Search NASA⌕ Search

SEARCH · Search NASA

Results for “parallel parameters optimization”

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 145 records · Page 8
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Direction-optimizing Label Propagation and its Application for Community Detection

Label Propagation is a machine learning algorithm typically used for classification. It has also been found to be an effective method for detecting communities in networks. It has two attractive features as a community detection method: it has nearly linear runtime and it requires no \textit{a priori} community information. We propose a new Direction Optimizing Label Propagation Algorithm (DOLPA) that relies on the use of {\em frontiers} and alternates between label {\em push} and label {\em pull} operations to enhance the performance of LPA. Specifically, DOLPA has parameters for tuning the processing order of vertices in a graph. This reduces the number of edges visited and improves the quality of solution. We apply DOLPA to community detection and present the design and implementation of the algorithm as well as its shared-memory parallelization using OpenMP. Empirically, we evaluate our algorithm using synthetic graphs as well as real-world networks. Compared with the state-of-art \textit{Parallel Label Propagation} algorithm, we achieve at least two times the F-Score while reducing the runtime by 50\% for synthetic graphs with overlapping communities. We also compare DOLPA against the state-of-art parallel implementation of the Louvain method using the same graphs and show that DOLPA achieves about three times the F-Score at 10\% the runtime. On real-world graphs, we get a speedup of up to $10\times$ using 64 threads.

Liu, X↗
DOE OSTI2022

Optimization of the moderators in the STS preliminary design

This report details the results for an optimization of the dimensions of the moderators in the preliminary design of the Spallation Neutron Source Second Target Station (STS). This study uses the optimization algorithms of Dakota and an unstructured mesh model for the moderators in MCNP. More details on the unstructured mesh model and the automated mesh generation can be found in [3]. Parallel to this effort, the same moderator geometries have been optimized using a constructive solid geometry (CSG) MCNP model. More details on this model and its results can be found in [4]. Three optimal designs are selected for each moderator: one that is optimized for maximum peak brightness, one for maximum time-integrated brightness, and one for a combination of peak and time-integrated brightness. The backbone of the optimization work flow is provided by Dakota. For each set of design parameters requested by Dakota, a new solid geometry is automatically built in Creo and SpaceClaim, and subsequently exported to Attila4MC to generate an unstructured mesh geometry for MCNP. After the MCNP calculation is finished, the objective function (e.g., brightness metric) is returned to Dakota. After the new design has been evaluated, a result-file is written, and Dakota proposes the next set of design parameters to be evaluated. The loop continues until a specified convergence criterion has been met. The design parameters of the cylindrical (upper) moderator include the hydrogen radius, the premoderator thickness (top, bottom, radial), the beryllium radius and the horizontal position of the moderator. The crucial design choice is the hydrogen radius. A radius of 62 mm is shown to provide the maximum time-integrated brightness. The maximum peak brightness occurs with a radius of 40 mm. A combined (middle) design, which balances peak and time-integrated brightnesses, is obtained with a hydrogen radius of 50 mm. The premoderator thicknesses and the beryllium radius are slightly larger in the design optimized for time-integrated brightness than in the design optimized for peak brightness. The sensitivity to these two parameters is relatively small close to the optimal configurations. The hydrogen vessel and vacuum vessel wall thicknesses are dependent on the radius of the liquid hydrogen due to structural integrity requirements. The increased wall thicknesses for larger vessels significantly penalize the time-integrated brightness, with the maximum obtainable value reduced by more than 10% relative to earlier studies which used fixed vessel wall thicknesses. The impact of the variable wall thicknesses is much less for the peak brightness and combined brightness designs. The design parameters of the tube (lower) moderator selected for the optimization are the tube length, the annular premoderator thickness, the beryllium radius and the horizontal position of the moderator. The tube length is the crucial parameter and is chosen large (210 mm) and small (125 mm) in the designs optimized for time-integrated and peak brightness respectively. A combined optimal design has a tube length of 170 mm. The premoderator thickness and the beryllium radius are chosen larger in the design optimized for time-integrated brightness.

