Engineering Papers⌕ Search

Engineering topics

Foster, Ian T.

Publications and source records attributed to Foster, Ian T..

At least 19 records
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Assessment of fine-tuned large language models for real-world chemistry and material science applications

The current generation of large language models (LLMs) has limited chemical knowledge. Recently, it has been shown that these LLMs can learn and predict chemical properties through fine-tuning. Using natural language to train machine learning models opens doors to a wider chemical audience, as field-specific featurization techniques can be omitted. In this work, we explore the potential and limitations of this approach. We studied the performance of fine-tuning three open-source LLMs (GPT-J-6B, Llama-3.1-8B, and Mistral-7B) for a range of different chemical questions. We benchmark their performances against “traditional” machine learning models and find that, in most cases, the fine-tuning approach is superior for a simple classification problem. Depending on the size of the dataset and the type of questions, we also successfully address more sophisticated problems. The most important conclusions of this work are that, for all datasets considered, their conversion into an LLM fine-tuning training set is straightforward and that fine-tuning with even relatively small datasets leads to predictive models. These results suggest that the systematic use of LLMs to guide experiments and simulations will be a powerful technique in any research study, significantly reducing unnecessary experiments or computations.

Van Herck, Joren↗
DOE OSTI2024

Frontiers in Scientific Workflows: Pervasive Integration With High-Performance Computing

Herein we address the increasing complexity of scientific workflows in the context of high-performance computing (HPC) and their associated need for robust, adaptable, and flexible computational support systems. We explore five key trends as well as future challenges and opportunities for scientific workflows and HPC technologies.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Uncovering novel liquid organic hydrogen carriers: a systematic exploration of chemical compound space using cheminformatics and quantum chemical methods

We present a comprehensive, in silico-based discovery approach to identifying novel liquid organic hydrogen carrier (LOHC) candidates using cheminformatics methods and quantum chemical calculations. We screened over 160 billion molecules from ZINC15 and GDB-17 chemical databases for structural similarity to known LOHCs and employed a data-driven selection criterion connecting molecular features with dehydrogenation enthalpy. This scoring criterion effectively predicts dehydrogenation enthalpies from SMILES strings, streamlining the LOHC screening process. After rigorous screening and down-selection, we compiled a database of 3000 dehydrogenation reactions for the most promising LOHC candidates, setting the stage for future selection based on kinetics and catalysis. This work demonstrates the significant impact of integrating quantum chemistry and cheminformatics in materials discovery, accelerating the selection process while reducing experimental efforts and time. By proposing new molecules as prospective LOHC candidates, our study provides a valuable resource for researchers and engineers in the development of advanced LOHC systems and showcases a successful approach for high-throughput discovery, contributing to more efficient and sustainable energy storage solutions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

A Combined Machine Learning and High-Energy X-ray Diffraction Approach to Understanding Liquid and Amorphous Metal Oxides

Determining the structure-property relations of liquid and amorphous metal oxides is challenging, due to their variable short-range order and polyhedral connectivity. To predict chemically realistic structures, we have developed a Machine Learned, Gaussian Approximation Potential (GAP) for HfO 2 , with a focus on enhanced sampling of the training database and accurate density functional theory calculations. By using training datasets for the GAP model at the level of Density Functional Theory-Strongly Constrained and Appropriately Normed (DFT-SCAN) level of theory, our results show that the topology of both the low viscosity liquid and the amorphous form are dominated by edge-shared chains and small corner-shared rings of polyhedra. This topology is shown to be consistent with the structure of other liquid and amorphous transition metal oxides of variable ion size, such as TiO 2 and ZrO 2 . Current limitations of the ML-GAP modeling method for obtaining glass structures and future perspectives are also discussed.

