Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Inference”

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 523 records · Page 29

A Variational Inference Approach to Inverse Problems with Gamma Hyperpriors

Hierarchical models with gamma hyperpriors provide a flexible, sparse-promoting framework to bridge L 1 and L 2 regularizations in Bayesian formulations to inverse problems. Despite the Bayesian motivation for these models, existing methodologies are limited to maximum a posteriori estimation. The potential to perform uncertainty quantification has not yet been realized. This paper introduces a variational iterative alternating scheme for hierarchical inverse problems with gamma hyperpriors. The proposed variational inference approach yields accurate reconstruction, provides meaningful uncertainty quantification, and is easy to implement. In addition, it lends itself naturally to conduct model selection for the choice of hyperparameters. Here, we illustrate the performance of our methodology in several computed examples, including a deconvolution problem and sparse identification of dynamical systems from time series data.

Bayesian shrinkage↗

Efficient Neural Network Approaches for Conditional Optimal Transport with Applications in Bayesian Inference

In this work, we present two neural network approaches that approximate the solutions of static and dynamic conditional optimal transport (COT) problems. Both approaches enable conditional sampling and conditional density estimation, which are core tasks in Bayesian inference—particularly in the simulation-based (“likelihood-free”) setting. Our methods represent the target conditional distribution as a transformation of a tractable reference distribution. Obtaining such a transformation, chosen here to be an approximation of the COT map, is computationally challenging even in moderate dimensions. To improve scalability, our numerical algorithms use neural networks to parameterize candidate maps and further exploit the structure of the COT problem. Our static approach approximates the map as the gradient of a partially input convex neural network. It uses a novel numerical implementation to increase computational efficiency compared to state-of-the-art alternatives. Our dynamic approach approximates the conditional optimal transport via the flow map of a regularized neural ODE; compared to the static approach, it is slower to train but offers more modeling choices and can lead to faster sampling. We demonstrate both algorithms numerically, comparing them with competing state-of-the-art approaches, using benchmark datasets and simulation-based Bayesian inverse problems.

97 MATHEMATICS AND COMPUTING↗

Relation Inference among Sensor Time Series in Smart Buildings with Metric Learning

Smart Building Technologies hold promise for better livability for residents and lower energy footprints. Yet, the rollout of these technologies, from demand response controls to fault detection and diagnosis, significantly lags behind and is impeded by the current practice of manual identification of sensing point relationships, e.g., how equipment is connected or which sensors are co-located in the same space. This manual process is still error-prone, albeit costly and laborious.We study relation inference among sensor time series. Our key insight is that, as equipment is connected or sensors co-locate in the same physical environment, they are affected by the same real-world events, e.g., a fan turning on or a person entering the room, thus exhibiting correlated changes in their time series data. To this end, we develop a deep metric learning solution that first converts the primitive sensor time series to the frequency domain, and then optimizes a representation of sensors that encodes their relations. Built upon the learned representation, our solution pinpoints the relationships among sensors via solving a combinatorial optimization problem. Extensive experiments on real-world buildings demonstrate the effectiveness of our solution.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

Bayes Inference Engine (BIE) Test Plan: TiltConeBeamCombo

The Bayes Inference Engine (BIE) is a general software tool intended to be used primarily in the analysis of radiographic data for density. An analysis is set up in the BIE by representing the problem as a collection of modules called glyphs. All glyphs to be used in creating the forward model of the experiment will be tested separately for a range of inputs. The testing of the adjoint code is performed in a separate document.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Bayes Inference Engine (BIE) Test Plan: TiltConeBeamComboDMLinear

The Bayes Inference Engine (BIE) is a general software tool intended to be used primarily in the analysis of radiographic data for density. An analysis is set up in the BIE by representing the problem as a collection of modules called glyphs. All glyphs to be used in creating the forward model of the experiment will be tested separately for a range of inputs.