42 ENGINEERING↗
DOE OSTI2024

District heating utilizing waste heat of a data center: High-temperature heat pumps

Data centers are energy-intensive facilities with substantial low-grade waste heat. High-temperature heat pumps can be critical in boosting the data center’s waste heat for district heating, improving the system-level energy efficiency of data centers, and reducing CO 2 emissions in district heating. This study built thermodynamic models to assess high-temperature heat pumps with six configurations using low global warming potential refrigerants to supply heat up to 120 °C. The heat pump configurations include single-stage or two-stage cycles with advanced components, such as internal heat exchanger, economizer, flash tank, or parallel compressor. The refrigerants include R1234ze(Z), R1233ed(E), R1224yd(Z), R600, and R600a, and R245fa is used as a reference. A case study was carried out to recover the waste heat from the Frontier high-performance computing data center and provide hot water for district heating at the US Department of Energy’s Oak Ridge National Laboratory campus. The optimized performance of high-temperature heat pumps is characterized with various effectiveness of internal heat exchangers, and the operating parameters of economizer or flash tank, as well as their combination. The results show that the configurations of two-stage cycles with internal heat exchanger + flash tank and internal heat exchanger + economizer/parallel-compressor provide the highest coefficient of performance under scenarios of the maximum allowable value and a fixed value (0.3) of the internal heat exchangers’ effectiveness, respectively. R1234ze(Z) and R600a are the most promising refrigerants, considering trade-offs between the coefficient of performance and the volumetric heating capacity. The single-stage cycle with internal heat exchanger + economizer/parallel-compressor using R1234ze(Z) is recommended for utilizing Fronter’s waste heat in district heating. A one mega-watt high-temperature heat pump will reduce 33,100–33,200 metric tons of CO2 emission annually, corresponding to 85.4 %–85.6 % of equivalent CO2 emissions from natural gas boilers. Here, this study provides good guidelines for designing and deploying high-temperature heat pumps to support sustainable data centers and decarbonize district heating in the US.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2017

Aspen-based performance and energy modeling frameworks

Now with the anticipation of exascale architectures, energy consumption is becoming one of the critical design parameters, especially in light of the energy budget of 20-30 Megawatts set by the U.S. Department of Energy. Understanding an application’s execution pattern and its energy footprint is critical to improving application operation on a diverse heterogeneous architecture. Applying application-specific performance optimization can consequently improve energy consumption. However, this approach is only applicable to current systems. As we enter a new era of exascale architectures that is projected to contain more complex memory hierarchies, increased levels of parallelism, heterogeneity in hardware, and complex programming models and techniques, energy and performance management is getting more cumbersome. We therefore propose techniques that predict the energy consumption beforehand or at runtime to enable proactive tuning. Such energy prediction approaches must be generic and adapt themselves at runtime to changing application and hardware configurations. Most existing energy estimation and prediction approaches are empirical in nature and thus tied to current systems. To overcome this limitation, we propose two energy estimation techniques: ACEE (Algorithmic and Categorical Energy Estimation), which uses a combination of analytical and empirical modeling techniques; and AEEM (Aspen’s Embedded Energy Estimation), a system-level analytical energy estimation technique. Both of these models incorporate the Aspen domain specific language for performance modeling. We present the methodologies of these two models and test their accuracy using five proxy applications. We also describe three use cases.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Flexible Stellarator Physics Facility

We propose to build a Flexible Stellarator Physics Facility to explore promising regions of the vast parameter space of disruption-free stellarator solutions for Fusion Pilot Plants (FPPs). The FESAC Long Range Plan recognized the quasi-symmetric stellarator as “the leading US approach to developing disruption-free, low-recirculating-power fusion configurations”. To deliver the ambitious Decadal Vision for Commercial Fusion Energy, we must establish a persuasive stellarator program in parallel to the tokamak one: the stellarator will prove to be a better path to a reactor if theoretical predictions are confirmed and novel optimization techniques and strategies work as desired. Since the release of the Long Range Plan in 2020, stellarators have arguably made the most significant advances of all fusion concepts. Groundbreaking results from W7-X demonstrated low neoclassical transport and the successful operation of the island divertor. Advances in theory and modeling now allow us to minimize turbulent transport, to achieve equilibria with precise quasisymmetry, to reduce neoclassical transport and fast ion loss to levels far below what has been previously achieved, and to minimize the effect of coil manufacturing errors. If realized, these advances will lead to cost-effective stellarator designs with confinement comparable to tokamaks but without the fundamental challenges of disruptions and current drive.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2017

