Search NASA⌕ Search

SEARCH · Search NASA

Results for “Deep generative model”

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 919 records · Page 51

Sea quark transverse momentum distributions and dynamical chiral symmetry breaking

Recent theoretical studies have provided new insight into the intrinsic transverse momentum distributions of valence and sea quarks in the nucleon at a low scale. The valence quark transverse momentum distributions (q - qbar) are governed by the nucleon's inverse hadronic size R{sup -1} ~ 0.2 GeV and drop steeply at large p{sub T}. The sea quark distributions (qbar) are in large part generated by non-perturbative chiral-symmetry breaking interactions and extend up to the scale rho{sup -1} ~ 0.6 GeV. These findings have many implications for modeling the initial conditions of perturbative QCD evolution of TMD distributions (starting scale, shape of p{sub T}. distributions, coordinate-space correlation functions). The qualitative difference between valence and sea quark intrinsic p{sub T}. distributions could be observed experimentally, by comparing the transverse momentum distributions of selected hadrons in semi-inclusive deep-inelastic scattering, or those of dileptons produced in pp and pbar-p scattering.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Evaluation of SNS Beamline Shielding Configurations using MCNPX Accelerated by ADVANTG

Shielding analyses for the Spallation Neutron Source (SNS) at Oak Ridge National Laboratory pose significant computational challenges, including highly anisotropic high-energy sources, a combination of deep penetration shielding and an unshielded beamline, and a desire to obtain well-converged nearly global solutions for mapping of predicted radiation fields. The majority of these analyses have been performed using MCNPX with manually generated variance reduction parameters (source biasing and cell-based splitting and Russian roulette) that were largely based on the analyst's insight into the problem specifics. Development of the variance reduction parameters required extensive analyst time, and was often tailored to specific portions of the model phase space. We previously applied a developmental version of the ADVANTG code to an SNS beamline study to perform a hybrid deterministic/Monte Carlo analysis and showed that we could obtain nearly global Monte Carlo solutions with essentially uniform relative errors for mesh tallies that cover extensive portions of the model with typical voxel spacing of a few centimeters. The use of weight window maps and consistent biased sources produced using the FW-CADIS methodology in ADVANTG allowed us to obtain these solutions using substantially less computer time than the previous cell-based splitting approach. While those results were promising, the process of using the developmental version of ADVANTG was somewhat laborious, requiring user-developed Python scripts to drive much of the analysis sequence. In addition, limitations imposed by the size of weight-window files in MCNPX necessitated the use of relatively coarse spatial and energy discretization for the deterministic Denovo calculations that we used to generate the variance reduction parameters. We recently applied the production version of ADVANTG to this beamline analysis, which substantially streamlined the analysis process. We also tested importance function collapsing (in space and energy) capabilities in ADVANTG. These changes, along with the support for parallel Denovo calculations using the current version of ADVANTG, give us the capability to improve the fidelity of the deterministic portion of the hybrid analysis sequence, obtain improved weight-window maps, and reduce both the analyst and computational time required for the analysis process.

43 PARTICLE ACCELERATORS↗

Real-time sparse-sampled Ptychographic imaging through deep neural networks

Ptychography has rapidly grown in the fields of X-ray and electron imaging for its unprecedented ability to achieve nano or atomic scale resolution while simultaneously retrieving chemical or magnetic information from a sample. A ptychographic reconstruction is achieved by means of solving a complex inverse problem that imposes constraints both on the acquisition and on the analysis of the data, which typically precludes real-time imaging due to computational cost involved in solving this inverse problem. In this work we propose PtychoNN, a novel approach to solve the ptychography reconstruction problem based on deep convolutional neural networks. We demonstrate how the proposed method can be used to predict real-space structure and phase at each scan point solely from the corresponding far-field diffraction data. The presented results demonstrate how PtychoNN can effectively be used on experimental data, being able to generate high quality reconstructions of a sample up to hundreds of times faster than state-of-the-art ptychography reconstruction solutions once trained. By surpassing the typical constraints of iterative model-based methods, we can significantly relax the data acquisition sampling conditions and produce equally satisfactory reconstructions. Besides drastically accelerating acquisition and analysis, this capability can enable new imaging scenarios that were not possible before, in cases of dose sensitive, dynamic and extremely voluminous samples.

