Designing for Interpretability and Adaptability by Using Weighted Averages.
Abstract not provided.
Agarwal, Sapan↗
SEARCH · Engineering Papers
Results for “Abstract Interpretation”
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.
A detailed study of interpretability of deep neural network based top taggers
Abstract Recent developments in the methods of explainable artificial intelligence (XAI) allow researchers to explore the inner workings of deep neural networks (DNNs), revealing crucial information about input–output relationships and realizing how data connects with machine learning models. In this paper we explore interpretability of DNN models designed to identify jets coming from top quark decay in high energy proton–proton collisions at the Large Hadron Collider. We review a subset of existing top tagger models and explore different quantitative methods to identify which features play the most important roles in identifying the top jets. We also investigate how and why feature importance varies across different XAI metrics, how correlations among features impact their explainability, and how latent space representations encode information as well as correlate with physically meaningful quantities. Our studies uncover some major pitfalls of existing XAI methods and illustrate how they can be overcome to obtain consistent and meaningful interpretation of these models. We additionally illustrate the activity of hidden layers as neural activation pattern diagrams and demonstrate how they can be used to understand how DNNs relay information across the layers and how this understanding can help to make such models significantly simpler by allowing effective model reoptimization and hyperparameter tuning. These studies not only facilitate a methodological approach to interpreting models but also unveil new insights about what these models learn. Incorporating these observations into augmented model design, we propose the particle flow interaction network model and demonstrate how interpretability-inspired model augmentation can improve top tagging performance.
97 MATHEMATICS AND COMPUTING↗
Voltage controlled dual-wavelength ZnO/Au/MgZnO UV photodetectors
Highlights: • A dual-wavelength UV photodetector has been constructed with a sandwich structure. • The ratio of the two response peaks can be changed by changing the bias voltage. • The principle was investigated by interpretation of the depletion widths. - Abstract: In this work, dual-wavelength UV photodetectors were successfully constructed by growing ZnO and MgZnO layers on SiO{sub 2} substrates with ZnO/Au/MgZnO structure. The photodetectors exhibited high performance and the influence of different thickness of MgZnO layers was investigated. Moreover, simply by adjusting the applied voltage on the devices, the ratio of the two response peaks (caused by MgZnO and ZnO layers) changed therewith. The detecting properties may originate from the two layers having different increasing rates where the existence of Schottky junctions plays an important role. The photodetectors are meaningful for fabricating dual-wavelength optoelectronic devices for practical applications.
75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
On stability theory
It is found that under mild assumptions, feedback system stability can be concluded if one can 'topologically separate' the infinite-dimensional function space containing the system's dynamical input-output relations into two regions, one region containing the dynamical input-output relation of the 'feedforward' element of the system and the other region containing the dynamical output-input relation of the 'feedback' element. Nonlinear system stability criteria of both the input-output type and the state-space (Liapunov) type are interpreted in this context. The abstract generality and conceptual simplicity afforded by the topological separation perspective clarifies some of the basic issues underlying stability theory and serves to suggest improvements in existing stability criteria. A generalization of Zames' (1966) conic-relation stability criterion is proved, laying the foundation for improved multivariable generalizations of the frequency-domain circle stability criterion for nonlinear systems.
Safonov, M. G.↗
TIGRE: Combinator graph reduction on the RTX 2000
An efficient evaluation technique is examined for lazy functional programs based on combinator graph reduction. Graph reduction is widely believed to be slow and inefficient, but an abstract machine called the Threaded Interpretive Graph Reduction Engine (TIGRE) achieves a substantial speedup over previous reduction techniques. The runtime system of TIGRE is a threaded system that permits self-modifying program execution with compiler-guaranteed safety. This paper describes an implementation of TIGRE in Forth for the Harris RTX 2000 stack processor.
Koopman, Philip, Jr.↗
On interpretation of fluctuations of conserved charges at high T
Abstract Fluctuations of conserved charges calculated on the lattice which can be measured experimentally, are well reproduced by a hadron resonanse gas model at temperatures below$$T_{ch} \sim 155$$ T ch ∼ 155 MeV and radically deviate from the hadron resonance gas predictions above the chiral restoration crossover. This behaviour is typically interpreted as an indication of deconfinement in the quark-gluon plasma regime. We present an argument that this interpretation may be too simple. The argument is based on the scaling of quantities with the number of colors: demonstration of deconfinement and QGP requires observable that is sensitive to$$\sim N_c^2$$ ∼ N c 2 gluons while the conserved charges are sensitive only to quarks and above$$T_{ch}$$ T ch scale as$$N_c^1$$ N c 1 . The latter scaling is consistent with the existence of an intermediate regime characterized by restored chiral symmetry and by approximate chiral spin symmetry which is a symmetry of confining interaction. In this regime the energy density, pressure and entropy density scale as$$N_c^1$$ N c 1 . In the large$$N_c$$ N c limit this regime might become a distinct phase separated from the hadron gas and from QGP by phase transitions. A natural observable that associates with deconfinement and is directly sensitive to deconfined$$N_c^2-1$$ N c 2 - 1 gluons is the Polyakov loop; in the$$N_c=3$$ N c = 3 world it remains very close to 0 at temperatures well above chiral crossover, reaches the value$$\sim 0.5$$ ∼ 0.5 around$$\sim 3T_{ch}$$ ∼ 3 T ch and the value close to 1 at temperatures$$\sim 1$$ ∼ 1 GeV.