97 MATHEMATICS AND COMPUTING↗

Bayes Inference Engine (BIE) Test Plan: TiltConeBeamComboDMGeometric

The Bayes Inference Engine (BIE) is a general software tool intended to be used primarily in the analysis of radiographic data for density. An analysis is set up in the BIE by representing the problem as a collection of modules called glyphs. All glyphs to be used in creating the forward model of the experiment will be tested separately for a range of inputs.

97 MATHEMATICS AND COMPUTING↗

Bayes Inference Engine (BIE) Test Plan: TiltConeBeamFVT

The Bayes Inference Engine (BIE) is a general software tool intended to be used primarily in the analysis of radiographic data for density. An analysis is set up in the BIE by representing the problem as a collection of modules called glyphs. All glyphs to be used in creating the forward model of the experiment will be tested separately for a range of inputs. The testing of the adjoint code is performed in a separate document.

97 MATHEMATICS AND COMPUTING↗

Scaling Inference Using Triton to Accelerate Particle Physics at the LHC and DUNE

DUNE and the LHC experiments consist of unique and cutting-edge particle detectors that create massive, complex, and rich datasets with billions of events. They require sophisticated algorithms to reconstruct and interpret the data. Modern machine learning algorithms provide a powerful toolset to detect and classify particles, from familiar image processing convolutional neural networks to newer graph neural network architectures. A full reconstruction of these particle collisions requires novel approaches to handle the computing challenge of processing so much raw data. In a series of studies, physicists from Fermilab, CERN, and university groups explored how to accelerate their data processing using the Triton Inference Server.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Bayesian Inference of State-Level COVID-19 Basic Reproduction Numbers across the United States

Although many persons in the United States have acquired immunity to COVID-19, either through vaccination or infection with SARS-CoV-2, COVID-19 will pose an ongoing threat to non-immune persons so long as disease transmission continues. We can estimate when sustained disease transmission will end in a population by calculating the population-specific basic reproduction number ℛ0, the expected number of secondary cases generated by an infected person in the absence of any interventions. The value of ℛ0 relates to a herd immunity threshold (HIT), which is given by 1−1/ℛ0. When the immune fraction of a population exceeds this threshold, sustained disease transmission becomes exponentially unlikely (barring mutations allowing SARS-CoV-2 to escape immunity). Here, we report state-level ℛ0 estimates obtained using Bayesian inference. Maximum a posteriori estimates range from 7.1 for New Jersey to 2.3 for Wyoming, indicating that disease transmission varies considerably across states and that reaching herd immunity will be more difficult in some states than others. ℛ0 estimates were obtained from compartmental models via the next-generation matrix approach after each model was parameterized using regional daily confirmed case reports of COVID-19 from 21 January 2020 to 21 June 2020. Our ℛ0 estimates characterize the infectiousness of ancestral strains, but they can be used to determine HITs for a distinct, currently dominant circulating strain, such as SARS-CoV-2 variant Delta (lineage B.1.617.2), if the relative infectiousness of the strain can be ascertained. On the basis of Delta-adjusted HITs, vaccination data, and seroprevalence survey data, we found that no state had achieved herd immunity as of 20 September 2021.

60 APPLIED LIFE SCIENCES↗

ABRF Proteome Informatics Research Group (iPRG) 2016 Study: Inferring Proteoforms from Bottom-up Proteomics Data

This report presents the results from the 2016 ABRF Proteome Informatics Research Group (iPRG) study on proteoform inference and FDR estimation from bottom-up proteomics data. For this study, three replicate Q Exactive Orbitrap LC-MS/MS datasets were generated from each of four E. coli samples spiked with different equimolar mixtures of small recombinant proteins selected to mimic pairs of homologous proteins. Participants were given raw data and a sequence file, and asked to identify the proteins and provide estimates on the false discovery rate at the proteoform level. As part of this study, we tested a new submission system with a format validator running on a virtual private server (VPS) and allowed methods to be provided as executable R Markdown or IPython Notebooks. The task was perceived as difficult, and only eight unique submissions were received, though those who participated did well, with no one method performing best on all samples. However, none of the submissions included a complete Markdown or Notebook, even though examples were provided. Future iPRG studies need to be more successful in promoting and encouraging participation. The VPS and submission validator easily scale to much larger numbers of participants in these types of studies. The unique “ground-truth” dataset for proteoform identification generated for this study is now available to the research community, as are the server-side scripts for validating and managing submissions.

