Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Computational graph”

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 1,693 records · Page 94
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Sparse invariant domain preserving discontinuous Galerkin methods with subcell convex limiting

In this paper, we develop high-order nodal discontinuous Galerkin (DG) methods for hyperbolic conservation laws that satisfy invariant domain preserving properties using subcell flux corrections and convex limiting. These methods are based on a subcell flux corrected transport (FCT) methodology that involves blending a high-order target scheme with a robust, low-order invariant domain preserving method that is obtained using a graph viscosity technique. Furthermore, the new low-order discretizations are based on sparse stencils which do not increase with the polynomial degree of the high-order DG method. As a result, the accuracy of the low-order method does not degrade when used with high-order target methods. The method is applied to both scalar conservation laws, for which the discrete maximum principle is naturally enforced, and to systems of conservation laws such as the Euler equations, for which positivity of density and a minimum principle for specific entropy are enforced. Numerical results are presented on a number of benchmark test cases.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Analyzing the Quantum Approximate Optimization Algorithm: Ansätze, Symmetries, and Lie Algebras

The quantum approximate optimization algorithm (QAOA) has been proposed as a method to obtain approximate solutions for combinatorial optimization tasks. In this work, we study the underlying algebraic properties of three QAOA ansätze for the maximum-cut problem on connected graphs, while focusing on the generated Lie algebras as well as their invariant subspaces. Specifically, we analyze the standard QAOA ansatz as well as the orbit and multiangle ansätze. We are able to fully characterize the Lie algebras of the multiangle ansatz across arbitrary connected graphs, finding that they only fall into one of just six families. Aside from the cycle and path graphs, the Lie dimensions for every graph are exponentially large in the system size, meaning that multiangle ansätze are extremely prone to exhibiting barren plateaus. Then, a similar quasi-graph-independent Lie-algebraic characterization beyond the multiangle ansatz is impeded as the circuit exhibits additional “hidden” symmetries besides those naturally arising from a certain parity-superselection operator and all automorphisms of the considered graph. Disregarding the “hidden” symmetries, we can upper bound the dimensions of the orbit and the standard Lie algebras, and the dimensions of the associated invariant subspaces are determined via explicit character formulas. To finish, we conjecture that (for most graphs) the standard Lie algebras have only components that are either exponential or that grow, at most, polynomially with the system size. This would imply that the QAOA is either prone to barren plateaus or classically simulable. More generally, our work provides a symmetry framework and tools to analyze any desired variational quantum algorithm.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
NASA NTRS2004

Adaptive Bio-Inspired Wireless Network Routing for Planetary Surface Exploration

Wireless mobile networks suffer connectivity loss when used in a terrain that has hills, and valleys when line of sight is interrupted or range is exceeded. To resolve this problem and achieve acceptable network performance, we have designed an adaptive, configurable, hybrid system to automatically route network packets along the best path between multiple geographically dispersed modules. This is very useful in planetary surface exploration, especially for ad-hoc mobile networks, where computational devices take an active part in creating a network infrastructure, and can actually be used to route data dynamically and even store data for later transmission between networks. Using inspiration from biological systems, this research proposes to use ant trail algorithms with multi-layered information maps (topographic maps, RF coverage maps) to determine the best route through ad-hoc network at real time. The determination of best route is a complex one, and requires research into the appropriate metrics, best method to identify the best path, optimizing traffic capacity, network performance, reliability, processing capabilities and cost. Real ants are capable of finding the shortest path from their nest to a food source without visual sensing through the use of pheromones. They are also able to adapt to changes in the environment using subtle clues. To use ant trail algorithms, we need to define the probability function. The artificial ant is, in this case, a software agent that moves from node to node on a network graph. The function to calculate the fitness (evaluate the better path) includes: length of the network edge, the coverage index, topology graph index, and pheromone trail left behind by other ant agents. Each agent modifies the environment in two different ways: 1) Local trail updating: As the ant moves between nodes it updates the amount of pheromone on the edge; and 2) Global trail updating: When all ants have completed a tour the ant that found the shortest route updates the edges in its path.