Physics↗
On-the-fly active learning of interpretable Bayesian force fields for atomistic rare events
Abstract Machine learned force fields typically require manual construction of training sets consisting of thousands of first principles calculations, which can result in low training efficiency and unpredictable errors when applied to structures not represented in the training set of the model. This severely limits the practical application of these models in systems with dynamics governed by important rare events, such as chemical reactions and diffusion. We present an adaptive Bayesian inference method for automating the training of interpretable, low-dimensional, and multi-element interatomic force fields using structures drawn on the fly from molecular dynamics simulations. Within an active learning framework, the internal uncertainty of a Gaussian process regression model is used to decide whether to accept the model prediction or to perform a first principles calculation to augment the training set of the model. The method is applied to a range of single- and multi-element systems and shown to achieve a favorable balance of accuracy and computational efficiency, while requiring a minimal amount of ab initio training data. We provide a fully open-source implementation of our method, as well as a procedure to map trained models to computationally efficient tabulated force fields.
36 MATERIALS SCIENCE↗
Mapping Planetary Volcanic Deposits: Identifying Vents and Distinguishing between Effects of Eruption Conditions and Local Storage and Release on Flow Field Morphology
Terrestrial geologic mapping techniques are regularly used for "photogeologic" mapping of other planets, but these approaches are complicated by the diverse type, areal coverage, and spatial resolution of available data sets. When available, spatially-limited in-situ human and/or robotic surface observations can sometimes introduce a level of detail that is difficult to integrate with regional or global interpretations. To assess best practices for utilizing observations acquired from orbit and on the surface, our team conducted a comparative study of geologic mapping and interpretation techniques. We compared maps generated for the same area in the San Francisco Volcanic Field (SFVF) in northern Arizona using 1) data collected for reconnaissance before and during the 2010 Desert Research And Technology Studies campaign, and 2) during a traditional, terrestrial field geology study. The operations, related results, and direct mapping comparisons are discussed in companion LPSC abstracts. Here we present new geologic interpretations for a volcanic cone and related lava flows as derived from all approaches involved in this study. Mapping results indicate a need for caution when interpreting past eruption conditions on other planetary surfaces from orbital data alone.
Bleacher, J. E.↗
Systolic wall thickening and prone acquisition as tools to reduce false positives in the interpretation of the SPECT myocardial perfusion study
No abstract prepared.
62 RADIOLOGY AND NUCLEAR MEDICINE↗
Emergent Trends Complicate the Interpretation of the United States Drought Monitor (USDM)
Abstract Effective drought management must be informed by an understanding of whether and how current drought monitoring and assessment practices represent underlying nonstationary climate conditions, either naturally occurring or forced by climate change. Here we investigate the emerging climatology and associated trends in drought classes defined by the United States Drought Monitor (USDM), a weekly product that, since 2000, has been used to inform drought management in the United States. The USDM classifies drought intensity based in part on threshold percentiles in key hydroclimate quantities. Here we assess how those USDM‐defined drought threshold percentiles have changed over the last 23 years, examining precipitation, runoff, soil moisture (SM), terrestrial water storage (TWS), vapor pressure deficit (VPD), and near‐surface air temperature. We also assess underlying trends in the frequency of drought classifications across the U.S. Our analysis suggests that the frequency of drought class occurrence is exceeding the threshold percentiles defined by the USDM in a number of regions in the United States, particularly in the American West, where the last 23 years have emerged as a prolonged dry period. These trends are also reflected in percentile‐based thresholds in precipitation, runoff, SM, TWS, VPD, and temperature. Our results emphasize that while the USDM appears to be accurately reflecting observed nonstationarity in the physical climate, such trends raise critical questions about whether and how drought diagnosis, classification, and monitoring should address long‐term intervals of wet and dry periods or trends.
Geology↗
Soviet-French working group interpretation of the scientific information during the search for celestial sources of gamma pulses, abstract of reports, 24-30 March 1977
The progress made and techniques used by the Soviet-French group in the study of gamma and X ray pulses are described in abstracts of 16 reports. Experiments included calibration and operation of various recording instruments designed for measurements involving these pulses, specifically the location of sources of such pulses in outer space. Space vehicles are utilized in conjunction with ground equipment to accomplish these tests.
