Search NASA⌕ Search

SEARCH · Search NASA

Results for “Generative modeling”

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 253 records · Page 14

Performance and Architecture Lab Modeling Tool

Analytical application performance models are critical for diagnosing performance-limiting resources, optimizing systems, and designing machines. Creating models, however, is difficult. Furthermore, models are frequently expressed in forms that are hard to distribute and validate. The Performance and Architecture Lab Modeling tool, or Palm, is a modeling tool designed to make application modeling easier. Palm provides a source code modeling annotation language. Not only does the modeling language divide the modeling task into sub problems, it formally links an application's source code with its model. This link is important because a model's purpose is to capture application behavior. Furthermore, this link makes it possible to define rules for generating models according to source code organization. Palm generates hierarchical models according to well-defined rules. Given an application, a set of annotations, and a representative execution environment, Palm will generate the same model. A generated model is a an executable program whose constituent parts directly correspond to the modeled application. Palm generates models by combining top-down (human-provided) semantic insight with bottom-up static and dynamic analysis. A model's hierarchy is defined by static and dynamic source code structure. Because Palm coordinates models and source code, Palm's models are 'first-class' and reproducible. Palm automates common modeling tasks. For instance, Palm incorporates measurements to focus attention, represent constant behavior, and validate models. Palm's workflow is as follows. The workflow's input is source code annotated with Palm modeling annotations. The most important annotation models an instance of a block of code. Given annotated source code, the Palm Compiler produces executables and the Palm Monitor collects a representative performance profile. The Palm Generator synthesizes a model based on the static and dynamic mapping of annotations to program behavior. The model -- an executable program -- is a hierarchical composition of annotation functions, synthesized functions, statistics for runtime values, and performance measurements.

Tallent, Nathan↗

Next-Gen 3 : Sequencing, Modeling, and Advanced Biofuels - Final Technical Report

Successful, scalable implementation of biofuels is dependent on the efficient and near complete utilization of diverse biomass sources. One approach is to utilize the large recalcitrant biomass fraction (or any organic waste stream) through the thermochemical conversion of organic compounds to syngas, a mixture of carbon monoxide (CO), carbon dioxide (CO 2 ), and hydrogen (H 2 ), which can subsequently be metabolized by acetogenic microorganisms to produce next-gen biofuels. The goal of this proposal was to advance the development of the acetogen Clostridium ljungdahlii as a chassis organism for next-gen biofuel production from cheap, renewable sources and to detail the interconnectivity of metabolism, energy conservation, and regulation of acetogens using next-gen sequencing and next-gen modeling. To achieve this goal we determined optimization of carbon and energy utilization through differential translational efficiency in C. ljungdahlii. Furthermore, we reconstructed a next-generation model of all major cellular processes, such as macromolecular synthesis and transcriptional regulation and deployed this model to predicting proteome allocation, overflow metabolism, and metal requirements in this model acetogen. In addition we explored the evolutionary significance of tRNA operon structure using the next-gen model and determined the optimal operon structure for bioproduction. Our study substantially enhanced the knowledgebaase for chemolithoautotrophs and their potential for advanced biofuel production. It provides next-generation modeling capability, offer innovative tools for genome-scale engineering, and provide novel methods to utilize next-generation models for the design of tunable systems that produce commodity chemicals from inexpensive sources.

09 BIOMASS FUELS↗

The Aeras Next Generation Global Atmosphere Model

The Next Generation Global Atmosphere Model LDRD project developed a suite of atmosphere models: a shallow water model, an x-z hydrostatic model, and a 3D hydrostatic model, by using Albany, a finite element code. Albany provides access to a large suite of leading-edge Sandia high-performance computing technologies enabled by Trilinos, Dakota, and Sierra. The next-generation capabilities most relevant to a global atmosphere model are performance portability and embedded uncertainty quantification (UQ). Performance portability is the capability for a single code base to run efficiently on diverse set of advanced computing architectures, such as multi-core threading or GPUs. Embedded UQ refers to simulation algorithms that have been modified to aid in the quantifying of uncertainties. In our case, this means running multiple samples for an ensemble concurrently, and reaping certain performance benefits. We demonstrate the effectiveness of these approaches here as a prelude to introducing them into ACME.