Alena, Richard I.↗
DOE OSTI2019

Software Defined Architectures for Data Analytics

Data analytics applications increasingly are complex workflows composed of phases with very different program behaviors (e.g., graph algorithms and machine learning, algorithms operating on sparse an dense data structures, etc). To reach the levels of efficiency required to process these workflows in real time, upcoming architectures will need to leverage even more workload specialization. If, at one end, we may find even more heterogenous processors composed by a myriad of specialized processing elements, at the other end we may see novel reconfigurable architectures, composed of sets of functional units and memories interconnected with (re)configurable on-chip networks, able to adapt dynamically to adapt the workload characteristics. Field Programmable Gate Arrays are more and more used for accelerating various workloads and, in particular, inferencing in machine learning, providing higher efficiency than other solutions. However, their fine grained nature still leads to issues for the design software and still makes dynamic reconfiguration impractical. Future, more coarse-grained architectures could offer the features to execute diverse workloads at high efficiency while providing better reconfiguration mechanisms for dynamic adaptability. Nevertheless, we argue that the challenges for reconfigurable computing remain in the software. In this position paper, we describe a possible toolchain for reconfigurable architectures targeted at data analytics.

FPGA, Dataflow, Reconfigurable Architectures, Soft↗
NASA NTRS1999

A Study of Direct and Cloud-Mediated Radiative Forcing of Climate Due to Aerosols

The Intergovernmental Panel on Climate Change (IPCC) has reported that in the southeastern US and eastern China, the general greenhouse warming due to anthropogenic gaseous emissions is dominated by the cooling effect of anthropogenic aerosols. To verify this model prediction in eastern China and southeastern US, we analyzed regional patterns of climate changes at 72 stations in eastern China during 1951- 94 (44 years), and at 52 stations in the southeastern US during 1949-94 (46 years) to detect the fingerprint of aerosol radiative forcing. It was found that the mean rates of change of annual mean daily, maximum, minimum temperatures and diurnal temperature range (DTR) in eastern China were 0.8, -0.2, 1.8, and -2.0 C/100 years respectively, while the mean rates of change of annual mean daily, maximum, minimum temperatures and DTR in the southeastern US were -0.2, -0.6, 0.2, and -0.8 C/100 years, respectively. This indicates that the high rate of increase in annual mean minimum temperature in eastern China results in a slightly warming trend of daily temperature, while the high rate of decrease in annual mean maximum temperature and low rate of increase in annual mean minimum temperature lead to the cooling trend of daily temperature in the southeastern US. We found that the warming from the longwave forcing due to both greenhouse gases and aerosols was completely counteracted by the shortwave aerosol forcing in the southeastern US in the past 46 years. A slightly overall warming trend in eastern China is evident; winters have become milder. This finding is explained by hypothesizing that increasing energy usage during the past 44 years has resulted in more coal and biomass burning, thus increasing the emission of absorbing soot and organic aerosols in eastern China. Such emissions, in addition to well-known Asia dust and greenhouse gases, may be responsible for the winter warming trend in eastern China that we have reported here. The sensitivity of aerosol radiative properties to aerosol composition, size distribution, relative humidity (RH) is examined for the following aerosol systems: inorganic and organic ions (Cl-, Br-, NO3 -, SO4 2-, Na+, NH4 +, K+, Ca2+, Mg2+, HCOO-, CH3COO-, CH3CH2COO-, CH3COCOO-, OOCCOO2-, MSA-1); water-insoluble inorganic and organic compounds (elemental carbon, n-alkanes, SiO2, Al2O3, Fe2O3 and other organic compounds). The partial molar refraction method was used to calculate the real part of the refractive index. It was found that the asymmetry factor increased by approximately 48% with the real part varying from 1.40 to 1.65, and the single scattering albedo decreased by 24% with the imaginary part varying from -0.005 to -0.1. The asymmetry factor increased by 5.4 times with the geometric standard deviation varying from 1.2 to 3.0. The radiation transmission is very sensitive to the change in size distribution; other factors are not as significant. To determine the aerosol direct radiative forcing (ADRF), the aerosol optical depth (AOD) values at the three operational wavelengths (415, 500 and 673 nm) were determined at a regionally representative site, namely, Mt. Gibbs (35.78 deg N, 82.29 deg W, elevation 2006 m) in Mt. Mitchell State Park, NC, and a site located in an adjacent valley (Black Mountain, 35.66 deg N, 82.38 deg W, elevation 951 m) in the southeastern US. The two sites are separated horizontally by 10 km and vertically by 1 km. It was found that the representative total AOD values at 500 nm at the valley site for highly polluted (HP), marine (M) and continental (C) air masses were 0.68 +/- 0.33, 0.29 +/- 0.19 and 0.10 +/- 0.04, respectively. A search-graph method was used to retrieve the columnar size distribution (number concentration N, effective radius reff and geometric standard deviation=?g) from the optical depth observations at three operational wavelengths. The ground albedo, single scattering albedo and imaginary part of the refractive index were calculated using a mathematically unique procedure involving a Mie code and a radiative transfer code in conjunction with the retrieved aerosol size distribution, AOD, and diffuse-direct irradiance ratio. It was found that N, r(eff) and sigma(g) were in the ranges of 10 to 1.7 x 10(exp 4)/cubic cm, 0.09 to 0.68 micrometers and 1.12 to 2.95, respectively. The asymmetry factor and single scattering albedo were in the ranges of 0.63 to 0.75 and 0.74 to 0.97 respectively. The ground albedo for the forested terrain and imaginary part of refractive index were found to be in the ranges of 0.06 to 0.29 and 0.005 to 0.051 respectively. On the basis of these aerosol radiative properties obtained at the research sites and computations using the Column Radiation Model (CRM) of National Center of Atmospheric Research (NCAR) Community Climate Model (CCM3), it was found that the average cloud-free 24-hour ADRF values were -13 +/- 8, -8 +/- 3, -33 +/- 16 W/square m for marine, continental, and polluted air masses, respectively. On the assumption that the fractional coverage of clouds is 0.61, it was estimated that the annual mean ADRF was 7 +/- 2 W/square m in the southeastern US. The review with respect to the current knowledge of organic acids shows that aerosol formate and acetate concentrations range from 0.02 to 5.3 nmol/cubic m and from 0.03 to 12.4 nmol/cubic m respectively, and that between 34% to 77% of formate and between 21% to 66% of acetate are present in the fine fraction of aerosols. It was found that although most (98-99%) of these volatile organic acids were present in the gas phase, their concentrations in the aerosol particles were sufficient to make them a good candidate for cloud condensation nuclei (CCN). It is hypothesized that organic acids are at least one of the primary sources of CCN in the atmosphere due to their ubiquitous presence in the troposphere, especially over the continental forested areas. The results of our measurements at Palmer Station, Antarctica show that the daily average CCN concentrations at 0.3% and 1% supersaturations ranged from 0.3 to 160/cubic cm and from 4 to 168/cubic cm, respectively, during the period from 17 January to 26 February, 1994. New evidence for substantial and definitive CCN enhancement near and within cloud has been observed at Mt. Mitchell, North Carolina. The results show that the average monthly CCN concentrations were 460 +/- 217, 386 +/- 286, 429 +/- 228 and 238 +/- 134/cubic cm for in-cloud, overcast, clear and rainy conditions, respectively. The typical CCN spectra show that there were a lot of small CCN produced and the ion concentrations (especially H+ and SO4 2-) were very high during the CCN enhancement period. The significantly positive correlation between black carbon (BC) and CCN at 1% supersaturation indicates that a percentage of the BC measured at the site may be in the form of an internal mixture and participated in the formation of CCN.

