Search

Monitoring carbon dioxide (CO2) injected and stored in subsurface reservoirs is critical for avoiding failure scenarios and enables real-time optimization of CO2 injection rates. Sequential Bayesian data assimilation (DA) is a statistical method for combining information over time from multiple sources to estimate a hidden state, such as the spread of the subsurface CO2 plume. An example of scalable and efficient sequential Bayesian DA is the ensemble Kalman filter (EnKF). We improve upon existing DA literature in the seismic-CO2 monitoring domain by applying this scalable DA algorithm to a high-dimensional CO2 reservoir using two-phase flow dynamics and time-lapse full waveform seismic data with a realistic surface-seismic survey design. We show more accurate estimates of the CO2 saturation field using the EnKF compared to using either the seismic data or the fluid physics alone. Furthermore, we test a range of values for the EnKF hyperparameters and give guidance on their selection for seismic CO2 reservoir monitoring., Comment: This work has been submitted to the IEEE for possible publication

In this paper, we present a comprehensive analysis of the posterior covariance field in Gaussian processes, with applications to the posterior covariance matrix. The analysis is based on the Gaussian prior covariance but the approach also applies to other covariance kernels. Our geometric analysis reveals how the Gaussian kernel's bandwidth parameter and the spatial distribution of the observations influence the posterior covariance as well as the corresponding covariance matrix, enabling straightforward identification of areas with high or low covariance in magnitude. Drawing inspiration from the a posteriori error estimation techniques in adaptive finite element methods, we also propose several estimators to efficiently measure the absolute posterior covariance field, which can be used for efficient covariance matrix approximation and preconditioning. We conduct a wide range of experiments to illustrate our theoretical findings and their practical applications., Comment: 22 papges

Gaussian Process Regression (GPR) is widely used in statistics and machine learning for prediction tasks requiring uncertainty measures. Its efficacy depends on the appropriate specification of the mean function, covariance kernel function, and associated hyperparameters. Severe misspecifications can lead to inaccurate results and problematic consequences, especially in safety-critical applications. However, a systematic approach to handle these misspecifications is lacking in the literature. In this work, we propose a general framework to address these issues. Firstly, we introduce a flexible two-stage GPR framework that separates mean prediction and uncertainty quantification (UQ) to prevent mean misspecification, which can introduce bias into the model. Secondly, kernel function misspecification is addressed through a novel automatic kernel search algorithm, supported by theoretical analysis, that selects the optimal kernel from a candidate set. Additionally, we propose a subsampling-based warm-start strategy for hyperparameter initialization to improve efficiency and avoid hyperparameter misspecification. With much lower computational cost, our subsampling-based strategy can yield competitive or better performance than training exclusively on the full dataset. Combining all these components, we recommend two GPR methods-exact and scalable-designed to match available computational resources and specific UQ requirements. Extensive evaluation on real-world datasets, including UCI benchmarks and a safety-critical medical case study, demonstrates the robustness and precision of our methods.

Solving a linear system $Ax=b$ is a fundamental scientific computing primitive for which numerous solvers and preconditioners have been developed. These come with parameters whose optimal values depend on the system being solved and are often impossible or too expensive to identify; thus in practice sub-optimal heuristics are used. We consider the common setting in which many related linear systems need to be solved, e.g. during a single numerical simulation. In this scenario, can we sequentially choose parameters that attain a near-optimal overall number of iterations, without extra matrix computations? We answer in the affirmative for Successive Over-Relaxation (SOR), a standard solver whose parameter $\omega$ has a strong impact on its runtime. For this method, we prove that a bandit online learning algorithm--using only the number of iterations as feedback--can select parameters for a sequence of instances such that the overall cost approaches that of the best fixed $\omega$ as the sequence length increases. Furthermore, when given additional structural information, we show that a contextual bandit method asymptotically achieves the performance of the instance-optimal policy, which selects the best $\omega$ for each instance. Our work provides the first learning-theoretic treatment of high-precision linear system solvers and the first end-to-end guarantees for data-driven scientific computing, demonstrating theoretically the potential to speed up numerical methods using well-understood learning algorithms., Comment: ICLR 2024 Spotlight

SPARC is an accurate, efficient, and scalable real-space electronic structure code for performing ab initio Kohn-Sham density functional theory calculations. Version 2.0.0 of the software provides increased efficiency, and includes spin-orbit coupling, dispersion interactions, and advanced semilocal as well as hybrid exchange-correlation functionals, where it outperforms state-of-the-art planewave codes by an order of magnitude and more, with increasing advantages as the number of processors is increased. These new features further expand the range of physical applications amenable to first principles investigation., Comment: 13 pages, 2 figures