Cherukara, Mathew J↗

Automatic Detection and Classification of Radio Galaxy Images by Deep Learning

Abstract Surveys conducted by radio astronomy observatories, such as SKA, MeerKAT, Very Large Array, and ASKAP, have generated massive astronomical images containing radio galaxies (RGs). This generation of massive RG images has imposed strict requirements on the detection and classification of RGs and makes manual classification and detection increasingly difficult, even impossible. Rapid classification and detection of images of different types of RGs help astronomers make full use of the observed astronomical image data for further processing and analysis. The classification of FRI and FRII is relatively easy, and there are more studies and literature on them at present, but FR0 and FRI are similar, so it is difficult to distinguish them. It poses a greater challenge to image processing. At present, deep learning has made breakthrough progress in the field of image analysis and processing and has preliminary applications in astronomical data processing. Compared with classification algorithms that can only classify galaxies, object detection algorithms that can locate and classify RGs simultaneously are preferred. In target detection algorithms, YOLOv5 has outstanding advantages in the classification and positioning of small targets. Therefore, we propose a deep-learning method based on an improved YOLOv5 object detection model that makes full use of multisource data, combining FIRST radio with SDSS optical image data, and realizes the automatic detection of FR0, FRI, and FRII RGs. The innovation of our work is that on the basis of the original YOLOv5 object detection model, we introduce the SE Net attention mechanism, increase the number of preset anchors, adjust the network structure of the feature pyramid, and modify the network structure, thereby allowing our model to demonstrate galaxy classification and position detection effects. Our improved model produces satisfactory results, as evidenced by experiments. Overall, the mean average precision (mAP@0.5) of our improved model on the test set reaches 89.4%, which can determine the position (R.A. and decl.) and automatically detect and classify FR0s, FRIs, and FRIIs. Our work contributes to astronomy because it allows astronomers to locate FR0, FRI, and FRII galaxies in a relatively short time and can be further combined with other astronomically generated data to study the properties of these galaxies. The target detection model can also help astronomers find FR0s, FRIs, and FRIIs in future surveys and build a large-scale star RG catalog. Moreover, our work is also useful for the detection of other types of galaxies.

Astronomy & Astrophysics↗

Experimental Soil Warming Impacts Soil Moisture and Plant Water Stress and Thereby Ecosystem Carbon Dynamics (Blodgett, CA)

This dataset contains data on daily soil temperature, moisture and flux, and soil carbon stock and root biomass across a soil profile down to 100 cm depth at Blodgett Forest Research Station, CA, USA. These data were generated to determine if modeling of an experimental soil warming of 4C showed increased soil CO2 emissions and changes in bulk soil carbon stocks with depth consistent with field observations, as part of the study: Riley et al. (2025) Experimental Soil Warming Impacts Soil Moisture and Plant Water Stress and Thereby Ecosystem Carbon Dynamics in Journal of Advances in Modeling Earth Systems. This research was performed within the framework of the TES Belowground Biogeochemistry SFA project, in particular association with a 1 m-deep experimental soil heating experiment at the University of California Blodgett Forest Research Station, California (120 ° 39′40′′W; 38 ° 54′43′′N). Continuous data were collected at the plot level, and bulk soil carbon and root biomass were sampled once a year from each plot from 0-100 cm, in 10 cm intervals. Measurements relevant to the current study include soil temperature and soil volumetric water content measured continuously at multiple depths in the top meter; fine root biomass and SOC stocks measured from annual soil cores. Soil flux was continuously monitored using a LI-8100 Automated CO2 Flux System in conjunction with the LI-8150 Multiplexer (Licor, Nebraska, USA). Soil flux was determined using SoilFluxPro software, with flux values showing an R² fit of less than 0.9 being excluded from the analysis. Data were collected from each paired plot (1-3): one control (C) and one heated (H).