54 ENVIRONMENTAL SCIENCES↗

Unsupervised quantum circuit learning in high energy physics

Unsupervised training of generative models is a machine learning task that has many applications in scientific computing. Here, in this work, we evaluate the efficacy of using quantum circuit-based generative models to generate synthetic data of high energy physics processes. We use nonadversarial, gradient-based training of quantum circuit Born machines to generate joint distributions over two and three variables.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Learning effective physical laws for generating cosmological hydrodynamics with Lagrangian deep learning

The goal of generative models is to learn the intricate relations between the data to create new simulated data, but current approaches fail in very high dimensions. When the true data-generating process is based on physical processes, these impose symmetries and constraints, and the generative model can be created by learning an effective description of the underlying physics, which enables scaling of the generative model to very high dimensions. In this work, we propose Lagrangian deep learning (LDL) for this purpose, applying it to learn outputs of cosmological hydrodynamical simulations. The model uses layers of Lagrangian displacements of particles describing the observables to learn the effective physical laws. The displacements are modeled as the gradient of an effective potential, which explicitly satisfies the translational and rotational invariance. The total number of learned parameters is only of order 10, and they can be viewed as effective theory parameters. Further, we combine N-body solver fast particle mesh (FastPM) with LDL and apply it to a wide range of cosmological outputs, from the dark matter to the stellar maps, gas density, and temperature. The computational cost of LDL is nearly four orders of magnitude lower than that of the full hydrodynamical simulations, yet it outperforms them at the same resolution. We achieve this with only of order 10 layers from the initial conditions to the final output, in contrast to typical cosmological simulations with thousands of time steps. This opens up the possibility of analyzing cosmological observations entirely within this framework, without the need for large dark-matter simulations.

79 ASTRONOMY AND ASTROPHYSICS↗

Magnetic Gears for a Marine Hydrokinetic Generator Component Model

The goal of this project is to design, fabricate, and test a hermetically sealed 50 kilowatt (kW) multistage magnetically geared generator (MGG). The Component Content Model provides data submitters with an easy and consistent means of uploading data and associated meta data about a component that is currently under development. The data fields include generic information about the component, technology classifications, current costs and performance, proposed target goals, and the environment that the component is operated in. These data are important to DOE and will be used to develop data products that provide quantitative information to guide and support programmatic decisions. Data will also be used by DOE in general assessments of MHK component readiness, performance, costs, and proposed plans. The ultimate goal is to use these data to perform research and tailor programs to best benefit the industry.

16 TIDAL AND WAVE POWER↗

Applying Particle Swarm Optimization and Extended Kalman Filtering to Model Kaplan Generation Dynamics for Hydropower Systems

Variable renewable generation is increasing the need for hydropower plants to provide fast and flexible grid support, which places new demands on plant-level dynamic models used for monitoring, control, and operational decision-making. This need is especially important for hydroelectric systems, where turbine and generator dynamics are strongly coupled, nonlinear, and time-varying, making accurate real-time representation difficult. To address this problem, this paper develops a digital twin (DT) framework for a synchronous generator–Kaplan turbine system using an explicit separation of slow turbine dynamics and fast generator dynamics. The turbine subsystem is represented by a six-coefficient model, whose parameters are identified offline using particle swarm optimization, while the generator subsystem is updated online through an extended Kalman filter for real-time state and parameter estimation. These models are integrated within a closed-loop simulation that includes a proportional–integral–derivative–double-derivative governor and excitation system, allowing the DT to track plant behavior under realistic operating conditions. Unlike prior studies that treat turbine and generator modeling separately or rely mainly on simulated inputs, the proposed framework is validated using real operational data from a hydropower plant. Results show that the DT reproduces terminal voltage, active power, and reactive power with a normalized root mean square error of approximately 5%. This hybrid offline–online formulation constitutes the main contribution of the work, providing an adaptive and practically deployable DT for hydropower systems with direct relevance to control improvement, performance monitoring, and grid-support applications under high renewable penetration.