The spectrum of a kernel matrix significantly depends on the parameter values of the kernel function used to define the kernel matrix. This makes it challenging to design a preconditioner for a regularized kernel matrix that is robust across different parameter values. This paper proposes the Adaptive Factorized Nystr\"om (AFN) preconditioner. The preconditioner is designed for the case where the rank k of the Nystr\"om approximation is large, i.e., for kernel function parameters that lead to kernel matrices with eigenvalues that decay slowly. AFN deliberately chooses a well-conditioned submatrix to solve with and corrects a Nystr\"om approximation with a factorized sparse approximate matrix inverse. This makes AFN efficient for kernel matrices with large numerical ranks. AFN also adaptively chooses the size of this submatrix to balance accuracy and cost.

We present a GPU-accelerated version of the real-space SPARC electronic structure code for performing Kohn-Sham density functional theory calculations within the local density and generalized gradient approximations. In particular, we develop a modular math kernel based implementation for NVIDIA architectures wherein the computationally expensive operations are carried out on the GPUs, with the remainder of the workload retained on the CPUs. Using representative bulk and slab examples, we show that GPUs enable speedups of up to 6x relative to CPU-only execution, bringing time to solution down to less than 30 seconds for a metallic system with over 14,000 electrons, and enabling significant reductions in computational resources required for a given wall time., Comment: 11 pages, 3 figures

A general, {\em rectangular} kernel matrix may be defined as $K_{ij} = \kappa(x_i,y_j)$ where $\kappa(x,y)$ is a kernel function and where $X=\{x_i\}_{i=1}^m$ and $Y=\{y_i\}_{i=1}^n$ are two sets of points. In this paper, we seek a low-rank approximation to a kernel matrix where the sets of points $X$ and $Y$ are large and are arbitrarily distributed, such as away from each other, ``intermingled'', identical, etc. Such rectangular kernel matrices may arise, for example, in Gaussian process regression where $X$ corresponds to the training data and $Y$ corresponds to the test data. In this case, the points are often high-dimensional. Since the point sets are large, we must exploit the fact that the matrix arises from a kernel function, and avoid forming the matrix, and thus ruling out most algebraic techniques. In particular, we seek methods that can scale linearly or nearly linear with respect to the size of data for a fixed approximation rank. The main idea in this paper is to {\em geometrically} select appropriate subsets of points to construct a low rank approximation. An analysis in this paper guides how this selection should be performed.

Hierarchical matrices provide a powerful representation for significantly reducing the computational complexity associated with dense kernel matrices. For general kernel functions, interpolation-based methods are widely used for the efficient construction of hierarchical matrices. In this paper, we present a fast hierarchical data reduction (HiDR) procedure with $O(n)$ complexity for the memory-efficient construction of hierarchical matrices with nested bases where $n$ is the number of data points. HiDR aims to reduce the given data in a hierarchical way so as to obtain $O(1)$ representations for all nearfield and farfield interactions. Based on HiDR, a linear complexity $\mathcal{H}^2$ matrix construction algorithm is proposed. The use of data-driven methods enables {better efficiency than other general-purpose methods} and flexible computation without accessing the kernel function. Experiments demonstrate significantly improved memory efficiency of the proposed data-driven method compared to interpolation-based methods over a wide range of kernels. Though the method is not optimized for any special kernel, benchmark experiments for the Coulomb kernel show that the proposed general-purpose algorithm offers competitive performance for hierarchical matrix construction compared to several state-of-the-art algorithms for the Coulomb kernel., Comment: 26 pages, 20 figures

This paper presents some of the current challenges in designing deep learning artificial intelligence (AI) and integrating it with traditional high-performance computing (HPC) simulations. We evaluate existing packages for their ability to run deep learning models and applications on large-scale HPC systems efficiently, identify challenges, and propose new asynchronous parallelization and optimization techniques for current large-scale heterogeneous systems and upcoming exascale systems. These developments, along with existing HPC AI software capabilities, have been integrated into MagmaDNN, an open-source HPC deep learning framework. Many deep learning frameworks are targeted at data scientists and fall short in providing quality integration into existing HPC workflows. This paper discusses the necessities of an HPC deep learning framework and how those needs can be provided (e.g., as in MagmaDNN) through a deep integration with existing HPC libraries, such as MAGMA and its modular memory management, MPI, CuBLAS, CuDNN, MKL, and HIP. Advancements are also illustrated through the use of algorithmic enhancements in reduced- and mixed-precision, as well as asynchronous optimization methods. Finally, we present illustrations and potential solutions for enhancing traditional compute- and data-intensive applications at ORNL and UTK with AI. The approaches and future challenges are illustrated in materials science, imaging, and climate applications.

