating to the scenario bank tables and their associated requirements. Two conceptual groupings were identified for the contents requirements of the scenario bank. The first, called ProjectTemplate, shall consist of <Project, Scenarios, and Miscellaneous Files> groups. The second, ProjectArchive, shall consist of groups of <Project, Scenarios, Results, and Miscellaneous Files>. The figure below illustrates the multiplicity of the associations between the different tables, with color coding used to distinguish between current MESA (brown) and USDA (light green) requirements. Scenario bank tables are shown in black with their general contents specified within the box. The metadata associated with each table is expected to include database key information as well as relevant timestamps. Each File is expected to be a file with an arbitrary format.
Date: November 8, 2006
Creator: Dahlgren, T L; Hazlett, S G; Slone, D M & Smith, S G
Partner: UNT Libraries Government Documents Department

Final Report on Subcontract B591217: Multigrid Methods for Systems of PDEs

Description: Progress is summarized in the following areas of study: (1) Compatible relaxation; (2) Improving aggregation-based MG solver performance - variable cycle; (3) First Order System Least Squares (FOSLS) for LQCD; (4) Auxiliary space preconditioners; (5) Bootstrap algebraic multigrid; and (6) Practical applications of AMG and fast auxiliary space preconditioners.
Date: October 25, 2011
Creator: Xu, J; Brannick, J J & Zikatanov, L
Partner: UNT Libraries Government Documents Department

SC'11 Poster: A Highly Efficient MGPT Implementation for LAMMPS; with Strong Scaling

Description: The MGPT potential has been implemented as a drop in package to the general molecular dynamics code LAMMPS. We implement an improved communication scheme that shrinks the communication layer thickness, and increases the load balancing. This results in unprecedented strong scaling, and speedup continuing beyond 1/8 atom/core. In addition, we have optimized the small matrix linear algebra with generic blocking (for all processors) and specific SIMD intrinsics for vectorization on Intel, AMD, and BlueGene CPUs.
Date: December 7, 2011
Creator: Oppelstrup, T; Stukowski, A & Marian, J
Partner: UNT Libraries Government Documents Department

Rapid Exploitation and Analysis of Documents

Description: Analysts are overwhelmed with information. They have large archives of historical data, both structured and unstructured, and continuous streams of relevant messages and documents that they need to match to current tasks, digest, and incorporate into their analysis. The purpose of the READ project is to develop technologies to make it easier to catalog, classify, and locate relevant information. We approached this task from multiple angles. First, we tackle the issue of processing large quantities of information in reasonable time. Second, we provide mechanisms that allow users to customize their queries based on latent topics exposed from corpus statistics. Third, we assist users in organizing query results, adding localized expert structure over results. Forth, we use word sense disambiguation techniques to increase the precision of matching user generated keyword lists with terms and concepts in the corpus. Fifth, we enhance co-occurrence statistics with latent topic attribution, to aid entity relationship discovery. Finally we quantitatively analyze the quality of three popular latent modeling techniques to examine under which circumstances each is useful.
Date: November 28, 2011
Creator: Buttler, D J; Andrzejewski, D; Stevens, K D; Anastasiu, D & Gao, B
Partner: UNT Libraries Government Documents Department

Combination of evidence in recommendation systems characterized by distance functions

Description: Recommendation systems for different Document Networks (DN) such as the World Wide Web (WWW), Digitnl Libarries, or Scientific Databases, often make use of distance functions extracted from relationships among documents and between documents and semantic tags. For instance, documents In the WWW are related via a hyperlink network, while documents in bibliographic databases are related by citation and collaboration networks.Furthermore, documents can be related to semantic tags such as keywords used to describe their content, The distance functions computed from these relations establish associative networks among items of the DN, and allow recommendation systems to identify relevant associations for iudividoal users. The process of recommendation can be improved by integrating associative data from different sources. Thus we are presented with a problem of combining evidence (about assochaons between items) from different sonrces characterized by distance functions. In this paper we summarize our work on (1) inferring associations from semi-metric distance functions and (2) combining evidence from different (distance) associative DN.
Date: January 1, 2002
Creator: Rocha, L. M. (Luis Mateus)
Partner: UNT Libraries Government Documents Department


Description: Let D be an arbitrary (not necessarily finite) nonempty set, let C be a finite set of constant symbols denoting arbitrary elements of D, and let S and T be an arbitrary finite set of finite-arity relations on D. We denote the problem of determining the satisfiability of finite conjunctions of relations in S applied to variables (to variables and symbols in C) by SAT(S) (by SATc(S).) Here, we study simultaneously the complexity of decision, counting, maximization and approximate maximization problems, for unquantified, quantified and stochastically quantified formulas. We present simple yet general techniques to characterize simultaneously, the complexity or efficient approximability of a number of versions/variants of the problems SAT(S), Q-SAT(S), S-SAT(S),MAX-Q-SAT(S) etc., for many different such D,C ,S, T. These versions/variants include decision, counting, maximization and approximate maximization problems, for unquantified, quantified and stochastically quantified formulas. Our unified approach is based on the following two basic concepts: (i) strongly-local replacements/reductions and (ii) relational/algebraic represent ability. Some of the results extend the earlier results in [Pa85,LMP99,CF+93,CF+94O]u r techniques and results reported here also provide significant steps towards obtaining dichotomy theorems, for a number of the problems above, including the problems MAX-&-SAT( S), and MAX-S-SAT(S). The discovery of such dichotomy theorems, for unquantified formulas, has received significant recent attention in the literature [CF+93,CF+94,Cr95,KSW97]
Date: January 1, 2001
Creator: Hunt, H. B. (Harry B.); Marathe, M. V. (Madhav V.) & Stearns, R. E. (Richard E.)
Partner: UNT Libraries Government Documents Department

Learning with the ratchet algorithm.

Description: This paper presents a randomized algorithm called Ratchet that asymptotically minimizes (with probability 1) functions that satisfy a positive-linear-dependent (PLD) property. We establish the PLD property and a corresponding realization of Ratchet for a generalized loss criterion for both linear machines and linear classifiers. We describe several learning criteria that can be obtained as special cases of this generalized loss criterion, e.g. classification error, classification loss and weighted classification error. We also establish the PLD property and a corresponding realization of Ratchet for the Neyman-Pearson criterion for linear classifiers. Finally we show how, for linear classifiers, the Ratchet algorithm can be derived as a modification of the Pocket algorithm.
Date: January 1, 2003
Creator: Hush, D. R. (Donald R.) & Scovel, James C.
Partner: UNT Libraries Government Documents Department