13 HYDRO ENERGY↗

SU-G-TeP4-14: Quality Control of Treatment Planning Using Knowledge-Based Planning Across a System of Radiation Oncology Practices

Purpose: There is potentially a wide variation in plan quality for a certain disease site, even for clinics located in the same system of hospitals. We have used a prostate-specific knowledge-based planning (KBP) model as a quality control tool to investigate the variation in prostate treatment planning across a network of affiliated radiation oncology departments. Methods: A previously created KBP model was applied to 10 patients each from 4 community-based clinics (Clinics A, B, C, and D). The KBP model was developed using RapidPlan (Eclipse v13.5, Varian Medical Systems) from 60 prostate/prostate bed IMRT plans that were originally planned using an in-house treatment planning system at the central institution of the community-based clinics. The dosimetric plan quality (target coverage and normal-tissue sparing) of each model-generated plan was compared to the respective clinically-used plan. Each community-based clinic utilized the same planning goals to develop the clinically-used plans that were used at the main institution. Results: Across all 4 clinics, the model-generated plans decreased the mean dose to the rectum by varying amounts (on average, 12.5, 2.6, 4.5, and 2.7 Gy for Clinics A, B, C, and D, respectively). The mean dose to the bladder also decreased with the model-generated plans (5.4, 2.3, 3.0, and 4.1 Gy, respectively). The KBP model also identified that target coverage (D95%) improvements were possible for for Clinics A, B, and D (0.12, 1.65, and 2.75%) while target coverage decreased by 0.72% for Clinic C, demonstrating potentially different trade-offs made in clinical plans at different institutions. Conclusion: Quality control of dosimetric plan quality across a system of radiation oncology practices is possible with knowledge-based planning. By using a quality KBP model, smaller community-based clinics can potentially identify the areas of their treatment plans that may be improved, whether it be in normal-tissue sparing or improved target coverage. M. Matuszak has research funding for KBP from Varian Medical Systems.

60 APPLIED LIFE SCIENCES↗

Comparison of Doubly-Fed Induction Generator Machine Models for Real-Time Simulations

This work presents a detailed comparison among three implementations of the Doubly-Fed Induction Generator (DFIG) machine model. The alpha-beta-gamma, synchronous (dqo), and natural frame dynamic machine models were deployed on a Field-Programmable Gate Array (FPGA) and compared in terms of execution time and resource utilization. These machine models were also validated against a physical DFIG machine during balanced and unbalanced grid conditions. A benchmark is presented with the objective of selecting the machine model, orientation and state-space representation that is best suited for real-time simulations. The selected machine model is used to develop a real-time emulator of a Type-3 wind system. This emulator enables the study of wind turbine system in realtime using a validated platform while allowing controller testing without relying on expensive third-party solutions.

Ferrari Maglia, Max↗

Active learning for the design of polycrystalline textures using conditional normalizing flows

Generative modeling has opened new avenues for solving previously intractable materials design problems. However, these new opportunities are accompanied by a drastic increase in the required amount of training data. This is in stark juxtaposition to the high expense and difficulty in curating such large materials datasets. In this work, we propose a novel framework for integrating generative models within an active learning loop. Further, this enables the training of generative models with datasets significantly smaller than what has previously been demonstrated, providing a direct route for their application in data constrained environments. The functionality of this framework is then demonstrated by addressing the challenge of designing polycrystalline textures associated with target anisotropic mechanical properties. The developed protocol exhibited a cost reduction between 14 to 18 times over a randomly sampled experimental design.

36 MATERIALS SCIENCE↗

TestDose: A nuclear medicine software based on Monte Carlo modeling for generating gamma camera acquisitions and dosimetry