In an iterative approach for solving linear systems with ill-conditioned, symmetric positive definite (SPD) kernel matrices, both fast matrix-vector products and fast preconditioning operations are required. Fast (linear-scaling) matrix-vector products are available by expressing the kernel matrix in an $\mathcal{H}^2$ representation or an equivalent fast multipole method representation. Preconditioning such matrices, however, requires a structured matrix approximation that is more regular than the $\mathcal{H}^2$ representation, such as the hierarchically semiseparable (HSS) matrix representation, which provides fast solve operations. Previously, an algorithm was presented to construct an HSS approximation to an SPD kernel matrix that is guaranteed to be SPD. However, this algorithm has quadratic cost and was only designed for recursive binary partitionings of the points defining the kernel matrix. This paper presents a general algorithm for constructing an SPD HSS approximation. Importantly, the algorithm uses the $\mathcal{H}^2$ representation of the SPD matrix to reduce its computational complexity from quadratic to quasilinear. Numerical experiments illustrate how this SPD HSS approximation performs as a preconditioner for solving linear systems arising from a range of kernel functions.

Decentralized methods are gaining popularity for data-driven models in power systems as they offer significant computational scalability while guaranteeing full data ownership by utility stakeholders. However, decentralized methods still require sharing information about network flow estimates over public facing communication channels, which raises privacy concerns. In this paper we propose a differential privacy driven approach geared towards decentralized formulations of mixed integer operations and maintenance optimization problems that protects network flow estimates. We prove strong privacy guarantees by leveraging the linear relationship between the phase angles and the flow. To address the challenges associated with the mixed integer and dynamic nature of the problem, we introduce an exponential moving average based consensus mechanism to enhance convergence, coupled with a control chart based convergence criteria to improve stability. Our experimental results obtained on the IEEE 118 bus case demonstrate that our privacy preserving approach yields solution qualities on par with benchmark methods without differential privacy. To demonstrate the computational robustness of our method, we conduct experiments using a wide range of noise levels and operational scenarios.

Unit Commitment (UC) is a fundamental problem in power system operations. When coupled with generation maintenance, the joint optimization problem poses significant computational challenges due to coupling constraints linking maintenance and UC decisions. Obviously, these challenges grow with the size of the network. With the introduction of sensors for monitoring generator health and condition-based maintenance(CBM), these challenges have been magnified. ADMM-based decentralized methods have shown promise in solving large-scale UC problems, especially in vertically integrated power systems. However, in their current form, these methods fail to deliver similar computational performance and scalability when considering the joint UC and CBM problem. This paper provides a novel decentralized optimization framework for solving large-scale, joint UC and CBM problems. Our approach relies on the novel use of the subgradient method to temporally decouple various subproblems of the ADMM-based formulation of the joint problem along the maintenance horizon. By effectively utilizing multithreading, our decentralized subgradient approach delivers superior computational performance and eliminates the need to move sensor data thereby alleviating privacy and security concerns. Using experiments on large scale test cases, we show that our framework can provide a speedup of upto 50x as compared to various state of the art benchmarks without compromising on solution quality.

We consider asynchronous versions of the first and second order Richardson methods for solving linear systems of equations. These methods depend on parameters whose values are chosen a priori. We explore the parameter values that can be proven to give convergence of the asynchronous methods. This is the first such analysis for asynchronous second order methods. We find that for the first order method, the optimal parameter value for the synchronous case also gives an asynchronously convergent method. For the second order method, the parameter ranges for which we can prove asynchronous convergence do not contain the optimal parameter values for the synchronous iteration. In practice, however, the asynchronous second order iterations may still converge using the optimal parameter values, or parameter values close to the optimal ones, despite this result. We explore this behavior with a multithreaded parallel implementation of the asynchronous methods., Comment: 15 pages

We present SPARC: Simulation Package for Ab-initio Real-space Calculations. SPARC can perform Kohn-Sham density functional theory calculations for isolated systems such as molecules as well as extended systems such as crystals and surfaces, in both static and dynamic settings. It is straightforward to install/use and highly competitive with state-of-the-art planewave codes, demonstrating comparable performance on a small number of processors and increasing advantages as the number of processors grows. Notably, SPARC brings solution times down to a few seconds for systems with $\mathcal{O}(100-500)$ atoms on large-scale parallel computers, outperforming planewave counterparts by an order of magnitude and more., Comment: 17 pages, 3 figures