Tuning HDF5 subfiling performance on parallel file systems

Subfiling is a technique used on parallel file systems to reduce locking and contention issues when multiple compute nodes interact with the same storage target node. Subfiling provides a compromise between the single shared file approach that instigates the lock contention problems on parallel file systems and having one file per process, which results in generating a massive and unmanageable number of files. In this paper, we evaluate and tune the performance of recently implemented subfiling feature in HDF5. In specific, we explain the implementation strategy of subfiling feature in HDF5, provide examples of using the feature, and evaluate and tune parallel I/O performance of this feature with parallel file systems of the Cray XC40 system at NERSC (Cori) that include a burst buffer storage and a Lustre disk-based storage. We also evaluate I/O performance on the Cray XC30 system, Edison, at NERSC. Our results show performance benefits of 1.2X to 6X performance advantage with subfiling compared to writing a single shared HDF5 file. We present our exploration of configurations, such as the number of subfiles and the number of Lustre storage targets to storing files, as optimization parameters to obtain superior I/O performance. Based on this exploration, we discuss recommendations for achieving good I/O performance as well as limitations with using the subfiling feature.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Efficient First-Order Algorithms for Large-Scale, Non-Smooth Maximum Entropy Models with Application to Wildfire Science

Maximum entropy (MaxEnt) models are a class of statistical models that use the maximum entropy principle to estimate probability distributions from data. Due to the size of modern data sets, MaxEnt models need efficient optimization algorithms to scale well for big data applications. State-of-the-art algorithms for MaxEnt models, however, were not originally designed to handle big data sets; these algorithms either rely on technical devices that may yield unreliable numerical results, scale poorly, or require smoothness assumptions that many practical MaxEnt models lack. In this paper, we present novel optimization algorithms that overcome the shortcomings of state-of-the-art algorithms for training large-scale, non-smooth MaxEnt models. Our proposed first-order algorithms leverage the Kullback–Leibler divergence to train large-scale and non-smooth MaxEnt models efficiently. For MaxEnt models with discrete probability distribution of n elements built from samples, each containing m features, the stepsize parameter estimation and iterations in our algorithms scale on the order of O(mn) operations and can be trivially parallelized. Moreover, the strong ℓ1 convexity of the Kullback–Leibler divergence allows for larger stepsize parameters, thereby speeding up the convergence rate of our algorithms. To illustrate the efficiency of our novel algorithms, we consider the problem of estimating probabilities of fire occurrences as a function of ecological features in the Western US MTBS-Interagency wildfire data set. Our numerical results show that our algorithms outperform the state of the art by one order of magnitude and yield results that agree with physical models of wildfire occurrence and previous statistical analyses of wildfire drivers.

Physics↗
DOE OSTI2023

Data-Driven Optimization of the Processing Window for 316H Components Fabricated Using Laser Powder Bed Fusion

