Search NASA⌕ Search

SEARCH · Search NASA

Results for “LINEAR SCALING ALGORITHMS”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 91 records · Page 5

Variational quantum algorithms

Applications such as simulating complicated quantum systems or solving large-scale linear algebra problems are very challenging for classical computers, owing to the extremely high computational cost. Quantum computers promise a solution, although fault-tolerant quantum computers will probably not be available in the near future. Current quantum devices have serious constraints, including limited numbers of qubits and noise processes that limit circuit depth. Variational quantum algorithms (VQAs), which use a classical optimizer to train a parameterized quantum circuit, have emerged as a leading strategy to address these constraints. VQAs have now been proposed for essentially all applications that researchers have envisaged for quantum computers, and they appear to be the best hope for obtaining quantum advantage. Nevertheless, challenges remain, including the trainability, accuracy and efficiency of VQAs. Here we overview the field of VQAs, discuss strategies to overcome their challenges and finally, we highlight the exciting prospects for using them to obtain quantum advantage.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Potential quantum advantage for simulation of fluid dynamics

Numerical simulation of turbulent fluid dynamics needs to either parametrize turbulence—which introduces large uncertainties—or explicitly resolve the smallest scales—which is prohibitively expensive. Here, we provide evidence through analytic bounds and numerical studies that a potential quantum speedup can be achieved to simulate fluid dynamics using quantum computing. Specifically, we provide a lattice Boltzmann formulation of fluid dynamics for which we give evidence that low-order Carleman linearization is much more accurate than previously believed for these systems. This is achieved via a combination of reformulating the Navier-Stokes nonlinearity (u·$\triangledown$u) to lattice-Boltzmann nonlinearity (u 2 ) and accurately linearizing the dynamical equations, which effectively trades nonlinearity for additional degrees of freedom that add negligible expense in the quantum solver. Based on this, we apply a quantum algorithm for simulating the Carleman-linearized lattice Boltzmann equation and provide evidence that its cost scales logarithmically with system size compared with polynomial scaling in the best known classical algorithms. In this paper, we suggest that a quantum advantage may exist for simulating fluid dynamics, paving the way for simulating nonlinear multiscale transport phenomena in a wide range of disciplines using quantum computing.

42 ENGINEERING↗

Automatic Step Size Selection in Random Walk Metropolis Algorithms

Practitioners of Markov chain Monte Carlo (MCMC) may hesitate to use random walk Metropolis- Hastings algorithms, especially variable-at-a-time algorithms with many parameters, because these algorithms require users to select values of tuning parameters (step sizes). These algorithms perform poorly if the step sizes are set to be too low or too high. We show in this paper that it is not difficult for an algorithm to tune these step sizes automatically to obtain a desired acceptance probability, since the logit of the acceptance probability is very nearly linear in the log of the step size, with known slope coefficient. These ideas work in most applications, including single parameter or block moves on the linear, log, or logit scales. We discuss the implementation of this algorithm in the software package YADAS.

97 MATHEMATICS AND COMPUTING↗

Solving nuclear structure problems with the adaptive variational quantum algorithm

Here we use the Lipkin-Meshkov-Glick (LMG) model and the valence-space nuclear shell model to examine the likely performance of variational quantum eigensolvers in nuclear-structure theory. The LMG model exhibits both a phase transition and spontaneous symmetry breaking at the mean-field level in one of the phases, features that characterize collective dynamics in medium-mass and heavy nuclei. We show that with appropriate modifications, the ADAPT-VQE algorithm, a particularly flexible and accurate variational approach, is not troubled by these complications. We treat up to 12 particles and show that the number of quantum operations needed to approach the ground-state energy scales linearly with the number of qubits. We find similar scaling when the algorithm is applied to the nuclear shell model with realistic interactions in the sd and pf shells. Although most of these simulations contain no noise, we use a noise model from real IBM hardware to show that for the LMG model with four particles, weak noise has no effect on the efficiency of the algorithm.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