54 ENVIRONMENTAL SCIENCES↗

Deep Learning for Full Waveform Inversion of Elastic Active-Source Seismic Data to Estimate P-Wave Velocity Models

Seismic imaging methods are critical for Global Security and Energy & Homeland Security missions and activities that rely on subsurface characterization, but traditional methods remain computationally expensive and require significant labor hours and expertise to execute. Within the past few years, machine learning (ML), namely deep learning (DL), has been used to develop data-driven end-to-end full waveform inversion (FWI) methods to estimate 2D P-wave velocity (Vp) models in a fraction of the time as conventional FWI. These methods, however, are trained on simplistic acoustic wave seismic data and Vp models that are not realistic nor representative of real-world observations, leaving a large gap between the state-of-the-art and deployable, feasible, and practical DL FWI methods. Here, we generate a synthetic active-source, 3D, elastic wave seismic data set and a variety of Vp models with realistic geologic structure for training DL FWI methods. We evaluate six different methods that have performed well for acoustic DL FWI or medical imaging tasks using our more realistic dataset. We find that these six trained models do not match the performance of published acoustic end-to-end DL FWI methods, indicating more training data may be needed, physics may need to be incorporated to achieve good accuracy at the sacrifice of the end-to-end advantage, and/or novel methods need to be developed to enable end-to-end DL FWI methods to perform well for real-world seismic data.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

Computer-aided Abnormality Detection in Chest Radiographs in a Clinical Setting via Domain-adaptation

Deep learning (DL) models are being deployed at medical centers to aid radiologists for diagnosis of lung conditions from chest radiographs. Such models are often trained on a large volume of publicly available labeled radiographs. These pre-trained DL models’ ability to generalize in clinical settings is poor because of the changes in data distributions between publicly available and privately held radiographs. In chest radiographs, the heterogeneity in distributions arises from the diverse conditions in X-ray equipment and their configurations used for generating the images. In the machine learning community, the challenges posed by the heterogeneity in the data generation source is known as domain shift, which is a mode shift in the generative model. In this work, we introduce a domain-shift detection and removal method to overcome this problem. Our experimental results show the proposed method’s effectiveness in deploying a pre-trained DL model for abnormality detection in chest radiographs in a clinical setting.

Dubey, Abhishek↗

Thermal Equations of State of Magnesite: Implication for the Complex Mid‐Lower Mantle Seismic Scatterers

Abstract Magnesite (MgCO 3 ) entering the lower mantle together with the subducted oceanic crust is an important carbon carrier. The reaction between magnesite and mantle minerals has been documented, but its influence on the density and velocity profiles of lower mantle remains unexplored. To decipher the deep carbon transportation and its associated effect, here we determined the thermal equations of state of magnesite up to 120 GPa and 2600 K using X‐ray diffraction in laser‐heated diamond anvil cells. The obtained thermal elastic parameters of magnesite facilitated a comprehensive understanding on the influence of magnesite‐SiO 2 reaction, variation of carbon and SiO 2 content, and temperature on the origin of lower‐mantle scatterers at 1,000–1,800 km depth. Our modeling revealed that the depth of the lower‐mantleV S scatterers is mainly controlled by the Al 2 O 3 content in SiO 2 , while its magnitude depends on the SiO 2 content. Along normal geotherm, the magnesite‐SiO 2 reaction would occur before the post‐stishovite transition, consuming substantial SiO 2 in the subducted oceanic crust. Depending on the amount of residual SiO 2 , the post‐stishovite transition can produce a 2.5–5.2 (2)%V S reduction, compatible with the observed seismic scatterers in Izu‐Bonin and Mariana subduction zones. Along slab geotherm, this reaction occurs after the post‐stishovite transition, generating a greaterV S reduction of 4.4–6.4 (4)%. We thus propose that the reaction between sinking MgCO 3 and SiO 2 in the slab is one of the potential factors influencing the magnitude of the lower‐Vs scatterers at 1,000–1,900 km depth. Our results provide new insights into the deep‐mantle carbonate transportation influencing regional geophysics.

Geochemistry & Geophysics↗

Bi-fidelity modeling of uncertain and partially unknown systems using DeepONets

Recent advances in modeling large-scale, complex physical systems have shifted research focuses towards data-driven techniques. However, generating datasets by simulating complex systems can require significant computational resources. Similarly, acquiring experimental datasets can prove difficult. For these systems, often computationally inexpensive, but in general inaccurate models, known as the low-fidelity models, are available. Here in this paper, we propose a bi-fidelity modeling approach for complex physical systems, where we model the discrepancy between the true system's response and a low-fidelity response in the presence of a small training dataset from the true system's response using a deep operator network, a neural network architecture suitable for approximating nonlinear operators. We apply the approach to systems that have parametric uncertainty and are partially unknown. Three numerical examples are used to show the efficacy of the proposed approach to model uncertain and partially unknown physical systems.