Estulin, I. V.↗
Interpretable Sequence Classification via Sparse Multiple Instance Learning.
Abstract not provided.
Wang, Fulton↗
Predicting and understanding corrosion in molten chloride salts
Abstract Molten chloride salts are stable at higher temperatures than many other salts, including nitrate salts, and are thus promising for heat transfer and/or thermal energy storage in concentrating solar power, nuclear power, and other thermal energy storage applications. However, corrosion in molten chloride salts remains a significant problem. While many studies have been devoted to evaluation of corrosion, we find that a comprehensive method for predicting corrosion in molten chloride salts is lacking. Here, we present an evaluation of corrosion in molten chloride salts using Ellingham diagrams and chloride-oxide stability diagrams, which enable prediction of alloy performance in molten chloride salts and allow corrosion results to be interpreted at a fundamental level. Graphical abstract
32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
Physically Interpretable Neural Networks for the Geosciences: Applications to Earth System Variability
Abstract Neural networks have become increasingly prevalent within the geosciences, although a common limitation of their usage has been a lack of methods to interpret what the networks learn and how they make decisions. As such, neural networks have often been used within the geosciences to most accurately identify a desired output given a set of inputs, with the interpretation of what the network learns used as a secondary metric to ensure the network is making the right decision for the right reason. Neural network interpretation techniques have become more advanced in recent years, however, and we therefore propose that the ultimate objective of using a neural network can also be the interpretation of what the network has learned rather than the output itself. We show that the interpretation of neural networks can enable the discovery of scientifically meaningful connections within geoscientific data. In particular, we use two methods for neural network interpretation called backward optimization and layerwise relevance propagation, both of which project the decision pathways of a network back onto the original input dimensions. To the best of our knowledge, LRP has not yet been applied to geoscientific research, and we believe it has great potential in this area. We show how these interpretation techniques can be used to reliably infer scientifically meaningful information from neural networks by applying them to common climate patterns. These results suggest that combining interpretable neural networks with novel scientific hypotheses will open the door to many new avenues in neural network‐related geoscience research.
58 GEOSCIENCES↗
Strong Photothermoelectric Response and Contact Reactivity in the Dirac Semimetal ZrTe5.
Abstract not provided.
Leonard, Francois Leonard↗
Electricity Subsector Transmission Resilience Maturity Model (TRMM) User Guide
The electric transmission sector is facing a range of threats to its functionality that are either new, more severe than experienced in earlier years, or more well understood. Such threats include more frequent and more severe extreme weather events, wildfires, droughts, and human-caused physical and cyberattacks. They also include geological, electromagnetic, and biological events. The novelty or increasing severity of these threats creates a significant need for transmission owners to implement programs to prevent, prepare for, respond to, and recover from such incidents. The national and economic security of the United States depends on the reliable functioning of the Nation’s critical infrastructure in the face of such threats, and the transmission networks are essential components of that infrastructure. The Electricity Subsector Transmission Resilience Maturity Model (TRMM) is a tool that a transmission organization can use to objectively evaluate and benchmark its currently established transmission resilience strategies, programs, policies, and investments, in order to target and prioritize enhancements where needed. The TRMM was developed to address the unique characteristics of the transmission system. The model can enable users to: • evaluate and benchmark their organization’s resilience capabilities, effectively and consistently • prioritize actions and investments to improve the resilience of their systems • share transmission-related knowledge, best practices, and relevant references within their organization and with business partners as a means to improve resilience capabilities • contribute to increasing the overall resilience of the Nation’s transmission systems. The TRMM provides descriptive rather than prescriptive industry-focused guidance. The model content is presented at a high level of abstraction so that it can be interpreted by transmission organizations of various types, structures, and sizes. The model is designed to an be easy-to-use, self-assessment tool.
24 POWER TRANSMISSION AND DISTRIBUTION↗
Analysis Description Languages for the LHC
An analysis description language is a domain specific language capable of describing the contents of an LHC analysis in a standard and unambiguous way, independent of any computing framework. It is designed for use by anyone with an interest in, and knowledge of, LHC physics, i.e., experimentalists, phenomenologists and other enthusiasts. Adopting analysis description languages would bring numerous benefits for the LHC experimental and phenomenological communities ranging from analysis preservation beyond the lifetimes of experiments or analysis software to facilitating the abstraction, design, visualization, validation, combination, reproduction, interpretation and overall communication of the analysis contents. Here, we introduce the analysis description language concept and summarize the current efforts ongoing to develop such languages and tools to use them in LHC analyses.
72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