Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Spreading models”

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 289 records · Page 16
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Photometric and Kinematic Study of the Open Clusters SAI 44 and SAI 45

We carry out a detailed photometric and kinematic study of the poorly studied sparse open clusters SAI 44 and SAI 45 using ground-based BVR {sub c} I {sub c} data supplemented by archival data from Gaia eDR3 and Pan-STARRS. The stellar memberships are determined using a statistical method based on Gaia eDR3 kinematic data, and we found 204 members in SAI 44 while only 74 members are identified in SAI 45. The average distances to SAI 44 and SAI 45 are calculated to be 3670 ± 184 and 1668 ± 47 pc. The logarithmic age of the clusters are determined to be 8.82 ± 0.10 and 9.07 ± 0.10 yr for SAI 44 and SAI 45, respectively. The color–magnitude diagram of SAI 45 hosts an extended main-sequence turnoff (eMSTO). The apparent age spread is found to be similar to the apparent age spread predicted on the basis of the age spread and cluster age relation predicted by rotation models. This indicates that eMSTO is a stellar evolution rather than star formation phenomenon in SAI 45. We conclude that eMSTO in SAI 45 is mainly caused by the different rotation rates of stars as the SYCLIST synthetic population with different rotation rates was able to reproduce the observed eMSTO, and stars in the red part of the eMSTO were preferentially concentrated in the inner region, which again hints at different rotations being the reason for the extension in the upper MS. This finding supports the theory attributing the origin of eMSTO to the different rotations of eMSTO stars. The mass function slopes are obtained as -2.24 ± 0.66 and -2.58 ± 3.20 in the mass rages 2.426–0.990 M{sub ⊙} and 2.167–1.202 M{sub ⊙} for SAI 44 and SAI 45, respectively. SAI 44 exhibits the signature of mass segregation while we found weak evidence of mass segregation in SAI 45 possibly due to tidal stripping. The dynamical relaxation times of these clusters indicate that both clusters are in a dynamically relaxed state. Using the AD-diagram method, the apex coordinates are found to be ([[Formula]], -[[Formula]]) for SAI 44 and (-[[Formula]], -[[Formula]]) for SAI 45. The average space velocity components of the clusters SAI 44 and SAI 45 are calculated in units of km s{sup -1} as (-15.14 ± 3.90, -19.43 ± 4.41, -20.85 ± 4.57) and (28.13 ± 5.30, -9.78 ± 3.13, -19.59 ± 4.43), respectively.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2022

Dissecting Nearby Galaxies with piXedfit. I. Spatially Resolved Properties of Stars, Dust, and Gas as Revealed by Panchromatic SED Fitting

We study spatially resolved properties (on spatial scales of ~1–2 kpc out to at least 3 effective radii) of the stars, dust, and gas in 10 nearby spiral galaxies. The properties of the stellar population and dust are derived by fitting the spatially resolved spectral energy distribution (SED) with more than 20 photometric bands ranging from far-ultraviolet to far-infrared. Our newly developed software piXedfit performs point-spread function matching of images, pixel binning, and models the stellar light, dust attenuation, dust emission, and emission from a dusty torus heated by an active galactic nucleus simultaneously through the energy-balance approach. With this self-consistent analysis, we present the spatially resolved version of the IRX–βrelation, finding that it is consistent with the relationship from the integrated photometry. We show that the old stellar populations contribute to the dust heating, which causes an overestimation of the star formation rate (SFR) derived from the total ultraviolet and infrared luminosities on kiloparsec scales. With archival high-resolution maps of atomic and molecular gas, we study the radial variation of the properties of the stellar populations (including stellar mass, age, metallicity, and SFR), dust (including dust mass, dust temperature, and abundance of polycyclic aromatic hydrocarbon), and gas, as well as dust-to-stellar mass and dust-to-gas mass ratios. We observe a depletion of the molecular gas mass fraction in the central region of the majority of the galaxies, suggesting that the lack of available fuel is an important factor in suppressing the specific SFR at the center.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

Spatially distributed infection increases viral load in a computational model of SARS-CoV-2 lung infection