With increased reliance on cyber infrastructure, large scale power networks face new challenges owing to computational scalability. In this paper we focus on developing an asynchronous decentralized solution framework for the Unit Commitment(UC) problem for large scale power networks. We exploit the inherent asynchrony in a region based decomposition arising out of imbalance in regional subproblems to boost computational efficiency. A two phase algorithm is proposed that relies on the convex relaxation and privacy preserving valid inequalities in order to deliver algorithmic improvements. Our algorithm employs a novel interleaved binary mechanism that locally switches from the convex subproblem to its binary counterpart based on consistent local convergent behavior. We develop a high performance computing (HPC) oriented software framework that uses Message Passing Interface (MPI) to drive our benchmark studies. Our simulations performed on the IEEE 3012 bus case are benchmarked against the centralized and a state of the art synchronous decentralized method. The results demonstrate that the asynchronous method improves computational efficiency by a significant amount and provides a competitive solution quality rivaling the benchmark methods.

In the iterative solution of dense linear systems from boundary integral equations or systems involving kernel matrices, the main challenges are the expensive matrix-vector multiplication and the storage cost which are usually tackled by hierarchical matrix techniques such as $\mathcal{H}$ and $\mathcal{H}^2$ matrices. However, hierarchical matrices also have a high construction cost that is dominated by the low-rank approximations of the sub-blocks of the kernel matrix. In this paper, an efficient method is proposed to give a low-rank approximation of the kernel matrix block $K(X_0, Y_0)$ in the form of an interpolative decomposition (ID) for a kernel function $K(x,y)$ and two properly located point sets $X_0, Y_0$. The proposed method combines the ID using strong rank-revealing QR (sRRQR), which is purely algebraic, with analytic kernel information to reduce the construction cost of a rank-$r$ approximation from $O(r|X_0||Y_0|)$, for ID using sRRQR alone, to $O(r|X_0|)$ which is not related to $|Y_0|$. Numerical experiments show that $\mathcal{H}^2$ matrix construction with the proposed algorithm only requires a computational cost linear in the matrix dimension.

In constructing the $\mathcal{H}^2$ representation of dense matrices defined by the Laplace kernel, the interpolative decomposition of certain off-diagonal submatrices that dominates the computation can be dramatically accelerated using the concept of a proxy surface. We refer to the computation of such interpolative decompositions as the proxy surface method. We present an error bound for the proxy surface method in the 3D case and thus provide theoretical guidance for the discretization of the proxy surface in the method.

Parallel implementations of linear iterative solvers generally alternate between phases of data exchange and phases of local computation. Increasingly large problem sizes on more heterogeneous systems make load balancing and network layout very challenging tasks. In particular, global communication patterns such as inner products become increasingly limiting at scale. We explore the use of asynchronous communication based on one-sided MPI primitives in a multitude of domain decomposition solvers. In particular, a scalable asynchronous two-level method is presented. We discuss practical issues encountered in the development of a scalable solver and show experimental results obtained on state-of-the-art supercomputer systems that illustrate the benefits of asynchronous solvers in load balanced as well as load imbalanced scenarios. Using the novel method, we can observe speed-ups of up to 4x over its classical synchronous equivalent.

This paper presents an efficient method to perform Structured Matrix Approximation by Separation and Hierarchy (SMASH), when the original dense matrix is associated with a kernel function. Given points in a domain, a tree structure is first constructed based on an adaptive partitioning of the computational domain to facilitate subsequent approximation procedures. In contrast to existing schemes based on either analytic or purely algebraic approximations, SMASH takes advantage of both approaches and greatly improves the efficiency. The algorithm follows a bottom-up traversal of the tree and is able to perform the operations associated with each node on the same level in parallel. A strong rank-revealing factorization is applied to the initial analytic approximation in the separation regime so that a special structure is incorporated into the final nested bases. As a consequence, the storage is significantly reduced on one hand and a hierarchy of the original grid is constructed on the other hand. Due to this hierarchy, nested bases at upper levels can be computed in a similar way as the leaf level operations but on coarser grids. The main advantages of SMASH include its simplicity of implementation, its flexibility to construct various hierarchical rank structures and a low storage cost. Rigorous error analysis and complexity analysis are conducted to show that this scheme is fast and stable. The efficiency and robustness of SMASH are demonstrated through various test problems arising from integral equations, structured matrices, etc.