MATHEMATICS AND COMPUTING↗

Adaptive Reinforcement Learning (ARL) Control of a Multi-port Resonant Converter in UAV Systems

This study presents an adaptive reinforcement learning (ARL) control framework for a multi-port resonant converter used in hybrid unmanned aerial vehicle (UAV) power systems. The converter integrates high-frequency half-bridge input ports connected to a rectified engine–generator set and a battery energy storage system, along with a semi-bridgeless active rectifier supplying the propulsion load. A deep RL agent is trained to dynamically regulate inter-port phase-shift commands in real time based on flight conditions and load power demand. The ARL controller autonomously identifies phase-shift combinations that maximize conversion efficiency while maintaining stable and coordinated power flow, even under rapidly varying operating scenarios. This data-driven approach eliminates the need for explicit system modeling or extensive manual tuning and enables coordinated control among multiple power ports without inter-port communication. Experimental results validate that the ARL based strategy achieves reliable power sharing and consistently high-efficiency operation across diverse UAV operating conditions.

Asa, Erdem [ORNL] (ORCID:0000000190884812)↗

Latent-space time evolution of non-intrusive reduced-order models using Gaussian process emulation

Non-intrusive reduced-order models (ROMs) have recently generated considerable interest for constructing computationally efficient counterparts of nonlinear dynamical systems emerging from various domain sciences. They provide a low-dimensional emulation framework for systems that may be intrinsically high-dimensional. This is accomplished by utilizing a construction algorithm that is purely data-driven. It is no surprise, therefore, that the algorithmic advances of machine learning have led to non-intrusive ROMs with greater accuracy and computational gains. However, in bypassing the utilization of an equation-based evolution, it is often seen that the interpretability of the ROM framework suffers. This becomes more problematic when black-box deep learning methods are used which are notorious for lacking robustness outside the physical regime of the observed data. In this article, we propose the use of a novel latent-space interpolation algorithm based on Gaussian process regression. Notably, this reduced-order evolution of the system is parameterized by control parameters to allow for interpolation in space. The use of this procedure also allows for a continuous interpretation of time which allows for temporal interpolation. The latter aspect provides information, with quantified uncertainty, about full-state evolution at a finer resolution than that utilized for training the ROMs. This research assesses the viability of this algorithm for an advection-dominated system given by the inviscid shallow water equations.

97 MATHEMATICS AND COMPUTING↗

Proceedings of RIKEN BNL Research Center Workshop: Brookhaven Summer Program on Nucleon Spin Physics

Understanding the structure of the nucleon is of fundamental importance in sub-atomic physics. Already the experimental studies on the electro-magnetic form factors in the 1950s showed that the nucleon has a nontrivial internal structure, and the deep inelastic scattering experiments in the 1970s revealed the partonic substructure of the nucleon. Modern research focuses in particular on the spin and the gluonic structure of the nucleon. Experiments using deep inelastic scattering or polarized p-p collisions are carried out in the US at the CEBAF and RHIC facilities, respectively, and there are other experimental facilities around the world. More than twenty years ago, the European Muon Collaboration published their first experimental results on the proton spin structure as revealed in polarized deep inelastic lepton-nucleon scattering, and concluded that quarks contribute very little to the proton's spin. With additional experimental and theoretical investigations and progress in the following years, it is now established that, contrary to naive quark model expectations, quarks and anti-quarks carry only about 30% of the total spin of the proton. Twenty years later, the discovery from the polarized hadron collider at RHIC was equally surprising. For the phase space probed by existing RHIC experiments, gluons do not seem to contribute any to the proton's spin. To find out what carries the remaining part of proton's spin is a key focus in current hadronic physics and also a major driving force for the new generation of spin experiments at RHIC and Jefferson Lab and at a future Electron Ion Collider. It is therefore very important and timely to organize a series of annual spin physics meetings to summarize the status of proton spin physics, to focus the effort, and to layout the future perspectives. This summer program on 'Nucleon Spin Physics' held at Brookhaven National Laboratory (BNL) on July 14-27, 2010 [http://www.bnl.gov/spnsp/] is the second one following the Berkeley Summer Program taken place in June of 2009. This program at BNL focused on theory and had many presentations on a wide range of theoretical aspects on nucleon spin, from perturbative-QCD calculations to models, and to the first principle lattice calculation. It also had a good number of summary talks from all major experimental collaborations on spin physics. The program facilitated many discussions between theorists as well as experimentalists. With five transparencies from each presentation at the Summer Program, this proceedings provides a valuable summary on the status and progress, as well as the future prospects of spin physics.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