A key question in SARS-CoV-2 infection is why viral loads and patient outcomes vary dramatically across individuals. Because spatial-temporal dynamics of viral spread and immune response are challenging to study in vivo, we developed Spatial Immune Model of Coronavirus (SIMCoV), a scalable computational model that simulates hundreds of millions of lung cells, including respiratory epithelial cells and T cells. SIMCoV replicates viral growth dynamics observed in patients and shows how spatially dispersed infections can lead to increased viral loads. The model also shows how the timing and strength of the T cell response can affect viral persistence, oscillations, and control. By incorporating spatial interactions, SIMCoV provides a parsimonious explanation for the dramatically different viral load trajectories among patients by varying only the number of initial sites of infection and the magnitude and timing of the T cell immune response. When the branching airway structure of the lung is explicitly represented, we find that virus spreads faster than in a 2D layer of epithelial cells, but much more slowly than in an undifferentiated 3D grid or in a well-mixed differential equation model. These results illustrate how realistic, spatially explicit computational models can improve understanding of within-host dynamics of SARS-CoV-2 infection.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2023

Swift and extensive Omicron outbreak in China after sudden exit from ‘zero-COVID’ policy

In late 2022, China transitioned from a strict ‘zero-COVID’ policy to rapidly abandoning nearly all interventions and data reporting. This raised great concern about the presumably-rapid but unreported spread of the SARS-CoV-2 Omicron variant in a very large population of very low pre-existing immunity. By modeling a combination of case count and survey data, we show that Omicron spread extremely rapidly, at a rate of 0.42/day (95% credibility interval: [0.35, 0.51]/day), translating to an epidemic doubling time of 1.6 days ([1.6, 2.0] days) after the full exit from zero-COVID on Dec. 7, 2022. Consequently, we estimate that the vast majority of the population (97% [95%, 99%], sensitivity analysis lower limit of 90%) was infected during December, with the nation-wide epidemic peaking on Dec. 23. Overall, our results highlight the extremely high transmissibility of the variant and the importance of proper design of intervention exit strategies to avoid large infection waves.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2026

Comparison of structurally diverse simulation models for prediction of epidemic outcomes caused by a long-distance dispersed pathogen

Long-distance dispersal (LDD) pathogens pose substantial challenges for epidemic control due to their ability to generate new infection foci at great distances. While various modeling approaches have been developed to understand and manage such outbreaks, little work has compared how models of different structures behave under shared conditions. Here, in this study, we compare four structurally distinct epidemiological models — EPIMUL, GEMF, PoPS, and Warwick — each adapted to simulate the spread of wheat stripe rust (WSR), a wind-dispersed LDD pathogen, under identical epidemiological parameters and dispersal kernel. Using data from a controlled field experiment, we evaluate the ability of each model to replicate disease prevalence under nine intervention scenarios that vary in timing and culling area. While the models differ substantially in design — ranging from spatial grid-based to network-based and raster-based frameworks — the shared dispersal kernel allowed for close alignment in their predictions. All models accurately captured general epidemic trends, particularly the strong effect of early intervention on disease suppression. We qualitatively compared their behavioral responses across scenarios and also evaluated an ensemble prediction by averaging across model outputs. Our findings highlight how integrating shared epidemiological components into distinct modeling frameworks can improve consistency and accuracy, while reinforcing the importance of early culling in managing LDD pathogen outbreaks.