Purpose: The TestDose platform was developed to generate scintigraphic imaging protocols and associated dosimetry by Monte Carlo modeling. TestDose is part of a broader project (www.dositest.com) whose aim is to identify the biases induced by different clinical dosimetry protocols. Methods: The TestDose software allows handling the whole pipeline from virtual patient generation to resulting planar and SPECT images and dosimetry calculations. The originality of their approach relies on the implementation of functional segmentation for the anthropomorphic model representing a virtual patient. Two anthropomorphic models are currently available: 4D XCAT and ICRP 110. A pharmacokinetic model describes the biodistribution of a given radiopharmaceutical in each defined compartment at various time-points. The Monte Carlo simulation toolkit GATE offers the possibility to accurately simulate scintigraphic images and absorbed doses in volumes of interest. The TestDose platform relies on GATE to reproduce precisely any imaging protocol and to provide reference dosimetry. For image generation, TestDose stores user’s imaging requirements and generates automatically command files used as input for GATE. Each compartment is simulated only once and the resulting output is weighted using pharmacokinetic data. Resulting compartment projections are aggregated to obtain the final image. For dosimetry computation, emission data are stored in the platform database and relevant GATE input files are generated for the virtual patient model and associated pharmacokinetics. Results: Two samples of software runs are given to demonstrate the potential of TestDose. A clinical imaging protocol for the Octreoscan™ therapeutical treatment was implemented using the 4D XCAT model. Whole-body “step and shoot” acquisitions at different times postinjection and one SPECT acquisition were generated within reasonable computation times. Based on the same Octreoscan™ kinetics, a dosimetry computation performed on the ICRP 110 model is also presented. Conclusions: The proposed platform offers a generic framework to implement any scintigraphic imaging protocols and voxel/organ-based dosimetry computation. Thanks to the modular nature of TestDose, other imaging modalities could be supported in the future such as positron emission tomography.

60 APPLIED LIFE SCIENCES↗

Hierarchical Conditioning of Diffusion Models Using Tree-of-Life for Studying Species Evolution

A central problem in biology is to understand how organisms evolve and adapt to their environment by acquiring variations in the observable characteristics or traits of species across the tree of life. With the growing availability of large-scale image repositories in biology and recent advances in generative modeling, there is an opportunity to accelerate the discovery of evolutionary traits automatically from images. Toward this goal, we introduce Phylo-Diffusion, a novel framework for conditioning diffusion models with phylogenetic knowledge represented in the form of HIERarchical Embeddings (HIER-Embeds). We also propose two new experiments for perturbing the embedding space of Phylo-Diffusion: trait masking and trait swapping, inspired by counterpart experiments of gene knockout and gene editing/swapping. Our work represents a novel methodological advance in generative modeling to structure the embedding space of diffusion models using tree-based knowledge. Our work also opens a new chapter of research in evolutionary biology by using generative models to visualize evolutionary changes directly from images. We empirically demonstrate the usefulness of Phylo-Diffusion in capturing meaningful trait variations for fishes and birds, revealing novel insights about the biological mechanisms of their evolution. (Model and code can be found at imageomics.github.io/phylo-diffusion)

Khurana, Mridul↗

Denoising diffusion probabilistic models for generative alloy design

Inverse material design is an extremely challenging optimization task made difficult by, in part, the highly nonlinear relationship linking performance with composition. Quantitative approaches have improved significantly owing to advances in high throughput experimentation and computational thermodynamics. However, existing physics-based tools are mostly forward models; input a chemistry and obtain a prediction. More recently the materials community has leveraged advances in the machine learning community to establish novel inverse design frameworks. Very recently denoising diffusion probabilistic models have been shown to be extremely powerful generators producing synthetic data of various modalities e.g. images, text, audio, tables, etc.. In this work a novel framework for alloy design and optimization is proposed leveraging these class of models. Five key generative tasks are demonstrated (1) unconditional generation (2) composition conditioned generation (3) property conditioned generation (4) multi-feedstock conditioned generation and (5) generative optimization. These methods were tested on three case studies: high entropy alloy design, superalloy binder jet additive manufacturing, and in-situ dual-feedstock wire-arc additive manufacturing. Results indicate that the established models are extremely flexible, expressive, and robust. The architecture’s flexibility and training procedure empower the model to learn complex intra-compositional and composition-property relationships. Furthermore, the probabilistic nature of these models makes them well suited for addressing solution non-uniqueness and tackling uncertainty quantification tasks. While the fidelity and quantity of the underlying training data is paramount, we envision that future alloy design frameworks will make extensive use of these kinds of machine learning models as “search” tools bolstering the utility of experimental and computational approaches.