Phase change memory (PCM) devices are known to reduce in power consumption as the bit volume and contact area of their electrodes are scaled down. Here, we demonstrate two types of low-power PCM devices with lateral graphene ribbon electrodes: one in which the graphene is patterned into narrow nanoribbons and the other where the phase change material is patterned into nanoribbons. The sharp graphene "edge" contacts enable switching with threshold voltages as low as ~3 V, low programming currents (<1 {\mu}A SET, <10 {\mu}A RESET) and ON/OFF ratios >100. Large-scale fabrication with graphene grown by chemical vapor deposition also enables the study of heterogeneous integration and that of variability for such nanomaterials and devices., Comment: submitted to Applied Physics Letters

We study graphene nanoribbon (GNR) interconnects obtained from graphene grown by chemical vapor deposition (CVD). We report low- and high-field electrical measurements over a wide temperature range, from 1.7 to 900 K. Room temperature mobilities range from 100 to 500 cm2/V/s, comparable to GNRs from exfoliated graphene, suggesting that bulk defects or grain boundaries play little role in devices smaller than the CVD graphene crystallite size. At high-field, peak current densities are limited by Joule heating, but a small amount of thermal engineering allows us to reach ~2 x 10^9 A/cm2, the highest reported for nanoscale CVD graphene interconnects. At temperatures below ~5 K, short GNRs act as quantum dots with dimensions comparable to their lengths, highlighting the role of metal contacts in limiting transport. Our study illustrates opportunities for CVD-grown GNRs, while revealing variability and contacts as remaining future challenges., Comment: Nano Letters (2012); supplement available online

Cathodoluminescence (CL) imaging spectroscopy is an important technique to understand resonant behavior of optical nanoantennas. We report high-resolution CL spectroscopy of triangular gold nanoantennas designed with near-vacuum effective index and very small metal-substrate interface. This design helped in addressing issues related to background luminescence and shifting of dipole modes beyond visible spectrum. Spatial and spectral investigations of various plasmonic modes are reported. Out-of-plane dipole modes excited with vertically illuminated electron beam showed high-contrast tip illumination in panchromatic imaging. By tilting the nanostructures during fabrication, in-plane dipole modes of antennas were excited. Finite-difference time-domain simulations for electron and optical excitations of different modes showed excellent agreement with experimental results. Our approach of efficiently exciting antenna modes by using low index substrates is confirmed both with experiments and numerical simulations. This should provide further insights into better understanding of optical antennas for various applications., Comment: To be published in JVST B (accepted, Sep 2010) (15 pages, 6 figures, originally presented at EIPBN 2010)

An important task for Homeland Security is the prediction of threat vulnerabilities, such as through the detection of relationships between seemingly disjoint entities. A structure used for this task is a "semantic graph", also known as a "relational data graph" or an "attributed relational graph". These graphs encode relationships as "typed" links between a pair of "typed" nodes. Indeed, semantic graphs are very similar to semantic networks used in AI. The node and link types are related through an ontology graph (also known as a schema). Furthermore, each node has a set of attributes associated with it (e.g., "age" may be an attribute of a node of type "person"). Unfortunately, the selection of types and attributes for both nodes and links depends on human expertise and is somewhat subjective and even arbitrary. This subjectiveness introduces biases into any algorithm that operates on semantic graphs. Here, we raise some knowledge representation issues for semantic graphs and provide some possible solutions using recently developed ideas in the field of complex networks. In particular, we use the concept of transitivity to evaluate the relevance of individual links in the semantic graph for detecting relationships. We also propose new statistical measures for semantic graphs and illustrate these semantic measures on graphs constructed from movies and terrorism data., Comment: 9 pages, 2 tables, 7 figures

Using as a thermometer the temperature dependent magneto-transport of a two-dimensional electron gas, we find that effective temperature scales with current as $T_{\rm e} \sim I^a$, where $a=0.4 \pm 2\%$ in the {\it Shubnikov de-Haas} regime, and $0.53 \pm 2\%$ in both the {\it integer and fractional} quantum Hall effect. This implies the phonon energy emission rate changes from the expected $P\sim T^5$ to $P\sim T^4$. We explain this, as well as the dramatic enhancement in phonon emission efficiency using a hydrodynamic model., Comment: 4 pages, 2 Postscript figures uuencoded with TeX file uses psfig macro. Submitted to Phys. Rev. Lett