NanoPSD: A software for automatic detection of Nano-Particle Shape Distribution in electron microscopy images

Accurate quantification of the size and morphology of nanoparticles from electron microscopy (EM) images is essential to understand growth mechanisms, surface reactivity, and functional behavior in nanoscale materials. Manual analysis remains slow, subjective, and difficult to reproduce in large datasets. We introduce NanoPSD (Nano-Particle Shape Distribution), an open-source and fully automated framework for quantitative particle detection and morphology analysis from EM images. NanoPSD integrates adaptive contrast enhancement, polarity-agnostic scale-bar detection, Optical Character Recognition (OCR)-based calibration, and classical segmentation via Otsu thresholding with morphological refinement. Particle contours are used to extract geometric descriptors, including equivalent circular diameter, aspect ratio, circularity, and solidity, enabling automated classification into spherical, rod-like, and aggregate morphologies. The framework supports both single-image and batch processing, generating publication-quality visualizations, LaTeX-ready tables, and structured comma-separated values (CSV) datasets. As a demonstration, we applied NanoPSD to plasma-synthesized nanoparticle samples diagnosed via transmission electron microscopy (TEM). The code produced statistically robust size and morphology distributions spanning a few to tens of nanometers with minimal user supervision. The pipeline demonstrates high reproducibility and scalability, processing large image collections with consistent calibration and output formatting. Its modular design enables seamless integration of future deep-learning-based segmentation models, providing a pathway toward intelligent, data-driven electron microscopy analysis.

36 MATERIALS SCIENCE↗

A DEEP NARROWBAND IMAGING SEARCH FOR C iv AND He ii EMISSION FROM Lyα BLOBS

We conduct a deep narrowband imaging survey of 13 Lyα blobs (LABs) located in the SSA22 proto-cluster at z ∼ 3.1 in the C iv and He ii emission lines in an effort to constrain the physical process powering the Lyα emission in LABs. Our observations probe down to unprecedented surface brightness (SB) limits of (2.1–3.4) × 10{sup −18} erg s{sup −1} cm{sup −2} arcsec{sup −2} per 1 arcsec{sup 2} aperture (5σ) for the He ii λ1640 and C iv λ1549 lines, respectively. We do not detect extended He ii and C iv emission in any of the LABs, placing strong upper limits on the He ii/Lyα and C iv/Lyα line ratios, of 0.11 and 0.16, for the brightest two LABs in the field. We conduct detailed photoionization modeling of the expected line ratios and find that, although our data constitute the deepest ever observations of these lines, they are still not deep enough to rule out a scenario where the Lyα emission is powered by the ionizing radiation from an obscured active galactic nucleus. Our models can accommodate He ii/Lyα and C iv/Lyα ratios as low as ≃0.05 and ≃0.07, respectively, implying that one needs to reach SB as low as (1–1.5) × 10{sup −18} erg s{sup −1} cm{sup −2} arcsec{sup −2} (at 5σ) in order to rule out a photoionization scenario. These depths will be achievable with the new generation of image-slicing integral field units such as the Multi Unit Spectroscopic Explorer (MUSE) on VLT and the Keck Cosmic Web Imager (KCWI). We also model the expected He ii/Lyα and C iv/Lyα in a different scenario, where Lyα emission is powered by shocks generated in a large-scale superwind, but find that our observational constraints can only be met for shock velocities v{sub s} ≳ 250 km s{sup −1}, which appear to be in conflict with recent observations of quiescent kinematics in LABs.