Adaptive Neuron Apoptosis for Accelerating Deep Learning on Large Scale Systems

Machine Learning and Data Mining (MLDM) algorithms are becoming ubiquitous in {\em model learning} from the large volume of data generated using simulations, experiments and handheld devices. Deep Learning algorithms -- a class of MLDM algorithms -- are applied for automatic feature extraction, and learning non-linear models for unsupervised and supervised algorithms. Naturally, several libraries which support large scale Deep Learning -- such as TensorFlow and Caffe -- have become popular. In this paper, we present novel techniques to accelerate the convergence of Deep Learning algorithms by conducting low overhead removal of redundant neurons -- {\em apoptosis} of neurons -- which do not contribute to model learning, during the training phase itself. We provide in-depth theoretical underpinnings of our heuristics (bounding accuracy loss and handling apoptosis of several neuron types), and present the methods to conduct adaptive neuron apoptosis. We implement our proposed heuristics with the recently introduced TensorFlow and using its recently proposed extension with MPI. Our performance evaluation on two difference clusters -- one connected with Intel Haswell multi-core systems, and other with nVIDIA GPUs -- using InfiniBand, indicates the efficacy of the proposed heuristics and implementations. Specifically, we are able to improve the training time for several datasets by 2-3x, while reducing the number of parameters by 30x (4-5x on average) on datasets such as ImageNet classification. For the Higgs Boson dataset, our implementation improves the accuracy (measured by Area Under Curve (AUC)) for classification from 0.88/1 to 0.94/1, while reducing the number of parameters by 3x in comparison to existing literature, while achieving a 2.44x speedup in comparison to the default (no apoptosis) algorithm.

Siegel, Charles M.↗

CLUE: A Fast Parallel Clustering Algorithm for High Granularity Calorimeters in High-Energy Physics

One of the challenges of high granularity calorimeters, such as that to be built to cover the endcap region in the CMS Phase-2 Upgrade for HL-LHC, is that the large number of channels causes a surge in the computing load when clustering numerous digitized energy deposits (hits) in the reconstruction stage. In this article, we propose a fast and fully parallelizable density-based clustering algorithm, optimized for high-occupancy scenarios, where the number of clusters is much larger than the average number of hits in a cluster. The algorithm uses a grid spatial index for fast querying of neighbors and its timing scales linearly with the number of hits within the range considered. We also show a comparison of the performance on CPU and GPU implementations, demonstrating the power of algorithmic parallelization in the coming era of heterogeneous computing in high-energy physics.

Rovere, Marco↗

Linear Actuator Has Long Stroke and High Resolution

The term precision linear actuator, direct drive ( PLADD ) refers to a robust linear actuator designed to be capable of repeatedly performing, over a lifetime of the order of 5 to 10 years, positioning maneuvers that include, variously, submicron increments or slews of the order of a centimeter. The PLADD is capable of both long stroke (120 mm) and high resolution (repeatable increments of 20 nm). Unlike precise linear actuators of prior design, the PLADD contains no gears, levers, or hydraulic converters. The PLADD, now at the prototype stage of development, is intended for original use as a coarse-positioning actuator in a spaceborne interferometer. The PLADD could also be adapted to terrestrial applications in which there are requirements for long stroke and high resolution: potential applications include medical imaging and fabrication of semiconductor devices. The PLADD (see figure) includes a commercially available ball-screw actuator driven directly by a commercially available three-phase brushless DC motor. The ball-screw actuator comprises a spring-preloaded ball nut on a ball screw that is restrained against rotation as described below. The motor is coupled directly (that is, without an intervening gear train) to a drive link that, in turn, is coupled to the ball nut. By eliminating the gear train, the direct-drive design eliminates the complexity, backlash, and potential for misalignment associated with a gear train. To prevent inadvertent movement, there is a brake that includes flexured levers compressed against the drive link by preload springs. This is a power-off brake: There are also piezoelectric stacks that can be activated to oppose the springs and push the levers away from the drive link. Hence, power must be applied to the piezoelectric stacks to release the drive link from braking. To help ensure long operational life, all of the mechanical drive components are immersed in an oil bath within hermetically sealed bellows. The outer end of the bellows holds the outer end of the ball screw, thereby preventing rotation of the ball screw. Positioning is controlled by an electronic control system that includes digital and analog subsystems that interact with the motor and brake and with two sensor/encoder units: a Hall-effect-sensor rotation encoder and a linear glass-scale encoder. This system implements a proportional + integral + derivative control algorithm that results in variation of voltage commands to each of the three pairs of windings of the brushless DC motor. In one of two alternative control modes, the voltages are applied to the windings in a trapezoidal commutation scheme on the basis of timing signals obtained from the Hall-effect sensors; this scheme yields relatively coarse positioning - 24 steps per motor revolution. The second control mode involves a sinusoidal commutation scheme in which the output of the linear glass-scale encoder is transposed to rotational increments to yield much finer position feedback - more than 400,000 steps per revolution.

