-
Machine Learning Visualization Tool for Exploring Parameterized Hydrodynamics
Authors:
C. F. Jekel,
D. M. Sterbentz,
T. M. Stitt,
P. Mocz,
R. N. Rieben,
D. A. White,
J. L. Belof
Abstract:
We are interested in the computational study of shock hydrodynamics, i.e. problems involving compressible solids, liquids, and gases that undergo large deformation. These problems are dynamic and nonlinear and can exhibit complex instabilities. Due to advances in high performance computing it is possible to parameterize a hydrodynamic problem and perform a computational study yielding…
▽ More
We are interested in the computational study of shock hydrodynamics, i.e. problems involving compressible solids, liquids, and gases that undergo large deformation. These problems are dynamic and nonlinear and can exhibit complex instabilities. Due to advances in high performance computing it is possible to parameterize a hydrodynamic problem and perform a computational study yielding $\mathcal{O}\left({\rm TB}\right)$ of simulation state data. We present an interactive machine learning tool that can be used to compress, browse, and interpolate these large simulation datasets. This tool allows computational scientists and researchers to quickly visualize "what-if" situations, perform sensitivity analyses, and optimize complex hydrodynamic experiments.
△ Less
Submitted 19 June, 2024;
originally announced June 2024.
-
Satyrn: A Platform for Analytics Augmented Generation
Authors:
Marko Sterbentz,
Cameron Barrie,
Shubham Shahi,
Abhratanu Dutta,
Donna Hooshmand,
Harper Pack,
Kristian J. Hammond
Abstract:
Large language models (LLMs) are capable of producing documents, and retrieval augmented generation (RAG) has shown itself to be a powerful method for improving accuracy without sacrificing fluency. However, not all information can be retrieved from text. We propose an approach that uses the analysis of structured data to generate fact sets that are used to guide generation in much the same way th…
▽ More
Large language models (LLMs) are capable of producing documents, and retrieval augmented generation (RAG) has shown itself to be a powerful method for improving accuracy without sacrificing fluency. However, not all information can be retrieved from text. We propose an approach that uses the analysis of structured data to generate fact sets that are used to guide generation in much the same way that retrieved documents are used in RAG. This analytics augmented generation (AAG) approach supports the ability to utilize standard analytic techniques to generate facts that are then converted to text and passed to an LLM. We present a neurosymbolic platform, Satyrn that leverages AAG to produce accurate, fluent, and coherent reports grounded in large scale databases. In our experiments, we find that Satyrn generates reports in which over 86% accurate claims while maintaining high levels of fluency and coherence, even when using smaller language models such as Mistral-7B, as compared to GPT-4 Code Interpreter in which just 57% of claims are accurate.
△ Less
Submitted 17 June, 2024;
originally announced June 2024.
-
Explosively driven Richtmyer--Meshkov instability jet suppression and enhancement via coupling machine learning and additive manufacturing
Authors:
Dane M. Sterbentz,
Dylan J. Kline,
Daniel A. White,
Charles F. Jekel,
Michael P. Hennessey,
David K. Amondson,
Abigail J. Wilson,
Max J. Sevcik,
Matthew F. L. Villena,
Steve S. Lin,
Michael D. Grapes,
Kyle T. Sullivan,
Jonathan L. Belof
Abstract:
The ability to control the behavior of fluid instabilities at material interfaces, such as the shock-driven Richtmyer--Meshkov instability, is a grand technological challenge with a broad number of applications ranging from inertial confinement fusion experiments to explosively driven shaped charges. In this work, we use a linear-geometry shaped charge as a means of studying methods for controllin…
▽ More
The ability to control the behavior of fluid instabilities at material interfaces, such as the shock-driven Richtmyer--Meshkov instability, is a grand technological challenge with a broad number of applications ranging from inertial confinement fusion experiments to explosively driven shaped charges. In this work, we use a linear-geometry shaped charge as a means of studying methods for controlling material jetting that results from the Richtmyer--Meshkov instability. A shaped charge produces a high-velocity jet by focusing the energy from the detonation of high explosives. The interaction of the resulting detonation wave with a hollowed cavity lined with a thin metal layer produces the unstable jetting effect. By modifying characteristics of the detonation wave prior to striking the lined cavity, the kinetic energy of the jet can be enhanced or reduced. Modifying the geometry of the liner material can also be used to alter jetting properties. We apply optimization methods to investigate several design parameterizations for both enhancing or suppressing the shaped-charge jet. This is accomplished using 2D and 3D hydrodynamic simulations to investigate the design space that we consider. We also apply new additive manufacturing methods for producing the shaped-charge assemblies, which allow for experimental testing of complicated design geometries obtained through computational optimization. We present a direct comparison of our optimized designs with experimental results carried out at the High Explosives Application Facility at Lawrence Livermore National Laboratory.