Proteomics, proteoforms, inference, false discover↗

On Inferring Extinction Laws in Z -approximately 6 Quasars as Signatures of Supernova Dust

Unusual extinction curves of high-redshift QSOs have been taken as evidence that dust is primarily produced by supernovae at high redshift. In particular, the 3000 ˚A Todini-Ferrara-Maiolino kink in the extinction curve of the z = 6.20 SDSS J1048+4637 has been attributed to supernova dust. Here we discuss the challenges in inferring robust extinction curves of high-redshift QSOs and critically assess previous claims of detection of supernova dust. In particular, we address the sensitivity to the choice of intrinsic QSO spectrum, the need for a long wavelength baseline, and the drawbacks in fitting theoretical extinction curves. In a sample of 21 QSOs at z ∼ 6 we detect significant ultraviolet extinction using existing broad-band optical, near-infrared, and Spitzer photometry. The median extinction curve is consistent with a Small Magellanic Cloud curve with A1450 ∼ 0.7 mag and does not exhibit any conspicuous (restframe) 2175 ˚A or 3000 ˚A features. For two QSOs, SDSS J1044−0125 at z = 5.78 and SDSS J1030+0524 at z = 6.31, we further present X-shooter spectra covering the wavelength range 0.9-2.5 μm. The resulting non-parametric extinction curves do not exhibit the 3000 ˚A kink. Finally, in a re-analysis of literature spectra of SDSS J1048+4637, we do not find evidence for a conspicuous kink. We conclude that the existing evidence for a 3000 ˚A feature is weak and that the overall dust properties at high and low redshift show no significant differences. This, however, does not preclude supernovae from dominating the dust budget at high redshift.

Extinction↗

Pointer Adaptation and Pruning of Min-Max Fuzzy Inference and Estimation

A new technique for adaptation of fuzzy membership functions in a fuzzy inference system is proposed. The pointer technique relies upon the isolation of the specific membership functions that contributed to the final decision, followed by the updating of these functions' parameters using steepset descent.

Fuzzy↗

Regulon inference without arbitrary thresholds: three levels of sensitivity

Reconstruction of transcriptional regulatory networks is one of the major challenges facing the bioinformatics community in view of constantly growing number of complete genomes. The comparative genomics approach has been successfully used for the analysis of the transcriptional regulation of many metabolic systems in various bacteria taxa. The key step in this approach is given a position weight matrix, find an optimal threshold for the search of potential binding sites in genomes. In our previous work we proposed an approach for automatic selection of TFBS score threshold coupled with inference of regulon content. In this study we developed two modifications of this approach providing two additional levels of sensitivity.

99 GENERAL AND MISCELLANEOUS↗

Adaptive Neuro Fuzzy Inference System for Cyber-Intrusion Detection in a Smart Grid

The evolution of the power grid has brought increasing deployment of advance metering infrastructure, penetration of intelligent electronic devices, and integration of physical power system components with information and communications technologies. With the fast-expanding connectivity, cyber vulnerabilities arise due to the use of internet-based communication systems. These systems are targets of cyber-intrusions which attempt to disturb the normal power system functions. Traditional intrusion detection algorithms have been developed without an explicit model of the cyber components. In this paper, an algorithm to detect false data injections in the power system is proposed considering both cyber and physical models of the power system. The algorithm is based on an Adaptive Neuro Fuzzy Inference System (ANFIS) which collects information from state variables of the cyber-physical system to meet the performance requirements of the grid. Simulations of the proposed approach using the IEEE 13-bus test system validate the effectiveness of this artificial intelligence-based algorithm.