The Advanced Materials and Manufacturing Technologies Program is focused on accelerating the development and deployment of advanced materials and components fabricated via additive manufacturing with a specific focus on laser powder bed fusion (LPBF). As an initial case study, the program has selected 316H stainless steel (SS) as an initial material around which to develop a code case development strategy. This strategy involves two parallel approaches: (1) an equivalency approach whereby round-robin testing across multiple collaborating laboratories demonstrates repeatability in processing and direct comparisons with conventional wrought 316H material and (2) a revolutionary approach to code qualification combining in situ data collection and high-fidelity modeling to capture, predict, and bound the performance of LPBF 316HSS components. As part of this campaign, this work package has initiated an extensive process optimization campaign across three laboratories, each printing variations of LPBF 316HSS using three different LPBF units (Concept Laser, EOS, and Renishaw). In FY23, ORNL has focused on unique experimental designs spanning wide ranges in energy inputs and turning knobs such as scan speed, laser power, hatch spacing, layer thickness, spot size, scan rotation, and more. On the Concept Laser M2, 72 different combinations of processing variables were investigated with duplicate samples and different powder compositions. In total, 252 samples were printed with combined in situ sensing data. A parallel design of experiments was conducted on the Renishaw AM400 with an additional 390 printed specimens for analysis. All 642 miniature specimens, each with unique features included in each print to capture geometry-related heterogeneity, were subjected to high-throughput x-ray computed tomography (XCT) analysis to enable the downselection of specific processing parameters of interest. Then, using electrical discharge machining (EDM), miniature tensile specimens were extracted for mechanical testing and microscopy investigations. From the analysis performed in FY23, it was found that powder composition drastically affects the resulting microstructure and mechanical performance of 316SS. Specifically, changing from 316L to 316HSS powder results in a wide range of grain sizes with varying degrees of preferred grain orientation, which increases as a function of energy density. It was also found that due to stored heat in thin fin–type features, large microstructural differences can be seen within one part printed with one set of processing parameters. These variations in microstructure features, including grain size, the nanoscale dislocation structure, and grain texture, will all affect the irradiation performance and high-temperature mechanical performance of LPBF 316HSS parts. Two sets of concept laser processing parameters, spanning both refined and columnar grain structures, were scaled to print larger 316H builds for campaign testing (high-temperature creep and irradiation). In addition, at least two optimized processing parameter sets were identified for the Renishaw AM400 for round-robin testing in FY24 with Argonne National Laboratory. Future work includes printing samples using identical parameters identified by partner institutions, providing material for corrosion and high-temperature mechanical testing, and continuing evaluations of heterogeneity in larger printed parts.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Data-Driven Optimization of the Processing Window for 316H Components Fabricated Using Laser Powder Bed Fusion

The Advanced Materials and Manufacturing Technologies Program is focused on accelerating the development and deployment of advanced materials and components fabricated via additive manufacturing with a specific focus on laser powder bed fusion (LPBF). As an initial case study, the program has selected 316H stainless steel (SS) as an initial material around which to develop a code case development strategy. This strategy involves two parallel approaches: (1) an equivalency approach whereby round-robin testing across multiple collaborating laboratories demonstrates repeatability in processing and direct comparisons with conventional wrought 316H material and (2) a revolutionary approach to code qualification combining in situ data collection and high-fidelity modeling to capture, predict, and bound the performance of LPBF 316HSS components. As part of this campaign, this work package has initiated an extensive process optimization campaign across three laboratories, each printing variations of LPBF 316HSS using three different LPBF units (Concept Laser, EOS, and Renishaw). In FY23, ORNL has focused on unique experimental designs spanning wide ranges in energy inputs and turning knobs such as scan speed, laser power, hatch spacing, layer thickness, spot size, scan rotation, and more. On the Concept Laser M2, 72 different combinations of processing variables were investigated with duplicate samples and different powder compositions. In total, 252 samples were printed with combined in situ sensing data. A parallel design of experiments was conducted on the Renishaw AM400 with an additional 390 printed specimens for analysis. All 642 miniature specimens, each with unique features included in each print to capture geometry-related heterogeneity, were subjected to high-throughput x-ray computed tomography (XCT) analysis to enable the downselection of specific processing parameters of interest. Then, using electrical discharge machining (EDM), miniature tensile specimens were extracted for mechanical testing and microscopy investigations. From the analysis performed in FY23, it was found that powder composition drastically affects the resulting microstructure and mechanical performance of 316SS. Specifically, changing from 316L to 316HSS powder results in a wide range of grain sizes with varying degrees of preferred grain orientation, which increases as a function of energy density. It was also found that due to stored heat in thin fin–type features, large microstructural differences can be seen within one part printed with one set of processing parameters. These variations in microstructure features, including grain size, the nanoscale dislocation structure, and grain texture, will all affect the irradiation performance and high-temperature mechanical performance of LPBF 316HSS parts. Two sets of concept laser processing parameters, spanning both refined and columnar grain structures, were scaled to print larger 316H builds for campaign testing (high-temperature creep and irradiation). In addition, at least two optimized processing parameter sets were identified for the Renishaw AM400 for round-robin testing in FY24 with Argonne National Laboratory. Future work includes printing samples using identical parameters identified by partner institutions, providing material for corrosion and high-temperature mechanical testing, and continuing evaluations of heterogeneity in larger printed parts.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Direction-optimizing Label Propagation Framework for Structure Detection in Graphs: Design, Implementation, and Experimental Analysis