Dispersal kernel↗
DOE OSTI2024

Droplet bioprinting of acellular and cell-laden structures at high-resolutions

Advances in digital light projection(DLP) based (bio) printers have made printing of intricate structures at high resolution possible using a wide range of photosensitive bioinks. A typical setup of a DLP bioprinter includes a vat or reservoir filled with liquid bioink, which presents challenges in terms of cost associated with bioink synthesis, high waste, and gravity-induced cell settling, contaminations, or variation in bioink viscosity during the printing process. Here, we report a vat-free, low-volume, waste-free droplet bioprinting method capable of rapidly printing 3D soft structures at high resolution using model bioinks and model cells. A multiphase many-body dissipative particle dynamics model was developed to simulate the dynamic process of droplet-based DLP printing and elucidate the roles of surface wettability and bioink viscosity. Process variables such as light intensity, photo-initiator concentration, and bioink formulations were optimized to print 3D soft structures (∼0.4–3 kPa) with a typical layer thickness of 50 µm, an XY resolution of 38 ± 1.5 μm and Z resolution of 237 ± 5.4 µm. To demonstrate its versatility, droplet bioprinting was used to print a range of acellular 3D structures such as a lattice cube, a Mayan pyramid, a heart-shaped structure, and a microfluidic chip with endothelialized channels. Droplet bioprinting, performed using model C3H/10T1/2 cells, exhibited high viability (90%) and cell spreading. Additionally, microfluidic devices with internal channel networks lined with endothelial cells showed robust monolayer formation while osteoblast-laden constructs showed mineral deposition upon osteogenic induction. Overall, droplet bioprinting could be a low-cost, no-waste, easy-to-use, method to make customized bioprinted constructs for a range of biomedical applications.

DLP↗
DOE OSTI2024

KCl-UCl 3 molten salts investigated by Ab Initio Molecular Dynamics (AIMD) simulations

Ab Initio Molecular Dynamics (AIMD) simulations are performed on molten KCl-UCl3 salt mixtures to determine energies, heat capacities, and densities. The density-dependent energy correction (DFT-dDsC), Grimme et al.'s DFT-D3, and Langreth & Lundqvist (vdW-cx) models are used for dispersion forces and combined with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation potential with a Hubbard U parameter for the 5f electrons of uranium. After validating predictions for the end-member systems to literature data, KCl-UCl 3 mixtures are studied at select temperatures. Densities and energies both deviate from ideal solution behavior, with the maximum deviation occurring around 36% UCl 3 for mixing energies and slightly lower (29% UCl3) for densities. Compared to the NaCl-UCl 3 system, which was previously investigated using the same simulation methodologies, the KCl-UCl 3 density and mixing energy deviations from ideal solution behavior are larger by almost a factor of two. No deviation from ideal solution behavior for heat capacity was observed. The AIMD predictions for mixing energies and densities agree qualitatively with experimental data, though the spread in data obtained from the various dispersion force models utilized, measurements, and empirical estimates makes strong conclusions difficult. The dependence of thermodynamic and thermophysical properties on composition is correlated with the local chemistry of the solution phase, in particular, the tendency of UCl 3 to form network structures.

36 MATERIALS SCIENCE↗
DOE OSTI2024

KCl-UCl 3 molten salts investigated by Ab Initio Molecular Dynamics (AIMD) simulations: A comparative study with three dispersion models

Ab Initio Molecular Dynamics (AIMD) simulations are performed on molten KCl-UCl 3 salt mixtures to determine energies, heat capacities, and densities. The density-dependent energy correction (DFT-dDsC), Grimme et al.’s DFT-D3, and Langreth & Lundqvist (vdW-cx) models are used for dispersion forces and combined with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation potential with a Hubbard U param eter for the 5$f$ electrons of uranium. After validating predictions for the end-member systems to literature data, KCl-UCl 3 mixtures are studied at select temperatures. Densities and energies both deviate from ideal solution behavior, with the maximum deviation occurring around 36% UCl 3 for mixing energies and slightly lower (29% UCl 3 ) for densities. Compared to the NaCl-UCl 3 system, which was previously investigated using the same simulation methodologies, the KCl-UCl 3 density and mixing energy deviations from ideal solution behavior are larger by almost a factor of two. No deviation from ideal solution behavior for heat capacity was observed. The AIMD predictions for mixing energies and densities agree qualitatively with experimental data, though the spread in data obtained from the various dispersion force models utilized, measurements, and empirical estimates makes strong conclusions difficult. The dependence of thermodynamic and thermophysical properties on composition is correlated with the local chemistry of the solution phase, in particular, the tendency of UCl 3 to form network structures.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2023

Development and formulation of physics based metallic fuel models and comparison to integral irradiation data