Adaptive Neuro Fuzzy Inference System (ANFIS), cyb↗

Rapid Inverse Parameter Inference Using Physics-Informed Neural Network

As Li-ion batteries become more essential in today's economy, tools need to be developed to accurately and rapidly diagnose a battery's internal state-of-health. Using a Li-ion battery's (high-rate) voltage response, it is proposed to determine a battery's internal state through Bayesian calibration. However, Bayesian calibration is notoriously slow and requires thousands of model runs. To accelerate parameter inference using Bayesian calibration, a surrogate model is developed to replace the underlying physics-based Li-ion model. Developing a surrogate model for rapid Bayesian calibration analysis is discussed for both the single particle model (SPM) and the pseudo two-dimensional (P2D) model. Surrogate models are constructed using physics-informed neural networks (PINNs) that encode the influence of internal properties on observed voltage responses. In practice, a neural network can be trained by: 1) using simulation results of the physics-based model (i.e., a data-loss approach); 2) using the residuals of the governing equations themselves (i.e., a physics-loss approach); or 3) using a combination of simulation results and governing equation residuals. In the present work, PINNs are developed using a variety of training losses and neural network architectures. In this analysis, it is shown that a PINN surrogate model can be reliably trained with only physics-informed loss. However, using a coupled data-informed and physics-loss approach produced the most accurate PINNs.

Bayesian calibration↗

Inferring magnetospheric heavy ion density using EMIC waves

Abstract We present a method to infer heavy ion concentration ratios from electromagnetic ion cyclotron (EMIC) wave observations that result from ion‐ion hybrid (IIH) resonance. A key feature of the IIH resonance is the concentration of wave energy in a field‐aligned resonant mode that exhibits linear polarization. These mode‐converted waves at the IIH resonance are localized at the location where the frequency of a compressional wave driver matches the IIH resonance condition, which depends sensitively on the heavy ion concentration. This dependence makes it possible to estimate the heavy ion concentration ratio. In this paper, we evaluate the absorption coefficients at the IIH resonance at Earth's geosynchronous orbit for variable concentrations of He + and wave frequencies using a dipole magnetic field model. We find that the resonance only occurs over a limited range of wave frequency such that the IIH resonance frequency is close to but not exactly the same as the crossover frequency. Using the wave absorption and EMIC waves observed from the GOES 12 satellite, we demonstrate how this technique can be used to estimate the He + concentration of around 4% near L = 6.6 assuming electron‐H + ‐He + plasma.

Kim, Eun‐Hwa↗

Contrasts in compliant fault zone properties inferred from geodetic measurements in the San Francisco Bay area

In crustal fault zones, regions of damaged rock characterized by reduced elastic shear modulus can influence patterns of near-field interseismic deformation. In order to study these compliant fault zones (CFZs) and how they might develop over the lifetimes of faults, we compare two fault segments with contrasting fault age and lithology along the San Andreas Fault in the San Francisco Bay Area. New geodetic measurements of the interseismic velocity fields at each location are used to constrain fault zone parameters through a Markov chain Monte Carlo method. At Black Mountain, in the Santa Cruz Mountains of the San Francisco Peninsula, we do not find evidence for a compliant fault zone; instead, we find that the geodetic data are more consistent with a model of a single fault in a homogeneous elastic half-space. At Lake San Andreas, a younger fault segment 35 km farther north, we find evidence for a compliant fault zone about 3.4 +1.1/-1.4 km wide, containing a shear modulus of about 40% of the shear modulus of the surrounding rock. We also find that the best fitting CFZ model at this location, unlike the best fitting homogeneous half-space model, has a locking depth that agrees well with the observed depth of microseismicity. Based on differences in fault age, cumulative displacement, and lithology between Black Mountain and Lake San Andreas, we infer that lithology plays an important and, in this case, perhaps a dominant role in the accumulation of fault zone damage structures and the development of CFZs over the lifetime of a fault.

58 GEOSCIENCES↗