Label Propagation is not only a well-known machine learning algorithm for classification but also an effective method for discovering communities and connected components in networks. We propose a new Direction-optimizing Label Propagation Algorithm (DOLPA) framework that enhances the performance of the standard Label Propagation Algorithm (LPA), increases its scalability, and extends its versatility and application scope. As a central feature, the DOLPA framework relies on the use of frontiers and alternates between label push and label pull operations to attain high performance. It is formulated in such a way that the same basic algorithm can be used for finding communities or connected components in graphs by only changing the objective function used. Additionally, DOLPA has parameters for tuning the processing order of vertices in a graph to reduce the number of edges visited and improve the quality of solution obtained. We present the design and implementation of the enhanced algorithm as well as our shared-memory parallelization of it using OpenMP. We also present an extensive experimental evaluation of our implementations using the LFR benchmark and real-world networks drawn from various domains. Compared with an implementation of LPA for community detection available in a widely used network analysis software, we achieve at most five times the F-Score while maintaining similar runtime for graphs with overlapping communities. We also compare DOLPA against an implementation of the Louvain method for community detection using the same LFR-graphs and show that DOLPA achieves about three times the F-Score at just 10% of the runtime. For connected component decomposition, our algorithm achieves orders of magnitude speedups over the basic LP-based algorithm on large-diameter graphs, up to 13.2× speedup over the Shiloach-Vishkin algorithm, and up to 1.6× speedup over Afforest on an Intel Xeon processor using 40 threads.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

Coping with silent and fail-stop errors at scale by combining replication and checkpointing

This paper provides a model and an analytical study of replication as a technique to cope with silent errors, as well as a mixture of both silent and fail-stop errors on large-scale platforms. Compared with fail-stop errors that are immediately detected when they occur, silent errors require a detection mechanism. To detect silent errors, many application-specific techniques are available, either based on algorithms (e.g., ABFT), invariant preservation or data analytics, but replication remains the most transparent and least intrusive technique. We explore the right level (duplication, triplication or more) of replication for two frameworks: (i) when the platform is subject to only silent errors, and (ii) when the platform is subject to both silent and fail-stop errors. A higher level of replication is more expensive in terms of resource usage but enables to tolerate more errors and to even correct some errors, hence there is a trade-off to be found. Replication is combined with checkpointing and comes with two flavors: process replication and group replication. Process replication applies to message-passing applications with communicating processes. Each process is replicated, and the platform is composed of process pairs, or triplets. Group replication applies to black-box applications, whose parallel execution is replicated several times. The platform is partitioned into two halves (or three thirds). In both scenarios, results are compared before each checkpoint, which is taken only when both results (duplication) or two out of three results (triplication) coincide. Otherwise, one or more silent errors have been detected, and the application rolls back to the last checkpoint, as well as when fail-stop errors have struck. We provide a detailed analytical study for all of these scenarios, with formulas to decide, for each scenario, the optimal parameters as a function of the error rate, checkpoint cost, and platform size. We also report a set of extensive simulation results that nicely corroborates the analytical model.

Checkpointing↗
DOE OSTI2025

Two-level overlapping additive Schwarz preconditioner for training scientific machine learning applications

In this work we introduce a novel two-level overlapping additive Schwarz preconditioner for accelerating the training of scientific machine learning applications. The design of the proposed preconditioner is motivated by the nonlinear two-level overlapping additive Schwarz preconditioner. The neural network parameters are decomposed into groups (subdomains) with overlapping regions. In addition, the network’s feed-forward structure is indirectly imposed through a novel subdomain-wise synchronization strategy and a coarse-level training step. Through a series of numerical experiments, which consider physicsinformed neural networks and operator learning approaches, we demonstrate that the proposed two-level preconditioner significantly speeds up the convergence of the standard (LBFGS) optimizer while also yielding more accurate machine learning models. Moreover, the devised preconditioner is designed to take advantage of model-parallel computations, which can further reduce the training time.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Optimizing Distributed Training on Frontier for Large Language Models