Yu, Shao-Cai↗
DOE OSTI2022

Efficient and interpretable graph network representation for angle-dependent properties applied to optical spectroscopy

Abstract Graph neural networks are attractive for learning properties of atomic structures thanks to their intuitive graph encoding of atoms and bonds. However, conventional encoding does not include angular information, which is critical for describing atomic arrangements in disordered systems. In this work, we extend the recently proposed ALIGNN (Atomistic Line Graph Neural Network) encoding, which incorporates bond angles, to also include dihedral angles (ALIGNN-d). This simple extension leads to a memory-efficient graph representation that captures the complete geometry of atomic structures. ALIGNN-d is applied to predict the infrared optical response of dynamically disordered Cu(II) aqua complexes, leveraging the intrinsic interpretability to elucidate the relative contributions of individual structural components. Bond and dihedral angles are found to be critical contributors to the fine structure of the absorption response, with distortions that represent transitions between more common geometries exhibiting the strongest absorption intensity. Future directions for further development of ALIGNN-d are discussed.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Electricity Market Dynamics as Smooth Games: Organizing Gradient-Based Learners and Identifying Instabilities

Wide-spread adoption of machine learning and distributed energy production has the potential to fundamentally change behaviours of energy markets. In practice, along with the benefits, previously unseen instabilities have begun emerging in markets currently embracing these technologies. In this report we study electricity markets as dynamical systems arising in smooth games, where energy consumers are modelled as gradient-based learners who interact on a graph through distributed energy producers. We give simple examples of how instabilities can arise even when the consumers’ local objectives seem reasonable and how these instabilities can be remedied. Drawing insights from these examples, we begin developing a general model of optimization-driven electricity market dynamics and give conditions for ensuring local Nash equilibria of the system are stable fixed points of the dynamics. Our results demonstrate how models of electricity market dynamics might be used for deriving formal guarantees in order to stabilize markets and enact regulatory policies in the face of market shocks.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2017