Metallic fuel has an important historical significance in the development of nuclear reactors and continues to be relevant to the progression of advanced test and power reactors. A number of models, ranging from empirical to mechanistic, have been developed and implemented in various fuel performance codes to describe U-Zr and U-Pu-Zr fuel and typical fast reactor cladding materials. One challenge of using these models to simulate fuel performance is the inevitable tangling of coupled phenomena that can cloud proper implementation, calibration, and eventual utilization of new models. Here in an effort to provide a baseline capability that will facilitate the use of advanced models, new capabilities have been implemented into the fuel performance code BISON specific to metallic fuel simulations, ranging from materials properties, fission gas release and swelling calculations, coolant channel models, and cladding correlations. These models have been applied to the X441/X441A EBR-II experimental assembly data, a set of irradiated metallic UPuZr fuel rods of varying pin designs. The models implemented in BISON are able to capture the general trend of the expected response of the fuel and cladding to irradiation in EBR-II, especially when considering the spread in experimental measurements and the uncertainties inherited from the historical material models. Ultimately, the models outlined here provide the baseline capabilities on which new models can build upon in order to improve the prediction of metallic fuel performance simulations in off-normal designs or operations.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2022

Separating the Influences of Low-Latitude Warming and Sea Ice Loss on Northern Hemisphere Climate Change

Abstract Analyzing a multimodel ensemble of coupled climate model simulations forced with Arctic sea ice loss using a two-parameter pattern-scaling technique to remove the cross-coupling between low- and high-latitude responses, the sensitivity to high-latitude sea ice loss is isolated and contrasted to the sensitivity to low-latitude warming. Despite some differences in experimental design, the Northern Hemisphere near-surface atmospheric sensitivity to sea ice loss is found to be robust across models in the cold season; however, a larger intermodel spread is found at the surface in boreal summer, and in the free tropospheric circulation. In contrast, the sensitivity to low-latitude warming is most robust in the free troposphere and in the warm season, with more intermodel spread in the surface ocean and surface heat flux over the Northern Hemisphere. The robust signals associated with sea ice loss include upward turbulent and longwave heat fluxes where sea ice is lost, warming and freshening of the Arctic Ocean, warming of the eastern North Pacific Ocean relative to the western North Pacific with upward turbulent heat fluxes in the Kuroshio Extension, and salinification of the shallow shelf seas of the Arctic Ocean alongside freshening in the subpolar North Atlantic Ocean. In contrast, the robust signals associated with low-latitude warming include intensified ocean warming and upward latent heat fluxes near the western boundary currents, freshening of the Pacific Ocean, salinification of the North Atlantic, and downward sensible and longwave fluxes over the ocean.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2024

GraMeR: Gra ph Me ta R einforcement learning for multi-objective influence maximization

Influence maximization (IM) is a combinatorial problem of identifying a subset of seed nodes in a network (graph), which when activated, provide a maximal spread of influence in the network for a given diffusion model and a budget for seed set size. IM has numerous applications such as viral marketing, epidemic control, sensor placement and other network-related tasks. However, its practical uses are limited due to the computational complexity of current algorithms. Recently, deep reinforcement learning has been leveraged to solve IM in order to ease the computational burden. However, there are serious limitations in current approaches, including narrow IM formulation that only consider influence via spread and ignore self-activation, low scalability to large graphs, and lack of generalizability across graph families leading to a large running time for every test network. In this work, we address these limitations through a unique approach that involves: (1) Formulating a generic IM problem as a Markov decision process that handles both intrinsic and influence activations; (2)incorporating generalizability via meta-learning across graph families. There are previous works that combine deep reinforcement learning with graph neural network, but this work solves a more realistic IM problem and incorporates generalizability across graphs via meta reinforcement learning. Extensive experiments are carried out in various standard networks to validate performance of the proposed Graph Meta Reinforcement learning (GraMeR) framework. Finally, the results indicate that GraMeR is multiple orders faster and generic than conventional approaches when applied on small to medium scale graphs.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Assessing mean climate change signals in the global CORDEX-CORE ensemble