Cook, Brant T.↗

An inertia-free filter line-search algorithm for large-scale nonlinear programming

We present a filter line-search algorithm that does not require inertia information of the linear system. This feature enables the use of a wide range of linear algebra strategies and libraries, which is essential to tackle large-scale problems on modern computing architectures. The proposed approach performs curvature tests along the search step to detect negative curvature and to trigger convexification. We prove that the approach is globally convergent and we implement the approach within a parallel interior-point framework to solve large-scale and highly nonlinear problems. Our numerical tests demonstrate that the inertia-free approach is as efficient as inertia detection via symmetric indefinite factorizations. We also demonstrate that the inertia-free approach can lead to reductions in solution time because it reduces the amount of convexification needed.

Chiang, Nai-Yuan↗

Stochastic density functional theory

Linear-scaling implementations of density functional theory (DFT) reach their intended efficiency regime only when applied to systems having a physical size larger than the range of their Kohn–Sham density matrix (DM). This causes a problem since many types of large systems of interest have a rather broad DM range and are therefore not amenable to analysis using DFT methods. For this reason, the recently proposed stochastic DFT (sDFT), avoiding exhaustive DM evaluations, is emerging as an attractive alternative linear-scaling approach. This review develops a general formulation of sDFT in terms of a (non)orthogonal basis representation and offers an analysis of the statistical errors (SEs) involved in the calculation. Using a new Gaussian-type basis-set implementation of sDFT, applied to water clusters and silicon nanocrystals, it demonstrates and explains how the standard deviation and the bias depend on the sampling rate and the system size in various types of calculations. We also develop a basis-set embedded-fragments theory, demonstrating its utility for reducing the SEs for energy, density of states and nuclear force calculations. Finally, we discuss the algorithmic complexity of sDFT, showing it has CPU wall-time linear-scaling. Lastly, the method parallelizes well over distributed processors with good scalability and therefore may find use in the upcoming exascale computing architectures.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Contact Multigraph Routing: Overview and Implementation