79 ASTRONOMY AND ASTROPHYSICS↗

Aspen Open Jets: unlocking LHC data for foundation models in particle physics

Foundation models are deep learning models pre-trained on large amounts of data which are capable of generalizing to multiple datasets and/or downstream tasks. This work demonstrates how data collected by the CMS experiment at the Large Hadron Collider can be useful in pre-training foundation models for HEP. Specifically, we introduce the AspenOpenJets (AOJs) dataset, consisting of approximately 178 M high p T jets derived from CMS 2016 Open Data. We show how pre-training the OmniJet-α foundation model on AOJs improves performance on generative tasks with significant domain shift: generating boosted top and QCD jets from the simulated JetClass dataset. In addition to demonstrating the power of pre-training of a jet-based foundation model on actual proton–proton collision data, we provide the ML-ready derived AOJs dataset for further public use.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Reimagine fire science for the anthropocene

Fire is an integral component of ecosystems globally and a tool that humans have harnessed for millennia. Altered fire regimes are a fundamental cause and consequence of global change, impacting people and the biophysical systems on which they depend. As part of the newly emerging Anthropocene, marked by human-caused climate change and radical changes to ecosystems, fire danger is increasing, and fires are having increasingly devastating impacts on human health, infrastructure, and ecosystem services. Increasing fire danger is a vexing problem that requires deep transdisciplinary, trans-sector, and inclusive partnerships to address. Here, we outline barriers and opportunities in the next generation of fire science and provide guidance for investment in future research. We synthesize insights needed to better address the long-standing challenges of innovation across disciplines to (i) promote coordinated research efforts; (ii) embrace different ways of knowing and knowledge generation; (iii) promote exploration of fundamental science; (iv) capitalize on the “firehose” of data for societal benefit; and (v) integrate human and natural systems into models across multiple scales. Fire science is thus at a critical transitional moment. We need to shift from observation and modeled representations of varying components of climate, people, vegetation, and fire to more integrative and predictive approaches that support pathways toward mitigating and adapting to our increasingly flammable world, including the utilization of fire for human safety and benefit. Only through overcoming institutional silos and accessing knowledge across diverse communities can we effectively undertake research that improves outcomes in our more fiery future.

54 ENVIRONMENTAL SCIENCES↗

Data Analysis of the 2020 Central Idaho Mainshock-Aftershock Sequence

In an effort to inform the Senior Seismic Hazard Analysis Committee for the Idaho National Laboratory, we provide an improved aftershock catalog related to the March 31, 2020, Mw6.5 Stanley, Idaho earthquake from picks related to a temporary network of two real-time and 15 non-telemetered seismometers within the epicentral area. From the permanent and temporary (XP) real-time network, the USGS cataloged 1,946 aftershocks between April 1, 2020 and October 31, 2020. To improve aftershock location and magnitudes, we manually picked arrival times of P and S waves from off-line stations in the XP temporary network, generated a new crustal velocity model, and independently relocated each event using the HypoDD double-difference earthquake algorithm. We created our new velocity model from existing broadband and active source seismic campaign data that were acquired near the epicentral region prior to the 2020 earthquake. We compare arrival time differences, epicentral locations and depths between aftershocks recorded with the two catalogs. We find the addition of local stations provides tighter aftershock clustering that suggests an improved aftershock locations. To detect lower magnitude events, we employed deep learning. Our method solves common problems associated with detecting many events that have a low signal-to-noise ratio. From the machine learning database, we detected more than 74,000 aftershocks. Based on the number of identified earthquakes and Gutenberg-Richter relationships derived from the USGS catalog, we estimate that we have reduced the completion magnitude for the Stanley earthquake sequence to below M1 using this machine learning approach. We located each aftershock with our new velocity model. Our new velocity model and picks suggests aftershocks occurred mostly at shallower depths than assessed in the USGS catalog. These aftershocks align along two linear trends that suggest the activation of two unnamed primary faults.

58 GEOSCIENCES↗

Robust Assessment of Real-World Adversarial Examples

We explore rigorous, systematic, and controlled experimental evaluation of adversarial examples in the real world and propose a testing regimen for evaluation of real world adversarial objects. We show that for small scene/ environmental perturbations, large adversarial performance differences exist. Current state of adversarial reporting exists largely as a frequency count over a dynamic collections of scenes. Our work underscores the need for either a more complete report or a score that incorporates scene changes and baseline performance for models and environments tested by adversarial developers. We put forth a score that attempts to address the above issues in a straight-forward exemplar application for multiple generated adversary examples. We contribute the following: 1. a testbed for adversarial assessment, 2. a score for adversarial examples, and 3. a collection of additional evaluations on testbed data.

deep learning, adversarial machine learning, machi↗