The new Coordinated Output for Regional Evaluations (CORDEX-CORE) ensemble provides high-resolution, consistent regional climate change projections for the major inhabited areas of the world. It serves as a solid scientific basis for further research related to vulnerability, impact, adaptation and climate services in addition to existing CORDEX simulations. The aim of this study is to investigate and document the climate change information provided by the CORDEX-CORE simulation ensemble, as a part of the World Climate Research Programme (WCRP) CORDEX community. An overview of the annual and monthly mean climate change information in selected regions in different CORDEX domains is presented for temperature and precipitation, providing the foundation for detailed follow-up studies and applications. Initially, two regional climate models (RCMs), REMO and RegCM were used to downscale global climate model output. The driving simulations by AR5 global climate models (AR5-GCMs) were selected to cover the spread of high, medium, and low equilibrium climate sensitivity at a global scale. The CORDEX-CORE ensemble has doubled the spatial resolution compared to the previously existing CORDEX simulations in most of the regions (25 $\,\mathrm {km}$ (0.22 $^{\circ }$ /jats:inline-formula>) versus 50 $\,\mathrm {km}$ (0.44 $^{\circ }$ )) leading to a potentially improved representation of, e.g., physical processes in the RCMs. The analysis focuses on changes in the IPCC physical climate reference regions. The results show a general reasonable representation of the spread of the temperature and precipitation climate change signals of the AR5-GCMs by the CORDEX-CORE simulations in the investigated regions in all CORDEX domains by mostly covering the AR5 interquartile range of climate change signals. The simulated CORDEX-CORE monthly climate change signals mostly follow the AR5-GCMs, although for specific regions they show a different change in the course of the year compared to the AR5-GCMs, especially for RCP8.5, which needs to be investigated further in region specific process studies.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2020

The impact of light polarization effects on weak lensing systematics

ABSTRACT A fraction of the light observed from edge-on disc galaxies is polarized due to two physical effects: selective extinction by dust grains aligned with the magnetic field and scattering of the anisotropic starlight field. Since the reflection and transmission coefficients of the reflecting and refracting surfaces in an optical system depend on the polarization of incoming rays, this optical polarization produces both (a) a selection bias in favour of galaxies with specific orientations and (b) a polarization-dependent point spread function (PSF). In this work, we build toy models to obtain for the first time an estimate for the impact of polarization on PSF shapes and the impact of the selection bias due to the polarization effect on the measurement of the ellipticity used in shear measurements. In particular, we are interested in determining if this effect will be significant for Wide-Field Infrared Survey Telescope (WFIRST). We show that the systematic uncertainties in the ellipticity components are 8 × 10−5 and 1.1 × 10−4 due to the selection bias and PSF errors respectively. Compared to the overall requirements on knowledge of the WFIRST PSF ellipticity (4.7 × 10−4 per component), both of these systematic uncertainties are sufficiently close to the WFIRST tolerance level that more detailed studies of the polarization effects or more stringent requirements on polarization-sensitive instrumentation for WFIRST are required.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2023

Emissions and Flame Structure Assessment of Aeroengine Micromixing Injectors for Lean Direct Injection of Hydrogen and Hydrogen/Natural Gas Blends

Abstract The present study focuses on whether adapting low emissions liquid fuel injection concepts can be applied to gaseous mixtures. The goal is to develop an injector with specific geometric characteristics that will achieve the lowest emissions. This is conducted by evaluating emissions and flame diagnostics on a set of test hardware that has been developed for combustion of pure hydrogen and hydrogen/methane mixtures. The current test hardware adapts lean direct injection concepts from aeroengine injectors to a ground-based engine based on target flows associated with a Solar Turbines Industrial Engine gas turbine. A statistical test plan was developed to assess these geometric characteristics using a three level Box Behnken design. The test plan investigates the effect of injector flow splits, air swirl, fuel swirl, pressure drop, preheat temperature, fuel composition, and the adiabatic flame temperature on the flame structure and emissions performance. A total of 16 injectors were considered for a total of 688 test points for emission and imaging data. Emissions data were reported on a mass basis in ng/J and chemical reactor network analysis were conducted to further interpret emissions behavior. Images were also captured to analyze the flame structure features of interest including reaction volume, flame length, flame spread, and flame width. With these results, a predictive model for emissions was made to conclude the most optimal injector configuration is to have the highest level of air split (more volume of air through the injector plate versus the injector) and air swirl angle for both methane and hydrogen gas mixtures. The predictive emissions level for all of these cases were below 10 ng/J.