36 MATERIALS SCIENCE↗
DOE OSTI2022

IMPROVING THE ACCURACY OF COMPOSITE METHODS: A G4MP2 METHOD WITH G4-LIKE ACCURACY AND IMPLICATIONS FOR MACHINE LEARNING

G4MP2 theory has proven to be a reliable and accurate quantum chemical composite method for the calculation of molecular energies using an approximation based on second-order perturbation theory to lower computational costs compared to G4 theory. However, it has been found to have significantly increased errors when applied to larger organic molecules with 10 or more nonhydrogen atoms. We report here on an investigation of the cause of the failure of G4MP2 theory for such larger molecules. One source of error is found to be the "higher-level correction (HLC)", which is meant to correct for deficiencies in correlation contributions to the calculated energies. This is because the HLC assumes that the contribution is independent of the element and the type of bonding involved, both of which become more important with larger molecules. We address this problem by adding an atom-specific correction, dependent on atom type but not bond type, to the higher-level correction. We find that a G4MP2 method that incorporates this modification of the higher-level correction, referred to as G4MP2A, becomes as accurate as G4 theory (for computing enthalpies of formation) for a test set of molecules with less than 10 nonhydrogen atoms as well as a set with 10-14 such atoms, the set of molecules considered here, with a much lower computational cost. The G4MP2A method is also found to significantly improve ionization potentials and electron affinities. Finally, we implemented the G4MP2A energies in a machine learning method to predict molecular energies.

Dandu, Naveen↗
DOE OSTI2021

MemXCT: Design, Optimization, Scaling, and Reproducibility of X-Ray Tomography Imaging

Here, this work extends our previous research entitled "MemXCT: Memory-centric X-ray CT Reconstruction with Massive Parallelization" that was originally published at SC19 conference (Hidayetoglu et al., 2019) with reproducibility of the computational imaging performance. X-ray computed tomography (XCT) is regularly used at synchrotron light sources to study the internal morphology of materials at high resolution. However, experimental constraints, such as radiation sensitivity, can result in noisy or undersampled measurements. Further, depending on the resolution, sample size and data acquisition rates, the resulting noisy dataset can be in the order of terabytes. Advanced iterative reconstruction techniques can produce high-quality images from noisy measurements, but their computational requirements have made their use an exception rather than the rule. We propose a novel memory-centric approach that avoids redundant computations at the expense of additional memory complexity. We develop a memory-centric iterative reconstruction system, MemXCT, that uses an optimized SpMV implementation with two-level pseudo-Hilbert ordering and multi-stage input buffering. We evaluate MemXCT on various supercomputer architectures involving KNL and GPU. MemXCT can reconstruct a large (11Kx11K) mouse brain tomogram in 10 seconds using 4096 KNL nodes (256K cores). The results presented in our original article at the SC19 were based on large-scale supercomputing resources. The MemXCT application was selected for the Student Cluster Competition (SCC) Reproducibility Challenge and evaluated on a variety of cloud computing resources by universities around the world in the SC20 conference. We summarize the results of the top-ranked SCC Reproducibility Challenge teams and identify the most pertinent measures for ensuring the reproducibility of our experiments in this article.

47 OTHER INSTRUMENTATION↗
DOE OSTI2021

2'-O methylation of RNA cap in SARS-CoV-2 captured by serial crystallography

The genome of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) coronavirus has a capping modification at the 5'-untranslated region (UTR) to prevent its degradation by host nucleases. These modifications are performed by the Nsp10/14 and Nsp10/16 heterodimers using S-adenosylmethionine as the methyl donor. Nsp10/16 heterodimer is responsible for the methylation at the ribose 2'-O position of the first nucleotide. To investigate the conformational changes of the complex during 2'-O methyltransferase activity, we used a fixed-target serial synchrotron crystallography method at room temperature. We determined crystal structures of Nsp10/16 with substrates and products that revealed the states before and after methylation, occurring within the crystals during the experiments. Here we report the crystal structure of Nsp10/16 in complex with Cap-1 analog ( m7 GpppA m2'-O ). Inhibition of Nsp16 activity may reduce viral proliferation, making this protein an attractive drug target.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

Crowd-Sourced Data and Analysis Tools for Advancing the Chemical Vapor Deposition of Graphene: Implications for Manufacturing

Industrial production of graphene by chemical vapor deposition (CVD) requires more than the ability to synthesize large domain, high-quality graphene in a lab reactor. The integration of graphene in the fabrication process of electronic devices requires the cost-effective and environmentally friendly production of graphene on dielectric substrates, but current approaches can only produce graphene on metal catalysts. Sustainable manufacturing of graphene should also conserve the catalyst and reaction gases, but today the metal catalysts are typically dissolved after synthesis. Progress toward these objectives is hindered by the hundreds of coupled synthesis parameters that can strongly affect CVD of low-dimensional materials and poor communication in the published literature of the rich experimental data that exists in individual laboratories. We report here on a platform, "graphene recipes for synthesis of high quality material" (Gr-ResQ: pronounced graphene rescue), which includes powerful new tools for data-driven graphene synthesis. At the core of Gr-ResQ is a crowd-sourced database of CVD synthesis recipes and associated experimental results. The database captures similar to 300 parameters ranging from synthesis conditions such as a catalyst material and preparation steps, to ambient lab temperature and reactor details, as well as resulting Raman spectra and microscopy images. These parameters are carefully selected to unlock the potential of machine-learning models to advance synthesis. A suite of associated tools enable fast, automated, and standardized processing of Raman spectra and scanning electron microscopy images. To facilitate community-based efforts, Gr-ResQ provides tools for cyber-physical collaborations among research groups, allowing experiments to be designed, executed, and analyzed by different teams. Gr-ResQ also allows publication and discovery of recipes via the Materials Data Facility, which assigns each recipe a unique identifier when published and collects parameters in a search index. We envision that this holistic approach to data-driven synthesis can accelerate CVD recipe discovery and production control and open opportunities for advancing not only graphene but also many other 1D and 2D materials.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Quantum-Chemically Informed Machine Learning: Prediction of Energies of Organic Molecules with 10 to 14 Non-hydrogen Atoms