36 MATERIALS SCIENCE↗

Advancing set-conditional set generation: Diffusion models for fast simulation of reconstructed particles

The computational intensity of detector simulation and event reconstruction poses a significant difficulty for data analysis in collider experiments. This challenge inspires the continued development of machine learning techniques to serve as efficient surrogate models. We propose a fast emulation approach that combines simulation and reconstruction. In other words, a neural network generates a set of reconstructed objects conditioned on input particle sets. To make this possible, we advance set-conditional set generation with diffusion models. Using a realistic, generic, and public detector simulation and reconstruction package (COCOA), we show how diffusion models can accurately model the complex spectrum of reconstructed particles inside jets.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Enhancing generative molecular design via uncertainty-guided fine-tuning of variational autoencoders

In recent years, deep generative models have been successfully applied to various molecular design tasks, particularly in the life and materials sciences. One critical challenge for pre-trained generative molecular design (GMD) models is to fine-tune them to be better suited for downstream design tasks that aim at optimizing specific molecular properties. However, redesigning and training an existing effective generative model from scratch for each new design task are impractical. Furthermore, the black-box nature of typical downstream tasks that involve property prediction makes it nontrivial to optimize the generative model in a task-specific manner. In this work, we propose an uncertainty-guided fine-tuning strategy that can effectively enhance a pre-trained variational autoencoder (VAE) for GMD through performance feedback in an active learning setting. The strategy begins by quantifying the model uncertainty of the generative model using an efficient active subspace-based UQ (uncertainty quantification) scheme. Next, the decoder diversity within the characterized model uncertainty class is explored to expand the viable space of molecular generation. The low-dimensionality of the active subspace makes this exploration tractable using a black-box optimization scheme, which in turn enables us to identify and leverage a diverse set of high-performing models to generate enhanced molecules. Empirical results across six target molecular properties using multiple VAE-based generative models demonstrate that our uncertainty-guided fine-tuning strategy consistently leads to improved models that outperform the original pre-trained models.

97 MATHEMATICS AND COMPUTING↗

Generalizing to new geometries with Geometry-Aware Autoregressive Models (GAAMs) for fast calorimeter simulation

Generation of simulated detector response to collision products is crucial to data analysis in particle physics, but computationally very expensive. One subdetector, the calorimeter, dominates the computational time due to the high granularity of its cells and complexity of the interactions. Generative models can provide more rapid sample production, but currently require significant effort to optimize performance for specific detector geometries, often requiring many models to describe the varying cell sizes and arrangements, without the ability to generalize to other geometries. Here, we develop a geometry-aware autoregressive model, which learns how the calorimeter response varies with geometry, and is capable of generating simulated responses to unseen geometries without additional training. The geometry-aware model outperforms a baseline unaware model by over 50% in several metrics such as the Wasserstein distance between the generated and the true distributions of key quantities which summarize the simulated response. A single geometry-aware model could replace the hundreds of generative models currently designed for calorimeter simulation by physicists analyzing data collected at the Large Hadron Collider. This proof-of-concept study motivates the design of a foundational model that will be a crucial tool for the study of future detectors, dramatically reducing the large upfront investment usually needed to develop generative calorimeter models.

47 OTHER INSTRUMENTATION↗

The Bias-Variance-Correlation Tradeoff and Its Implications for ML Applications in HEP

The bias-variance tradeoff is a well-recognized phenomenon in statistics and machine learning. In this talk, I will discuss an extension, dubbed the bias-variance-correlation tradeoff. Roughly speaking, as the flexibility of a model decreases, the correlations in the outputs of a trained model for different inputs increases. Such correlations have implications for several applications of machine learning in high energy physics, e.g., the use generative models for event generation. In particular, I will argue that claims in the literature of data amplification by generative models stem from ignoring important correlations between the model's outputs for different inputs.

Shyamsundar, Prasanth [Fermilab] (ORCID:0000000227↗

Multi-Channel Large Network Simulation Including Adversarial Activity

Using network simulation to generate high fidelity networks is essential to test adversarial search problems, in addition to their benefits of privacy preservation and benchmarking purposes. Different generative models have been developed to generate single-channel, homogeneous networks that can model domains such as social networks, communication, and co-authorship. Modeling adversarial activity compounds the complexity of network simulation as it requires generating multi-channel networks simultaneously with correlated channel attributes at scale. We present a methodology along with scalable graph modeling and generation tools to produce realistic large-scale synthetic background activity graphs with embedded an adversarial activity. We present a concrete use case based on an attack scenario developed with the help of subject matter experts (SMEs). We discuss challenges in modeling integration across channels, scalability of the generative models, and integration across such models. We also discuss required aspects to simulate high fidelity multi-channel networks up-to an order of billion edges.

Network simulation, Multi-channel simulation, Adve↗