Large language models (LLMs) have demonstrated remarkable success as foundational models, benefiting various downstream applications through fine-tuning. Loss scaling studies have demonstrated the superior performance of larger LLMs compared to their smaller counterparts. Nevertheless, training LLMs with billions of parameters poses significant challenges and requires considerable computational resources. For example, training a one trillion parameter GPT-style model on 20 trillion tokens requires a staggering 120 million exaflops. This research explores efficient distributed training strategies to extract this computation from Frontier, the world's first exascale supercomputer. We enable and investigate various model and data parallel training techniques, such as tensor parallelism, pipeline parallelism, and sharded data parallelism, to facilitate training a trillion-parameter model on Frontier. We empirically assess these techniques and their associated parameters to determine their impact on memory footprint, communication latency, and GPU's computational efficiency. We analyze the complex interplay among these techniques and find a strategy to combine them to achieve high throughput through hyperparameter tuning. We have identified efficient strategies for training large LLMs of varying sizes through empirical analysis and hyperparameter tuning. For 22 Billion, 175 Billion, and 1 Trillion parameters, we achieved GPU throughputs of 38.38%, 36.14%, and 31.96%, respectively. For the training of the 175 Billion parameter model and the 1 Trillion parameter model, we achieved 100% weak scaling efficiency on 1024 and 3072 Mi250X GPUs, respectively. We also achieved strong scaling efficiencies of 89% and 87% for these two models. We trained these models only tens of iterations instead of training till completion.

Yin, Junqi↗
DOE OSTI2022

Optimization of quasi-symmetric stellarators with self-consistent bootstrap current and energetic particle confinement

Quasi-symmetry can greatly improve the confinement of energetic particles and thermal plasma in a stellarator. The magnetic field of a quasi-symmetric stellarator at high plasma pressure is significantly affected by the bootstrap current, but the computational cost of accurate stellarator bootstrap calculations has precluded use inside optimization. Here, a new efficient method is demonstrated for optimization of quasi-symmetric stellarator configurations such that the bootstrap current profile is consistent with the geometry. The approach is based on the fact that all neoclassical phenomena in quasi-symmetry are isomorphic to those in axisymmetry. Therefore, accurate formulas for the bootstrap current in tokamaks, which can be evaluated rapidly, can be applied also in stellarators. The deviation between this predicted parallel current and the actual parallel current in the magnetohydrodynamic equilibrium is penalized in the objective function, and the current profile of the equilibrium is included in the parameter space. Quasi-symmetric configurations with significant pressure are thereby obtained with self-consistent bootstrap current and excellent confinement. In a comparison of fusion-produced alpha particle confinement across many stellarators, the new configurations have significantly lower alpha energy losses than many previous designs.

Landreman, M. (ORCID:000000027233577X)↗
DOE OSTI2024

Crystal structure of perfluorononanoic acid, C 9 HF 17 O 2

The crystal structure of perfluorononanoic acid (PFNA) was solved via parallel tempering using synchrotron powder diffraction data obtained from the Brockhouse X-ray Diffraction and Scattering (BXDS) Wiggler Lower Energy (WLE) beamline at the Canadian Light Source. PFNA crystallizes in monoclinic space group P2 1 /c (#14) with lattice parameters a = 26.172(1) Å, b = 5.6345(2) Å, c = 10.9501(4) Å, and β = 98.752(2)°. The crystal structure is composed of dimers, with pairs of PFNA molecules connected by hydrogen bonds via the carboxylic acid functional groups. The Rietveld-refined structure was compared to a density functional theory-optimized structure, and the root-mean-square Cartesian difference was larger than normally observed for correct powder structures. The powder data likely exhibited evidence of disorder which was not successfully modeled.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2019

Avoiding Communication in Primal and Dual Block Coordinate Descent Methods

Primal and dual block coordinate descent methods are iterative methods for solving regularized and unregularized optimization problems. Distributed-memory parallel implementations of these methods have become popular in analyzing large machine learning datasets. However, existing implementations communicate at every iteration, which, on modern data center and supercomputing architectures, often dominates the cost of floating-point computation. Recent results on communication-avoiding Krylov subspace methods suggest that large speedups are possible by re-organizing iterative algorithms to avoid communication. We show how applying similar algorithmic transformations can lead to primal and dual block coordinate descent methods that only communicate every $s$ iterations---where $s$ is a tuning parameter---instead of every iteration for the regularized least-squares problem. We show that the communication-avoiding variants reduce the number of synchronizations by a factor of $s$ on distributed-memory parallel machines without altering the convergence rate and attain strong scaling speedups of up to $6.1\times$ over the “standard algorithm" on a Cray XC30 supercomputer.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Versatile soil gas concentration and isotope monitoring: optimization and integration of novel soil gas probes with online trace gas detection