△ Less
Submitted 1 May, 2024;
originally announced May 2024.
-
A back-to-back diode model applied to MoS2 van der Waals Schottky diodes
Authors:
Jeffrey A. Cloninger,
Raine Harris,
Kristine L. Haley,
Randy M. Sterbentz,
Takashi Taniguchi,
Kenji Watanabe,
Joshua O. Island
Abstract:
The use of metal van der Waals contacts and the implicit reduction in Fermi-level pinning in contacted semiconductors has led to remarkable device optimizations. For example, using graphene as an electrical contact allows for tunable Schottky barriers in transistors and barristors. In this study, we present a double Schottky barrier model and apply it to barrier tunable all van der Waals transisto…
▽ More
The use of metal van der Waals contacts and the implicit reduction in Fermi-level pinning in contacted semiconductors has led to remarkable device optimizations. For example, using graphene as an electrical contact allows for tunable Schottky barriers in transistors and barristors. In this study, we present a double Schottky barrier model and apply it to barrier tunable all van der Waals transistors. In a molybdenum disulfide (MoS$_2$) transistor with graphene and few-layer graphene contacts, we find that the model can be applied to extract Schottky barrier heights that agree with the Schottky-Mott rule from simple two-terminal current-voltage measurements at room temperature. Furthermore, we show tunability of the Schottky barrier \textit{in-situ} using a regional contact gate. Our results show that a basic back-to-back diode model, applied to two terminal measurements, can capture the diode properties of all-van-der-Waals transistors relatively well.
△ Less
Submitted 5 February, 2024;
originally announced February 2024.
-
Lightweight Knowledge Representations for Automating Data Analysis
Authors:
Marko Sterbentz,
Cameron Barrie,
Donna Hooshmand,
Shubham Shahi,
Abhratanu Dutta,
Harper Pack,
Andong Li Zhao,
Andrew Paley,
Alexander Einarsson,
Kristian Hammond
Abstract:
The principal goal of data science is to derive meaningful information from data. To do this, data scientists develop a space of analytic possibilities and from it reach their information goals by using their knowledge of the domain, the available data, the operations that can be performed on those data, the algorithms/models that are fed the data, and how all of these facets interweave. In this w…
▽ More
The principal goal of data science is to derive meaningful information from data. To do this, data scientists develop a space of analytic possibilities and from it reach their information goals by using their knowledge of the domain, the available data, the operations that can be performed on those data, the algorithms/models that are fed the data, and how all of these facets interweave. In this work, we take the first steps towards automating a key aspect of the data science pipeline: data analysis. We present an extensible taxonomy of data analytic operations that scopes across domains and data, as well as a method for codifying domain-specific knowledge that links this analytics taxonomy to actual data. We validate the functionality of our analytics taxonomy by implementing a system that leverages it, alongside domain labelings for 8 distinct domains, to automatically generate a space of answerable questions and associated analytic plans. In this way, we produce information spaces over data that enable complex analyses and search over this data and pave the way for fully automated data analysis.
△ Less
Submitted 15 October, 2023;
originally announced November 2023.
-
Summarization from Leaderboards to Practice: Choosing A Representation Backbone and Ensuring Robustness
Authors:
David Demeter,
Oshin Agarwal,
Simon Ben Igeri,
Marko Sterbentz,
Neil Molino,
John M. Conroy,
Ani Nenkova
Abstract:
Academic literature does not give much guidance on how to build the best possible customer-facing summarization system from existing research components. Here we present analyses to inform the selection of a system backbone from popular models; we find that in both automatic and human evaluation, BART performs better than PEGASUS and T5. We also find that when applied cross-domain, summarizers exh…
▽ More
Academic literature does not give much guidance on how to build the best possible customer-facing summarization system from existing research components. Here we present analyses to inform the selection of a system backbone from popular models; we find that in both automatic and human evaluation, BART performs better than PEGASUS and T5. We also find that when applied cross-domain, summarizers exhibit considerably worse performance. At the same time, a system fine-tuned on heterogeneous domains performs well on all domains and will be most suitable for a broad-domain summarizer. Our work highlights the need for heterogeneous domain summarization benchmarks. We find considerable variation in system output that can be captured only with human evaluation and are thus unlikely to be reflected in standard leaderboards with only automatic evaluation.
△ Less
Submitted 18 June, 2023;
originally announced June 2023.