In Delay Tolerant Networking (DTN), the standard routing algorithm used to navigate time-varying networks has been Contact Graph Routing (CGR). In CGR, a globally distributed list of contacts, periods during which two DTN nodes may communicate, is used to construct a contact graph, in which contacts are vertices. A version of Dijkstra’s algorithm can then be used to find paths through this model of the timevarying network. However, since contact graphs may be large compared to the network, potentially growing with the square of the number of network nodes and linearly with the time interval represented, the resulting algorithm does not scale well with the size of the network or time. Any improvement to the routing algorithm will bring significant returns to scale. In a previous paper, we briefly introduced an alternative to the contact graph model for routing. This alternative model is based on a multigraph (a graph in which there may be multiple edges between a pair of vertices) where vertices represent network nodes instead of contacts. A version of Dijkstra’s algorithm in these multigraph models reduces the time needed to perform the same routing computations done in the existing CGR algorithm. Moreover, a modified version of Yen’s algorithm for multigraphs is included. Our variation of CGR, which we call Contact Multigraph Routing (CMR), provides an in-line replacement for the previously used pathfinding algorithms. This paper describes an implementation created based on the CMR approach, and experimental comparisons to traditional CGR are given. In addition, we explore some additional modifications to the routing pipeline traditionally assumed in CGR. These modifications range from the theoretical to the practical in terms of size and scope. We step forward our understanding of sheaftheoretic networking and describe how to model the routing pipeline using sheaves. We detail some enhanced route selection criteria that addresses some of the added complexity of DTNbased systems. We also include a future works section on future improvements and implementations that would be of service to the broader DTN community.

contact graph routing↗

Simple, direct and efficient multi-way spectral clustering

Abstract We present a new algorithm for spectral clustering based on a column-pivoted QR factorization that may be directly used for cluster assignment or to provide an initial guess for k-means. Our algorithm is simple to implement, direct and requires no initial guess. Furthermore, it scales linearly in the number of nodes of the graph and a randomized variant provides significant computational gains. Provided the subspace spanned by the eigenvectors used for clustering contains a basis that resembles the set of indicator vectors on the clusters, we prove that both our deterministic and randomized algorithms recover a basis close to the indicators in Frobenius norm. We also experimentally demonstrate that the performance of our algorithm tracks recent information theoretic bounds for exact recovery in the stochastic block model. Finally, we explore the performance of our algorithm when applied to a real-world graph.

Damle, Anil↗

Linear-Scaling Coupled Cluster with Perturbative Triple Excitations: The Divide–Expand–Consolidate CCSD(T) Model

Here, we propose a reformulation of the traditional (T) triples correction to the coupled cluster singles and doubles (CCSD) energy in terms of local Hartree–Fock (HF) orbitals such that its structural form aligns with our recently developed linear-scaling divide–expand–consolidate (DEC) coupled cluster family of local correlation methods. In a DEC-CCSD(T) calculation, a basis of local occupied and virtual HF orbitals is used to partition the correlated calculation on the full system into a number of independent atomic fragment and pair fragment calculations, each performed within a truncated set of the complete orbital space. In return, this leads to a massively parallel algorithm for the evaluation of the DEC-CCSD(T) correlation energy, which formally scales linearly with the size of the full system and has a tunable precision with respect to a conventional CCSD(T) calculation via a single energy-based input threshold. The theoretical developments are supported by proof of concept DEC-CCSD(T) calculations on a series of medium-sized molecular systems.