Master integrals for the NNLO virtual corrections to μe scattering in QED: the planar graphs

Here, we evaluate the master integrals for the two-loop, planar box-diagrams contributing to the elastic scattering of muons and electrons at next-to-next-to leading-order in QED. We adopt the method of differential equations and the Magnus exponential series to determine a canonical set of integrals, finally expressed as a Taylor series around four space-time dimensions, with coefficients written as combination of generalized poly-logarithms. The electron is treated as massless, while we retain full dependence on the muon mass. The considered integrals are also relevant for crossing-related processes, such as di-muon production at e + e – -colliders, as well as for the QCD corrections to top -pair production at hadron colliders.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2025

Physics-based stabilized finite element approximations of the Poisson–Nernst–Planck equations

We present and analyze two stabilized finite element methods for solving numerically the Poisson–Nernst–Planck equations. The stabilization we consider is carried out by using a shock detector and a discrete graph Laplacian operator for the ion equations, whereas the discrete equation for the electric potential need not be stabilized. Discrete solutions stemmed from the first algorithm preserve both maximum and minimum discrete principles. For the second algorithm, its discrete solutions are conceived so that they hold discrete principles and obey an entropy law provided that an acuteness condition is imposed for meshes. Remarkably the latter is found to be unconditionally stable. We validate our methodology through transient numerical experiments that show convergence toward steady-state solutions.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Active causal learning for decoding chemical complexities with targeted interventions

Abstract Predicting and enhancing inherent properties based on molecular structures is paramount to design tasks in medicine, materials science, and environmental management. Most of the current machine learning and deep learning approaches have become standard for predictions, but they face challenges when applied across different datasets due to reliance on correlations between molecular representation and target properties. These approaches typically depend on large datasets to capture the diversity within the chemical space, facilitating a more accurate approximation, interpolation, or extrapolation of the chemical behavior of molecules. In our research, we introduce an active learning approach that discerns underlying cause-effect relationships through strategic sampling with the use of a graph loss function. This method identifies the smallest subset of the dataset capable of encoding the most information representative of a much larger chemical space. The identified causal relations are then leveraged to conduct systematic interventions, optimizing the design task within a chemical space that the models have not encountered previously. While our implementation focused on the QM9 quantum-chemical dataset for a specific design task—finding molecules with a large dipole moment—our active causal learning approach, driven by intelligent sampling and interventions, holds potential for broader applications in molecular, materials design and discovery.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Deep Analysis Net with Causal Embedding for Coal-fired Power Plant Fault Detection and Diagnosis (DANCE4CFDD)

Fault detection and diagnosis is critical to power plant operation to ensure attaining high reliability while reducing operation cost. As more renewable power is introduced to the power grid, traditional fossil power plants take on the extra burden of excessive load cycling to compensate the generation variability from renewable power. Such load cycling will pose more reliability challenges to power plant operation. There are a number of challenges faced by today’s asset health management system in coal- fired (or gas) power plants: 1) high-dimensional nonlinear interaction among multiple time series measurements; 2) high measurement variance induced by operational conditions/modes; 3) variation among asset types and plant configurations; and 4) a small number of faulty events to learn from. To cope with these challenges, today’s fielded asset health management systems rely heavily on manual efforts from domain experts and hand-crafted features or rules based on domain knowledge. Despite its role in plant reliability, such a practice is costly and hinders its scalability and sustainability, particularly when a plant undergoes modifications. The objective of this project is to develop a novel end-to-end AI learning system that is trainable (i.e., the AI representation of a complex system behavior can be directly learned from properly labeled data) for accurate fault detection and root cause analysis. The ability to create a fault detection model directly from time series could alleviate the efforts associated with today’s asset management solution development. In the course of this project, we have achieved the following: Created an AI model development environment incorporating state-of-the-art neural network architectures for rapid model development and evaluation; Developed novel learning strategies for training of fault detection model; Developed special-purpose neural network architecture embedded with variable association graph aiming for better interpretability; Developed a learning strategy to leverage a small number of faulty events for enhanced fault detection capability; Conducted detailed experimental study based on public benchmark datasets and demonstrated the effectiveness of the proposed solution; and Validated the developed system with data from both a coal-fired plant boiler dynamic simulation model and real-world coal-fired power plant covering multiple asset and fault types. Overall, the project attained a technology readiness level of TRL 5 from TRL 2 at the beginning of the project.