-
Suppression of Richtmyer-Meshkov instability via special pairs of shocks and phase transitions
Authors:
W. J. Schill,
M. R. Armstrong,
J. H. Nguyen,
D. M. Sterbentz,
D. A. White,
L. X. Benedict,
R. N. Rieben,
A. Hoff,
H. E. Lorenzana,
B. M. La Lone,
M. D. Staska,
J. L. Belof
Abstract:
The classical Richtmyer-Meshkov instability is a hydrodynamic instability characterizing the evolution of an interface following shock loading. In contrast to other hydrodynamic instabilities such as Rayleigh-Taylor, it is known for being unconditionally unstable: regardless of the direction of shock passage, any deviations from a flat interface will be amplified. In this article, we show that for…
▽ More
The classical Richtmyer-Meshkov instability is a hydrodynamic instability characterizing the evolution of an interface following shock loading. In contrast to other hydrodynamic instabilities such as Rayleigh-Taylor, it is known for being unconditionally unstable: regardless of the direction of shock passage, any deviations from a flat interface will be amplified. In this article, we show that for negative Atwood numbers, there exist special sequences of shocks which result in a nearly perfectly suppressed instability growth. We demonstrate this principle computationally and experimentally with stepped fliers and phase transition materials. A fascinating immediate corollary is that in specific instances a phase transitioning material may self-suppress RMI.
△ Less
Submitted 23 March, 2023; v1 submitted 22 March, 2023;
originally announced March 2023.
-
Using Conservation Laws to Infer Deep Learning Model Accuracy of Richtmyer-meshkov Instabilities
Authors:
Charles F. Jekel,
Dane M. Sterbentz,
Sylvie Aubry,
Youngsoo Choi,
Daniel A. White,
Jonathan L. Belof
Abstract:
Richtmyer-Meshkov Instability (RMI) is a complicated phenomenon that occurs when a shockwave passes through a perturbed interface. Over a thousand hydrodynamic simulations were performed to study the formation of RMI for a parameterized high velocity impact. Deep learning was used to learn the temporal map** of initial geometric perturbations to the full-field hydrodynamic solutions of density a…
▽ More
Richtmyer-Meshkov Instability (RMI) is a complicated phenomenon that occurs when a shockwave passes through a perturbed interface. Over a thousand hydrodynamic simulations were performed to study the formation of RMI for a parameterized high velocity impact. Deep learning was used to learn the temporal map** of initial geometric perturbations to the full-field hydrodynamic solutions of density and velocity. The continuity equation was used to include physical information into the loss function, however only resulted in very minor improvements at the cost of additional training complexity. Predictions from the deep learning model appear to accurately capture temporal RMI formations for a variety of geometric conditions within the domain. First principle physical laws were investigated to infer the accuracy of the model's predictive capability. While the continuity equation appeared to show no correlation with the accuracy of the model, conservation of mass and momentum were weakly correlated with accuracy. Since conservation laws can be quickly calculated from the deep learning model, they may be useful in applications where a relative accuracy measure is needed.
△ Less
Submitted 18 July, 2022;
originally announced August 2022.
-
Requirements for Open Political Information: Transparency Beyond Open Data
Authors:
Andong Luis Li Zhao,
Andrew Paley,
Rachel Adler,
Harper Pack,
Sergio Servantez,
Alexander Einarsson,
Cameron Barrie,
Marko Sterbentz,
Kristian Hammond
Abstract:
A politically informed citizenry is imperative for a welldeveloped democracy. While the US government has pursued policies for open data, these efforts have been insufficient in achieving an open government because only people with technical and domain knowledge can access information in the data. In this work, we conduct user interviews to identify wants and needs among stakeholders. We further u…
▽ More
A politically informed citizenry is imperative for a welldeveloped democracy. While the US government has pursued policies for open data, these efforts have been insufficient in achieving an open government because only people with technical and domain knowledge can access information in the data. In this work, we conduct user interviews to identify wants and needs among stakeholders. We further use this information to sketch out the foundational requirements for a functional political information technical system.
△ Less
Submitted 6 December, 2021;
originally announced December 2021.
-
Universal image segmentation for optical identification of 2D materials
Authors:
Randy M. Sterbentz,
Kristine L. Haley,
Joshua O. Island
Abstract:
Machine learning methods are changing the way data is analyzed. One of the most powerful and widespread applications of these techniques is in image segmentation wherein disparate objects of a digital image are partitioned and classified. Here we present an image segmentation program incorporating a series of unsupervised clustering algorithms for the automatic thickness identification of two-dime…
▽ More
Machine learning methods are changing the way data is analyzed. One of the most powerful and widespread applications of these techniques is in image segmentation wherein disparate objects of a digital image are partitioned and classified. Here we present an image segmentation program incorporating a series of unsupervised clustering algorithms for the automatic thickness identification of two-dimensional materials from digital optical microscopy images. The program identifies mono- and few-layer flakes of a variety of materials on both opaque and transparent substrates with a pixel accuracy of roughly 95%. Contrasting with previous attempts, application generality is achieved through preservation and analysis of all three digital color channels and Gaussian mixture model fits to arbitrarily shaped data clusters. Our results provide a facile implementation of data clustering for the universal, automatic identification of two-dimensional materials exfoliated onto any substrate.
△ Less
Submitted 17 March, 2021;
originally announced March 2021.
