A5088379175- Parallel Computing and Optimization Techniques
- Distributed and Parallel Computing Systems
- Fluid Dynamics and Turbulent Flows
- Lattice Boltzmann Simulation Studies
- Computational Fluid Dynamics and Aerodynamics
- Machine Learning in Healthcare
- Advanced Data Storage Technologies
- Scientific Computing and Data Management
- Particle Dynamics in Fluid Flows
- Artificial Intelligence in Healthcare and Education
- Cloud Computing and Resource Management
- Fluid Dynamics and Vibration Analysis
- Machine Learning in Materials Science
- Gas Dynamics and Kinetic Theory
- Explainable Artificial Intelligence (XAI)
- Granular flow and fluidized beds
- Advanced Chemical Physics Studies
- Physics of Superconductivity and Magnetism
- Advanced Numerical Methods in Computational Mathematics
- Interconnection Networks and Systems
- Reservoir Engineering and Simulation Methods
- Model-Driven Software Engineering Techniques
- Logic, programming, and type systems
- Computational Drug Discovery Methods
- Groundwater flow and contamination studies
Los Alamos National Laboratory
2015-2024
Stanford University
2000
Cornell University
1993-2000
Anumericalmethodispresentedthatallowslargeeddysimulation (LES)ofturbulente owsincomplexgeometric cone gurations with moving boundaries and that retains the advantages of solving Navier ‐Stokes equations on e xed orthogonal grids. The boundary conditions are applied independently grid by assigning body forces over surfaces need not coincide coordinate lines. use orthogonal, nondeforming grids simplie es generation, facilitates theimplementation high-order,nondissipativediscretization schemes,...
This article explores the coupling of coarse and fine-grained parallelism for Finite Element simulations based on efficient parallel multigrid solvers.The focus lies both system performance a minimally invasive integration hardware acceleration into an existing software package, requiring no changes to application code.Because their excellent price ratio, we demonstrate viability our approach by using commodity graphics processors (GPUs) as preconditioners.We address issue limited precision...
Current multi-petaflop supercomputers are powerful systems, but present challenges when faced with problems requiring large machine learning workflows. Complex algorithms running at system scale, often different patterns that require disparate software packages and complex data flows cause difficulties in assembling managing experiments on these machines. This paper presents a workflow makes progress scaling ensembles, specifically this first release, ensembles of deep neural networks...
We present a graph-based methodology to reduce the computational cost of obtaining first passage times through sparse fracture networks. derive graph representations generic three-dimensional discrete networks (DFNs) using DFN topology and flow boundary conditions. Subgraphs corresponding union k shortest paths between inflow outflow boundaries are identified transport on their equivalent subnetworks is compared full network. The number included in subgraphs based scaling behavior edges with...
Protein-ligand docking is a computational method for identifying drug leads. The capable of narrowing vast library compounds down to tractable size downstream simulation or experimental testing and widely used in discovery. While there has been progress accelerating scoring with artificial intelligence, few works have bridged these successes back the virtual screening community terms utility forward-looking development. We demonstrate power high-speed ML models by 1 billion molecules under...
The Exascale Computing Project (ECP) is invested in co-design to assure that key applications are ready for exascale computing. Within ECP, the Co-design Center Particle Applications (CoPA) addressing challenges faced by particle-based across four sub-motifs: short-range particle-particle interactions (e.g., those which often dominate molecular dynamics (MD) and smoothed particle hydrodynamics (SPH) methods), long-range electrostatic MD gravitational N-body), particle-in-cell (PIC) methods,...
We show how graph theory can be combined with quantum to calculate the electronic structure of large complex systems. The formalism is general and applicable a broad range methods materials, including challenging systems such as biomolecules. methodology combines well-controlled accuracy, low computational cost, natural low-communication parallelism. This combination addresses substantial shortcomings linear scaling theory, in particular respect quantum-based molecular dynamics simulations.
To address the challenge of performance portability and facilitate implementation electronic structure solvers, we developed basic matrix library (BML) Parallel, Rapid O(N), Graph-based Recursive Electronic Structure Solver (PROGRESS) library. The BML implements linear algebra operations necessary for kernels using a unified user interface various formats (dense sparse) architectures (CPUs GPUs). Focusing on density functional theory tight-binding models, PROGRESS several solvers computing...
We have previously presented an approach to include graphics processing units as co-processors in a parallel Finite Element multigrid solver called FEAST. In this paper we show that the acceleration transfers real applications built on top of FEAST, without any modifications application code. The chosen solid mechanics code is well suited assess practicability our due higher accuracy requirements and more diverse CPU/co-processor interaction. demonstrate detail single precision execution...
We present an algorithm for the calculation of density matrix that insulators scales linearly with system size and parallelizes efficiently on multicore, shared memory platforms small controllable numerical errors. The is based implementation second-order spectral projection (SP2) [ Niklasson, A. M. N. Phys. Rev. B 2002 , 66 155115 ] in sparse algebra ELLPACK-R data format. illustrate performance within self-consistent tight binding theory by total energy calculations gas phase...
Machine learning regression can predict macroscopic fault properties such as shear stress, friction, and time to failure using continuous records of zone acoustic emissions. Here we show that a similar approach is successful event catalogs derived from the data. Our methods are applicable arbitrary scale magnitude completeness. We investigate how machine an catalog laboratory earthquakes performs function find strong model performance requires sufficiently low completeness, below this...
Rapid growth in data, computational methods, and computing power is driving a remarkable revolution what variously termed machine learning (ML), statistical learning, artificial intelligence. In addition to highly visible successes machine-based natural language translation, playing the game Go, self-driving cars, these new technologies also have profound implications for experimental science engineering, as well exascale systems that Department of Energy (DOE) developing support those...
Machine learning in biomedicine is reliant on the availability of large, high-quality data sets. These corpora are used for training statistical or deep learning-based models that can be validated against other sets and ultimately to guide decisions. The quality these an essential component their Thus, identifying inspecting outlier critical evaluating, curating, using biomedical Many techniques available look data, but it not clear how evaluate impact highly complex methods. In this paper,...
A case study is presented that provides computation caching (memoization) through a microservice architecture to high-performance computing (HPC) applications, particularly the ExMatEx proxy application CoEVP (Co-designed Embedded ViscoPlasticity Scale-bridging). represents class of multiscale physics methods in which inexpensive coarse-scale models are combined with expensive fine-scale simulate physical phenomena scalably across multiple time and length scales. Recently, has employed...
The push towards exascale computing has sparked a new set of explorations for providing productive programming environments. While many efforts are focusing on the design and development domain-specific languages (DSLs), few have addressed need fully domain-aware toolchain. Without such domain awareness critical features achieving acceptance adoption, as debugger support, pose long-term risk to overall success DSL approach. In this paper we explore use language extensions implement Scout...
Machine learning algorithms are expected to work side-by-side with humans in decision-making pipelines. Thus, the ability of classifiers make reliable decisions is paramount importance. Deep neural networks (DNNs) represent state-of-the-art models address real-world classification. Although strength activation DNNs often correlated network's confidence, in-depth analyses needed establish whether they well calibrated.
The push towards exascale computing has sparked a new set of explorations for providing productive programming environments. While many efforts are focusing on the design and development domain-specific languages (DSLs), few have addressed need fully domain-aware toolchain. Without such domain awareness critical features achieving acceptance adoption, as debugger support, pose long-term risk to overall success DSL approach. In this paper we explore use language extensions implement Scout...