20 FOSSIL-FUELED POWER PLANTS↗
DOE OSTI2025

Decision and Control of Complex Systems – A Data-Drive Framework

During the project period, we have collaborated with other team members and developed novel algorithms for novelty detection, continual learning, and graph learning algorithms for dynamic systems. The results are documented in publications and meeting notes. Moreover, we leverage virtual collaboration tools (such as Basecamp, Microsoft Teams and Zoom) for technical exchanges. Our research on novelty detection was published at AAAI 2022 and Lecture Notes in Artificial Intelligence, Springer Nature. The newly developed algorithms were successfully applied to realistic cases, including thermal data from buildings at Pacific Northwest National Lab and microelectronic data provided by GlobalFoundries. Multiple publications have been produced from this project, in collaboration with other team members. Three PhD students were supported in this project to conduct their research.

42 ENGINEERING↗
DOE OSTI2023

Dynamic conformational switching underlies TFIIH function in transcription and DNA repair and impacts genetic diseases

Transcription factor IIH (TFIIH) is a protein assembly essential for transcription initiation and nucleotide excision repair (NER). Yet, understanding of the conformational switching underpinning these diverse TFIIH functions remains fragmentary. TFIIH mechanisms critically depend on two translocase subunits, XPB and XPD. To unravel their functions and regulation, we build cryo-EM based TFIIH models in transcription- and NER-competent states. Using simulations and graph-theoretical analysis methods, we reveal TFIIH’s global motions, define TFIIH partitioning into dynamic communities and show how TFIIH reshapes itself and self-regulates depending on functional context. Our study uncovers an internal regulatory mechanism that switches XPB and XPD activities making them mutually exclusive between NER and transcription initiation. By sequentially coordinating the XPB and XPD DNA-unwinding activities, the switch ensures precise DNA incision in NER. Mapping TFIIH disease mutations onto network models reveals clustering into distinct mechanistic classes, affecting translocase functions, protein interactions and interface dynamics.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2026

Quantum Routing and Entanglement Dynamics Through Bottlenecks

To implement arbitrary quantum circuits in architectures with restricted interactions, one may effectively simulate all-to-all connectivity by routing quantum information. We consider the entanglement dynamics and routing between two regions only connected through an intermediate “bottleneck” region with few qubits. In such systems, where the entanglement rate is restricted by a vertex boundary rather than an edge boundary of the underlying interaction graph, existing results such as the small incremental entangling theorem give only a trivial constant lower bound on the routing time (the minimum time to perform an arbitrary permutation). We significantly improve the lower bound on the routing time in systems with a vertex bottleneck. Specifically, for any system with two regions 𝐿,𝑅 with 𝑁 𝐿 ,𝑁 𝑅 qubits, respectively, coupled only through an intermediate region 𝐶 with 𝑁 𝐶 qubits, for any 𝛿 > 0 we show a lower bound of Ω⁢(𝑁$^{1−𝛿}_{𝑅}$/√𝑁 𝐿⁢ 𝑁 𝐶 ) on the Hamiltonian quantum routing time when using piecewise time-independent Hamiltonians, or time-dependent Hamiltonians subject to a smoothness condition. We also prove an upper bound on the average amount of bipartite entanglement between 𝐿 and 𝐶,𝑅 that can be generated in time 𝑡 by such architecture-respecting Hamiltonians in systems constrained by vertex bottlenecks, improving the scaling in the system size from 𝑂⁡(𝑁 𝐿⁢ 𝑡) to 𝑂⁡(√𝑁 𝐿⁢ 𝑡). As a special case, when applied to the star graph (i.e., one vertex connected to 𝑁 leaves), we obtain an Ω⁡(√𝑁 1−𝛿 ) lower bound on the routing time and on the time to prepare 𝑁/2 Bell pairs between the vertices. We also show that, in systems of free particles, we can route optimally on the star graph in time Θ⁡(√𝑁) using Hamiltonian quantum routing, obtaining a speedup over gate-based routing, which takes time Θ⁡(𝑁).