High-fidelity quantum-chemical calculations can provide accurate predictions of molecular energies, but their high computational costs limit their utility, especially for larger molecules. We have shown in previous work that machine learning models trained on high-level quantum-chemical calculations (G4MP2) for organic molecules with one to nine non-hydrogen atoms can provide accurate predictions for other molecules of comparable size at much lower costs. Here we demonstrate that such models can also be used to effectively predict energies of molecules larger than those in the training set. To implement this strategy, we first established a set of 191 molecules with 10–14 non-hydrogen atoms having reliable experimental enthalpies of formation. We then assessed the accuracy of computed G4MP2 enthalpies of formation for these 191 molecules. The error in the G4MP2 results was somewhat larger than that for smaller molecules, and the reason for this increase is discussed. Two density functional methods, B3LYP and ωB97X-D, were also used on this set of molecules, with ωB97X-D found to perform better than B3LYP at predicting energies. The G4MP2 energies for the 191 molecules were then predicted using these two functionals with two machine learning methods, the FCHL-Δ and SchNet-Δ models, with the learning done on calculated energies of the one to nine non-hydrogen atom molecules. The better-performing model, FCHL-Δ, gave atomization energies of the 191 organic molecules with 10–14 non-hydrogen atoms within 0.4 kcal/mol of their G4MP2 energies. Thus, this work demonstrates that quantum-chemically informed machine learning can be used to successfully predict the energies of large organic molecules whose size is beyond that in the training set.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

Efficient I/O for Neural Network Training with Compressed Data

FanStore is a shared object store that enables efficient and scalable neural network training on supercomputers. By providing a global cache layer on node-local burst buffers using a compressed representation, it significantly enhances the processing capability of deep learning (DL) applications on existing hardware. In addition, FanStore allows POSIX-compliant file access to the compressed data in user space. We investigate the tradeoff between runtime overhead and data compression ratio using real-world datasets and applications, and propose a compressor selection algorithm to maximize storage capacity given performance constraints. We consider both asynchronous (i.e., with prefetching) and synchronous I/O strategies, and propose mechanisms for selecting compressors for both approaches. Using FanStore, the same storage hardware can host 2-13x more data for example applications without significant runtime overhead. Empirically, our experiments show that FanStore scales to 512 compute nodes with near linear performance scalability.

Zhang, Zhao↗
DOE OSTI2019

Creating Training Data for Scientific Named Entity Recognition with Minimal Human Effort

Scientific Named Entity Referent Extraction is often more complicated than traditional Named Entity Recognition (NER). For example, in polymer science, chemical structure may be encoded in a variety of nonstandard naming conventions, and authors may refer to polymers with conventional names, commonly used names, labels (in lieu of longer names), synonyms, and acronyms. As a result, accurate scientific NER methods are often based on task-specific rules, which are difficult to develop and maintain, and are not easily generalized to other tasks and fields. Machine learning models require substantial expert-annotated data for training. Here we propose polyNER: a semi-automated system for efficient identification of scientific entities in text. PolyNER applies word embedding models to generate entity-rich corpora for productive expert labeling, and then uses the resulting labeled data to bootstrap a context-based word vector classifier. Evaluation on materials science publications shows that the polyNER approach enables improved precision or recall relative to a state-of-the-art chemical entity extraction system at a dramatically lower cost: it required just two hours of expert time, rather than extensive and expensive rule engineering, to achieve that result. This result highlights the potential for human-computer partnership for constructing domain-specific scientific NER systems.