Abstract. Gas concentrations and isotopic signatures can unveil microbial metabolisms and their responses to environmental changes in soil. Currently, few methods measure in situ soil trace gases such as the products of nitrogen and carbon cycling or volatile organic compounds (VOCs) that constrain microbial biochemical processes like nitrification, methanogenesis, respiration, and microbial communication. Versatile trace gas sampling systems that integrate soil probes with sensitive trace gas analyzers could fill this gap with in situ soil gas measurements that resolve spatial (centimeters) and temporal (minutes) patterns. We developed a system that integrates new porous and hydrophobic sintered polytetrafluoroethylene (sPTFE) diffusive soil gas probes that non-disruptively collect soil gas samples with a transfer system to direct gas from multiple probes to one or more central gas analyzer(s) such as laser and mass spectrometers. Here, we demonstrate the feasibility and versatility of this automated multiprobe system for soil gas measurements of isotopic ratios of nitrous oxide (δ18O, δ15N, and the 15N site preference of N2O), methane, carbon dioxide (δ13C), and VOCs. First, we used an inert silica matrix to challenge probe measurements under controlled gas conditions. By changing and controlling system flow parameters, including the probe flow rate, we optimized recovery of representative soil gas samples while reducing sampling artifacts on subsurface concentrations. Second, we used this system to provide a real-time window into the impact of environmental manipulation of irrigation and soil redox conditions on in situ N2O and VOC concentrations. Moreover, to reveal the dynamics in the stable isotope ratios of N2O (i.e., 14N14N16O, 14N15N16O, 15N14N16O, and 14N14N18O), we developed a new high-precision laser spectrometer with a reduced sample volume demand. Our integrated system – a tunable infrared laser direct absorption spectrometry (TILDAS) in parallel with Vocus proton transfer reaction mass spectrometry (PTR-MS), in line with sPTFE soil gas probes – successfully quantified isotopic signatures for N2O, CO2, and VOCs in real time as responses to changes in the dry–wetting cycle and redox conditions. Broadening the collection of trace gases that can be monitored in the subsurface is critical for monitoring biogeochemical cycles, ecosystem health, and management practices at scales relevant to the soil system.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2022

Prediction of DIII-D Pedestal Structure from Externally Controllable Parameters

The sharp increase of pressure at the edge of a high confinement mode (H-mode) plasma, the pedestal, strongly impacts overall plasma performance. Predicting the pedestal is a necessity to control and optimize tokamak operations. An experimental data-driven machine learning (ML) approach is presented that predicts the pedestal heights and widths of electron density (ne) and electron temperature (Te) profiles as well as the separatrix ne from externally controllable parameters such as the plasma shape, heating method and power, and gas puff rate and integrated gas puff. The OMFIT framework was used with DIII-D data to efficiently, robustly, and automatically build a database of pedestal parameters to train machine learning models. Database creation was enabled by the search engine tool for DIII-D data, TokSearch, which parallelizes data fetching, enabling fast searches through basic signals of thousands of DIII-D shots and selection of relevant time intervals. Principal Component Analysis (PCA) separated the database into three clusters that represent classes of plasma shapes that are regularly used in DIII-D. The most important parameters for setting the pedestal structure were plasma current (Ip), toroidal magnetic field (Bφ), neutral beam heating power (PNBI) and shaping quantities. The Deep Jointly Informed Neural Networks (DJINN) algorithm was applied to identify suitable neural network (NN) architectures that appropriately capture the features of the pedestal database. Separate NNs were implemented for each pedestal parameter, and ensembling methods were used to improve the prediction accuracy and allowed estimation of the prediction uncertainty. The pedestal predictions of the test dataset lie within the measurement uncertainties of the pedestal parameters. The NN outperformed simple Linear Regression (LR) analysis, indicating non-linear dependencies in the pedestal structure. The presented achievements illustrate a promising path for future research, using feature extraction to infer experimental trends and thereby improve pedestal models as well as deploying NN for a fast pedestal prediction in DIII-D scenario development.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