Algorithms↗

Final Report: Towards Optimal Petascale Simulations (TOPS), ER25785

Multiscale, multirate scientific and engineering applications in the SciDAC portfolio possess resolution requirements that are practically inexhaustible and demand execution on the highest-capability computers available, which will soon reach the petascale. While the variety of applications is enormous, their needs for mathematical software infrastructure are surprisingly coincident; moreover the chief bottleneck is often the solver. At their current scalability limits, many applications spend a vast majority of their operations in solvers, due to solver algorithmic complexity that is superlinear in the problem size, whereas other phases scale linearly. Furthermore, the solver may be the phase of the simulation with the poorest parallel scalability, due to intrinsic global dependencies. This project brings together the providers of some of the world's most widely distributed, freely available, scalable solver software and focuses them on relieving this bottleneck for many specific applications within SciDAC, which are representative of many others outside. Solver software directly supported under TOPS includes: hypre, PETSc, SUNDIALS, SuperLU, TAO, and Trilinos. Transparent access is also provided to other solver software through the TOPS interface. The primary goals of TOPS are the development, testing, and dissemination of solver software, especially for systems governed by PDEs. Upon discretization, these systems possess mathematical structure that must be exploited for optimal scalability; therefore, application-targeted algorithmic research is included. TOPS software development includes attention to high performance as well as interoperability among the solver components. Support for integration of TOPS solvers into SciDAC applications is also directly supported by this proposal. The role of the UCSD PI in this overall CET, is one of direct interaction between the TOPS software partners and various DOE applications scientists' specifically toward magnetohydrodynamics (MHD) simulations with the Center for Extended Magnetohydrodynamic Modeling (CEMM) SciDAC and Applied Partial Differential Equations Center (APDEC) SciDAC, and toward core-collapse supernova simulations with the previous Terascale Supernova Initiative (TSI) SciDAC and in continued work on INCITE projects headed by Doug Swesty, SUNY Stony Brook. In addition to these DOE applications scientists, the UCSD PI works to bring leading-edge DOE solver technology to applications scientists in cosmology and large-scale galactic structure formation. Unfortunately, the funding for this grant ended after only two years of its five-year duration, in August 2008, due to difficulties at DOE in transferring the grant to the PI's new faculty position at Southern Methodist University. Therefore, this report only describes two years' worth of effort.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Linear-scaling quadruple excitations in local pair natural orbital coupled-cluster theory

Here, we present a fast, asymptotically linear-scaling implementation of the perturbative quadruples energy correction in coupled-cluster theory using local natural orbitals. Our work follows the domain-based local pair natural orbital (DLPNO) approach previously applied to lower levels of excitations in coupled-cluster theory. Our DLPNO-CCSDT(Q) algorithm uses converged doubles and triples amplitudes from a preceding DLPNO-CCSDT computation to compute the quadruples amplitude and energy in the quadruples natural orbital (QNO) basis. We demonstrate the compactness of the QNO space, showing that more than 95% of the (Q) correction can be recovered using relatively loose natural orbital cutoffs, compared to the tighter cutoffs used in pair and triples natural orbitals at lower levels of coupled-cluster theory. We also highlight the accuracy of our algorithm in the computation of relative energies, which yields deviations of sub-kJ mol −1 in relative energy compared to the canonical CCSDT(Q). Timings are conducted on a series of growing linear alkanes (up to 10 carbons and 608 basis functions) and water clusters (up to 49 water molecules and 2842 basis functions) to establish the asymptotic linear-scaling of our DLPNO-(Q) algorithm.

Auxiliary functions↗

Determining the Bayesian optimal sampling strategy in a hierarchical system.

Consider a classic hierarchy tree as a basic model of a 'system-of-systems' network, where each node represents a component system (which may itself consist of a set of sub-systems). For this general composite system, we present a technique for computing the optimal testing strategy, which is based on Bayesian decision analysis. In previous work, we developed a Bayesian approach for computing the distribution of the reliability of a system-of-systems structure that uses test data and prior information. This allows for the determination of both an estimate of the reliability and a quantification of confidence in the estimate. Improving the accuracy of the reliability estimate and increasing the corresponding confidence require the collection of additional data. However, testing all possible sub-systems may not be cost-effective, feasible, or even necessary to achieve an improvement in the reliability estimate. To address this sampling issue, we formulate a Bayesian methodology that systematically determines the optimal sampling strategy under specified constraints and costs that will maximally improve the reliability estimate of the composite system, e.g., by reducing the variance of the reliability distribution. This methodology involves calculating the 'Bayes risk of a decision rule' for each available sampling strategy, where risk quantifies the relative effect that each sampling strategy could have on the reliability estimate. A general numerical algorithm is developed and tested using an example multicomponent system. The results show that the procedure scales linearly with the number of components available for testing.