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Cluster expansion by transfer learning for phase stability predictions

Recent progress towards universal machine-learned interatomic potentials holds considerable promise for materials discovery. Yet the accuracy of these potentials for predicting phase stability may still be limited. In contrast, cluster expansions provide accurate phase stability predictions but are computationally demanding to parameterize from first principles, especially for structures of low dimension or with a large number of components, such as interfaces or multimetal catalysts. We overcome this trade-off via transfer learning. Using Bayesian inference, we incorporate prior statistical knowledge from machine-learned and physics-based potentials, enabling us to sample the most informative configurations and to efficiently fit first-principles cluster expansions. Furthermore, this algorithm is tested on Pt:Ni, showing robust convergence of the mixing energies as a function of sample size with reduced statistical fluctuations.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Density of states prediction for materials discovery via contrastive learning from probabilistic embeddings

Abstract Machine learning for materials discovery has largely focused on predicting an individual scalar rather than multiple related properties, where spectral properties are an important example. Fundamental spectral properties include the phonon density of states (phDOS) and the electronic density of states (eDOS), which individually or collectively are the origins of a breadth of materials observables and functions. Building upon the success of graph attention networks for encoding crystalline materials, we introduce a probabilistic embedding generator specifically tailored to the prediction of spectral properties. Coupled with supervised contrastive learning, our materials-to-spectrum (Mat2Spec) model outperforms state-of-the-art methods for predicting ab initio phDOS and eDOS for crystalline materials. We demonstrate Mat2Spec’s ability to identify eDOS gaps below the Fermi energy, validating predictions with ab initio calculations and thereby discovering candidate thermoelectrics and transparent conductors. Mat2Spec is an exemplar framework for predicting spectral properties of materials via strategically incorporated machine learning techniques.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

IRIS: Exploring Performance Scaling of the Intelligent Runtime System and its Dynamic Scheduling Policies

High-Performance Computing is becoming increasingly heterogeneous, relying on a diverse mix of hardware to achieve good performance. Paradoxically, current drivers and frameworks for these devices typically require separate languages and implementations for each vendor. Furthermore, there are few tools and little support to schedule codes between these devices in a truly heterogeneous manner-partly because of this fragmentation between vendors and the languages each supports. To overcome both limitations, the Intelligent Runtime System (IRIS) was developed. It allows a common task abstraction to automatically be shared among contemporary vendors and is run from a single host-side API. At runtime, IRIS queries the host system and registers which frameworks and drivers are available, these determine which kernels can be used by the scheduler-CPUs via OpenMP, Nvidia GPUs (CUDA), AMD GPUs (HIP), and Intel and Xilinx FPGAs with OpenCL. IRIS enables tasks to be scheduled to any heterogeneous device and resolves to the appropriate kernel binary at runtimeit only uses the devices supported by the system on which it is run. IRIS supports single-task and graph-based expressions of dependencies of tasks. Additionally, IRIS features a range of dynamic scheduling policies, allowing complex chains of tasks and interactions to be executed, relieving the programmer/user from considering the system to assign tasks to devices optimally. This paper presents the peak performance attainable by IRIS over a range of systems-each with different numbers and types of accelerator devices, it highlights the flexibility of IRIS since these devices are truly heterogeneous, relying on different backends (drivers, frameworks, and languages) which historically required unique implementations to utilize them. We then use this peak performance as a baseline to compare increasingly complex chains of tasks (with increasingly complex task dependencies) and evaluate how IRIS copes. Finally, we consider the performance of different IRIS scheduling policies on this range of task graphs.

Johnston, Beau↗
DOE OSTI2024

DLSIA: Deep Learning for Scientific Image Analysis

DLSIA (Deep Learning for Scientific Image Analysis) is a Python-based machine learning library that empowers scientists and researchers across diverse scientific domains with a range of customizable convolutional neural network (CNN) architectures for a wide variety of tasks in image analysis to be used in downstream data processing. DLSIA features easy-to-use architectures, such as autoencoders, tunable U-Nets and parameter-lean mixed-scale dense networks (MSDNets). Additionally, this article introduces sparse mixed-scale networks (SMSNets), generated using random graphs, sparse connections and dilated convolutions connecting different length scales. For verification, several DLSIA-instantiated networks and training scripts are employed in multiple applications, including inpainting for X-ray scattering data using U-Nets and MSDNets, segmenting 3D fibers in X-ray tomographic reconstructions of concrete using an ensemble of SMSNets, and leveraging autoencoder latent spaces for data compression and clustering. As experimental data continue to grow in scale and complexity, DLSIA provides accessible CNN construction and abstracts CNN complexities, allowing scientists to tailor their machine learning approaches, accelerate discoveries, foster interdisciplinary collaboration and advance research in scientific image analysis.

97 MATHEMATICS AND COMPUTING↗