Crowdsourcing↗
DOE OSTI2018

Profiling and Predicting Application Performance on the Cloud

Cloud providers continue to expand and diversify their collection of leasable resources to meet the needs of an increasingly wide range of applications. While this flexibility is a key benefit of the cloud, it also creates a complex landscape in which users are faced with many resource choices for a given application. Suboptimal selections can both degrade performance and increase costs. Given the rapidly evolving pool of resources, it is infeasible for users alone to select instance types; instead, automated methods are needed to simplify and guide resource provisioning. Here we present a method for the automatic prediction of application performance on arbitrary cloud instances. We combine offline and online profiling approaches, using historical data gathered from non-cloud environments and targeted profiling runs on cloud environments to create a composite application model that can predict run times on a given cloud instance type for a given input data size. We demonstrate average error of 17.2% across nine applications used in production bioinformatics workflows. Finally, we evaluate an experiment design approach to explore the trade-off between the cost of profiling and the accuracy of our models. Using this approach, with no prior knowledge, we show that using 4 selectively chosen experiments we can achieve performance within 30% of a model trained using all instance types.

Cloud profiling↗
DOE OSTI2018

Skluma: An extensible metadata extraction pipeline for disorganized data

To mitigate the effects of high-velocity data expansion and to automate the organization of filesystems and data repositories, we have developed Skluma-a system that automatically processes a target filesystem or repository, extracts content-and context-based metadata, and organizes extracted metadata for subsequent use. Skluma is able to extract diverse metadata, including aggregate values derived from embedded structured data; named entities and latent topics buried within free-text documents; and content encoded in images. Skluma implements an overarching probabilistic pipeline to extract increasingly specific metadata from files. It applies machine learning methods to determine file types, dynamically prioritizes and then executes a suite of metadata extractors, and explores contextual metadata based on relationships among files. The derived metadata, represented in JSON, describes probabilistic knowledge of each file that may be subsequently used for discovery or organization. Skluma's architecture enables it to be deployed both locally and used as an on-demand, cloud-hosted service to create and execute dynamic extraction workflows on massive numbers of files. It is modular and extensible-allowing users to contribute their own specialized metadata extractors. Thus far we have tested Skluma on local filesystems, remote FTP-accessible servers, and publicly-accessible Globus endpoints. We have demonstrated its efficacy by applying it to a scientific environmental data repository of more than 500,000 files. We show that we can extract metadata from those files with modest cloud costs in a few hours.

data swamp↗
DOE OSTI2018

Advance reservation access control using software-defined networking and tokens

Advance reservation systems allow users to reserve dedicated bandwidth connection resources from advanced high-speed networks. A common use case for such systems is data transfers in distributed science environments in which a user wants exclusive access to the reservation. However, current advance network reservation methods cannot ensure exclusive access of a network reservation to the specific flow for which the user made the reservation. We present here a novel network architecture that addresses this limitation and ensures that a reservation is used only by the intended flow. We achieve this by leveraging software-defined networking (SDN) and token-based authorization. We use SDN to orchestrate and automate the reservation of networking resources, end-to-end and across multiple administrative domains, and tokens to create a strong binding between the user or application that requested the reservation and the flows provisioned by SDN. We conducted experiments on the ESNet 100G SDN testbed, and demonstrated that our system effectively protects authorized flows from competing traffic in the network.

admission control↗
DOE OSTI2018

Scalable pCT Image Reconstruction Delivered as a Cloud Service

We describe a cloud-based medical image reconstruction service designed to meet a real-time and daily demand to reconstruct thousands of images from proton cancer treatment facilities worldwide. Rapid reconstruction of a three-dimensional Proton Computed Tomography (pCT) image can require the transfer of 100 GB of data and use of approximately 120 GPU-enabled compute nodes. The nature of proton therapy means that demand for such a service is sporadic and comes from potentially hundreds of clients worldwide. We thus explore the use of a commercial cloud as a scalable and cost-efficient platform for pCT reconstruction. To address the high performance requirements of this application we leverage Amazon Web Services' GPU-enabled cluster resources that are provisioned with high performance networks between nodes. To support episodic demand, we develop an on-demand multi-user provisioning service that can dynamically provision and resize clusters based on image reconstruction requirements, priorities, and wait times. We compare the performance of our pCT reconstruction service running on commercial cloud resources with that of the same application on dedicated local high performance computing resources. We show that we can achieve scalable and on-demand reconstruction of large scale pCT images for simultaneous multi-client requests, processing images in less than 10 minutes for less than $10 per image.

Cloud computing↗