99 GENERAL AND MISCELLANEOUS↗

VisIO: enabling interactive visualization of ultra-scale, time-series data via high-bandwidth distributed I/O systems

Petascale simulations compute at resolutions ranging into billions of cells and write terabytes of data for visualization and analysis. Interactive visuaUzation of this time series is a desired step before starting a new run. The I/O subsystem and associated network often are a significant impediment to interactive visualization of time-varying data; as they are not configured or provisioned to provide necessary I/O read rates. In this paper, we propose a new I/O library for visualization applications: VisIO. Visualization applications commonly use N-to-N reads within their parallel enabled readers which provides an incentive for a shared-nothing approach to I/O, similar to other data-intensive approaches such as Hadoop. However, unlike other data-intensive applications, visualization requires: (1) interactive performance for large data volumes, (2) compatibility with MPI and POSIX file system semantics for compatibility with existing infrastructure, and (3) use of existing file formats and their stipulated data partitioning rules. VisIO, provides a mechanism for using a non-POSIX distributed file system to provide linear scaling of 110 bandwidth. In addition, we introduce a novel scheduling algorithm that helps to co-locate visualization processes on nodes with the requested data. Testing using VisIO integrated into Para View was conducted using the Hadoop Distributed File System (HDFS) on TACC's Longhorn cluster. A representative dataset, VPIC, across 128 nodes showed a 64.4% read performance improvement compared to the provided Lustre installation. Also tested, was a dataset representing a global ocean salinity simulation that showed a 51.4% improvement in read performance over Lustre when using our VisIO system. VisIO, provides powerful high-performance I/O services to visualization applications, allowing for interactive performance with ultra-scale, time-series data.

97 MATHEMATICS AND COMPUTING↗

Magnetometer bias determination and attitude determination for near-earth spacecraft

A simple linear-regression algorithm is used to determine simultaneously magnetometer biases, misalignments, and scale factor corrections, as well as the dependence of the measured magnetic field on magnetic control systems. This algorithm has been applied to data from the Seasat-1 and the Atmosphere Explorer Mission-1/Heat Capacity Mapping Mission (AEM-1/HCMM) spacecraft. Results show that complete inflight calibration as described here can improve significantly the accuracy of attitude solutions obtained from magnetometer measurements. This report discusses the difficulties involved in obtaining attitude information from three-axis magnetometers, briefly derives the calibration algorithm, and presents numerical results for the Seasat-1 and AEM-1/HCMM spacecraft.

Lerner, G. M.↗

Sanibel Symposium in the Petascale-Exascale Computational Era

The 56 th Sanibel Symposium was held February 14-19 2016 at the King and Prince Hotel, St. Simons Island, GA. It successfully brought quantum chemists and chemical and condensed matter physicists together in presentations, posters, and informal discussions bridging those two communities. The Symposium has had a significant role in preparing generations of quantum theorists. As computational potency and algorithmic sophistication have grown, the Symposium has evolved to emphasize more heavily computationally oriented method development in chemistry and materials physics, including nanoscience, complex molecular phenomena, and even bio-molecular methods and problems. Given this context, the 56 th Sanibel meeting systematically and deliberately had sessions focused on exascale computation. A selection of outstanding theoretical problems that need serious attention was included. Five invited sessions, two contributed sessions (hot topics), and a poster session were organized with the exascale theme. This was a historic milestone in the evolution of the Symposia. Just as years ago linear algebra, perturbation theory, density matrices, and band-structure methods dominated early Sanibel Symposia, the exascale sessions of the 56 th meeting contributed a transformative influence to add structure and strength to the computational physical science community in an unprecedented way. A copy of the full program of the 56 th Symposium is attached. The exascale sessions were Linear Scaling, Non-Adabatic Dynamics, Interpretive Theory and Models, Computation, Software, and Algorithms, and Quantum Monte Carlo. The Symposium Proceedings will be published in Molecular Physics (2017). Note that the Sanibel proceedings from 2015 and 2014 were published as Molecular Physics vol. 114, issue 3-4 (2016) and vol. 113, issue 3-4 (2015) respectively.

36 MATERIALS SCIENCE↗