08 HYDROGEN↗
DOE OSTI2026

Transport of 12 GeV positron beams at Ce+BAF

Jefferson Lab (JLab) is developing a concept to upgrade the Continuous Electron Beam Accelerator Facility (CEBAF) to additionally deliver spin-polarized continuous-wave positron beams for its nuclear physics program users (Ce+BAF 12 GeV). The concept involves repurposing the Low Energy Recirculator Facility (LERF) at JLab as a dual injector, first producing 100-300 MeV spin-polarized electron beams which are subsequently used for the generation and formation of 123 MeV continuous-wave positron beams. The positron beams are transported to CEBAF and injected for acceleration up to 12 GeV, tailored to the requirements of its four experimental halls. Given the higher emittance of the secondary positron beams, the CEBAF optics are optimized for low dispersion and low beta functions to enhance transmission within the Ce+BAF acceptance limits and with an R56 to manage the positron beams bunch length and energy spread. Potential bottlenecks are being investigated through both optical modeling and measurements using an electron beam, as well as degraded electron beams, to map the 6d acceptance of CEBAF as it is today. This presentation shares preliminary results from multi-particle tracking simulations of the positron beam up to 12 GeV, including spatial, momentum, and spin characteristics, and explores the feasibility of delivering beams simultaneously to multiple experimental halls via extraction optics.

Accelerator Physics↗
DOE OSTI2023

The evolution of the Milky Way’s thin disc radial metallicity gradient with K2 asteroseismic ages

ABSTRACT The radial metallicity distribution of the Milky Way’s disc is an important observational constraint for models of the formation and evolution of our Galaxy. It informs our understanding of the chemical enrichment of the Galactic disc and the dynamical processes therein, particularly radial migration. We investigate how the metallicity changes with guiding radius in the thin disc using a sample of red giant stars with robust astrometric, spectroscopic, and asteroseismic parameters. Our sample contains 668 stars with guiding radii 4 < Rg < 11 kpc and asteroseismic ages covering the whole history of the thin disc with precision ${\approx} 25 {{\, \rm per\ cent}}$. We use MCMC analysis to measure the gradient and its intrinsic spread in bins of age and construct a hierarchical Bayesian model to investigate the evolution of these parameters independently of the bins. We find a smooth evolution of the gradient from ≈−0.07 dex kpc−1 in the youngest stars to ≈−0.04 dex kpc−1 in stars older than 10 Gyr, with no break at intermediate ages. Our results are consistent with those based on asteroseismic ages from CoRoT, with that found in Cepheid variables for stars younger than 1 Gyr, and with open clusters for stars younger than 6 Gyr. For older stars we find a significantly lower metallicity in our sample than in the clusters, suggesting a survival bias favouring more metal-rich clusters. We also find that the chemical evolution model of Chiappini '09 is too metal poor in the early stages of disc formation. Our results provide strong new constraints for the growth and enrichment of the thin disc and radial migration, which will facilitate new tests of model conditions and physics.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2021

Evaluating Deception Detection Model Robustness To Linguistic Variation

With the increasing use of automated, machine learning-driven tools and the downstream impact that algorithmic judgements can have, it is critical to develop models that are robust to evolving or manipulated inputs. Evaluating the reliability of multimodal models across linguistic variations to understand model susceptibility to intentional linguistic adversarial attacks as well as natural linguistic variations is essential in this pursuit. We present extensive analysis of model robustness and susceptibility to linguistic variations in the setting of deceptive news detection, a difficult classification task that is an increasingly important problem to solve with the impact of misinformation spread online. We evaluate the effectiveness of incorporating adversarial defense strategies and measure model susceptibility to state-of-the-art adversarial attacks using two types of linguistic attacks — character and word perturbations. We consider two multiclass prediction tasks — a 3-way classification of tweets as trustworthy, propaganda, or disinformation; and a 4-way classification as clickbait, hoax, satire, or conspiracy — and compare the performance of three embeddings that have been state-of-the-art for several NLP tasks — GloVe, ELMo, and BERT — to highlight consistent trends in susceptibility, high confidence misclassifications, and high impact failures. We find that character or mixed ensemble models are the most effective defense mechanisms and that character perturbations are a more effective attack than word perturbations for deception classification.

adversarial evaluation↗