SEARCH · Engineering Papers
Results for “explainability”
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.
Transcriptional bursting explains the noise–versus–mean relationship in mRNA and protein levels
Recent analysis demonstrates that the HIV-1 Long Terminal Repeat (HIV LTR) promoter exhibits a range of possible transcriptional burst sizes and frequencies for any mean-expression level. However, these results have also been interpreted as demonstrating that cell-tocell expression variability (noise) and mean are uncorrelated, a significant deviation from previous results. Here, we re-examine the available mRNA and protein abundance data for the HIV LTR and find that noise in mRNA and protein expression scales inversely with the mean along analytically predicted transcriptional burst-size manifolds. We then experimentally perturb transcriptional activity to test a prediction of the multiple burst-size model: that increasing burst frequency will cause mRNA noise to decrease along given burst-size lines as mRNA levels increase. In conclusion, the data show that mRNA and protein noise decrease as mean expression increases, supporting the canonical inverse correlation between noise and mean.
Mechanisms of QT prolongation by buprenorphine cannot be explained by direct hERG channel block
Buprenorphine is a μ-opioid receptor (MOR) partial agonist used to manage pain and addiction. QT C prolongation that crosses the 10 msec threshold of regulatory concern was observed at a supratherapeutic dose in two thorough QT studies for the transdermal buprenorphine product BUTRANS®. Because QT C prolongation can be associated with Torsades de Pointes (TdP), a rare but potentially fatal ventricular arrhythmia, these results have led to further investigation of the electrophysiological effects of buprenorphine. Drug-induced QT C prolongation and TdP are most commonly caused by acute inhibition of hERG current (I hERG ) that contribute to the repolarizing phase of the ventricular action potentials (APs). Concomitant inhibition of inward late Na + (I NaL ) and/or L-type Ca 2+ (I CaL ) current can offer some protection against proarrhythmia. Therefore, we characterized the effects of buprenorphine and its major metabolite norbuprenorphine on cardiac hERG, Ca 2+ , and Na + ion channels, as well as cardiac APs. For comparison, methadone, a MOR agonist associated with QT C prolongation and high TdP risk, and naltrexone and naloxone, two opioid receptor antagonists, were also studied. Whole cell recordings were performed at 37°C on cells stably expressing hERG, Ca V 1.2, and Na V 1.5 proteins. Microelectrode array (MEA) recordings were made on human induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs). The results showed that buprenorphine, norbuprenorphine, naltrexone, and naloxone had no effect on I hERG , I CaL , I NaL , and peak Na + current (I NaP ) at clinically relevant concentrations. In contrast, methadone inhibited I hERG , I CaL , and I NaL . Experiments on iPSC-CMs showed a lack of effect for buprenorphine, norbuprenorphine, naltrexone, and naloxone, and delayed repolarization for methadone at clinically relevant concentrations. The mechanism of QT C prolongation is opioid moiety-specific. This remains undefined for buprenorphine, while for methadone it involves direct hERG channel block. There is no evidence that buprenorphine use is associated with TdP. Whether this lack of TdP risk can be generalized to other drugs with QT C prolongation not mediated by acute hERG channel block warrants further study.
Explainable AI in Cybersecurity Operations: Lessons Learned from xAI Tool Deployment.
Abstract not provided.
Evaluating and Explaining Natural Language Generation with GenX
Current methods for evaluation of natural language generation models focus on measuring text quality but fail to probe the model creativity, i.e., its ability to generate novel but coherent text sequences not seen in the training corpus. We present the GenX tool which is designed to enable interactive exploration and explanation of natural language generation outputs with a focus on the detection of memorization. We demonstrate the utility of the tool on two domain-conditioned generation use cases - phishing emails and ACL abstracts.
Hypothesis to Explain the Size Effect Observed in APO-BMI Compression Tests
In 2013 compression tests were performed on cylindrical specimens of carbon-microballoon-APOBMI syntactic foam machined to different lengths (0.25, 0.5, and 2.8 inches1) (Kingston, 2013). In 2014 similar tests were performed on glass-microballoon-APO-BMI of different lengths (~0.15”, ~0.32”, and ~0.57”). In all these tests it was observed that, when strains were calculated from the platen displacement (corrected for machine compliance), the apparent Young’s modulus of the material decreased with specimen size, as shown in Table 1. The reason for this size effect was speculated to be a layer of damage on or near the top and bottom machined surfaces of the specimens (Kingston, Schembri, & Siranosian, 2014). This report examines that hypothesis in further detail.
Explaining jurisdictional compliance with California’s top-down streamlined solar permitting law (AB 2188)
In the U.S., the process by which residential rooftop solar photovoltaic (PV) system installations are permitted and inspected by an “authority having jurisdiction” (AHJ) varies across AHJs. As the U.S. has roughly 23,000 AHJs with varying degrees of local autonomy, improving the cross-jurisdictional consistency of permitting and inspection is not a straightforward task. One promising approach is a “top-down” streamlined solar permitting (SSP) mandate by U.S. States to their AHJs. This study set out to understand the factors associated with the likelihood that an AHJ either did or did not comply with California’s top-down streamlined solar permitting law, AB2188, and then to consider how these factors might be relevant to the future prospects of top-down SSP outside of California. This report provides background material on AB2188 and tests five hypotheses about the factors that affect the likelihood that an AHJ will comply or not comply with AB2188, as developed from the policy diffusion literature and background knowledge on AB2188 and its implementation.
Explaining Neural Networks with Functional DataUsing PCA and Feature Importance.
Abstract not provided.
Explaining the Explanation: In Search of Understanding for Saliency Maps [Slides]
Pixel attribution methods are the most popular way to present explanations for image classifiers.
Test and Evaluation of Reinforcement Learning Via Robustes Testing and Explainable AI.
Abstract not provided.
Explainable AI in Cybersecurity Operations: Lessons Learned from xAI Tool Deployment.
Abstract not provided.
An Explainable Pipeline for Machine Learning with Functional Data.
Abstract not provided.
A dynamic statistical model for explaining the spread of COVID-19: A New Mexico Case Study.
Abstract not provided.
Vision Transformers Explained Series
Since their introduction in 2017 with Attention is All You Need¹, transformers have established themselves as the state of the art for natural language processing (NLP). In 2021, An Image is Worth 16x16 Words successfully adapted transformers for computer vision tasks. Since then, numerous transformer-based architectures have been proposed for computer vision. This article walks through the Vision Transformer (ViT) as laid out in An Image is Worth 16x16 Words.
FORESTR: Finding, Organizing, Representing, Explaining, Summarizing, and Thinning Random forests
Random forests have become popular models used for data driven predictions. As a result, random forests are currently used or being considered for high-consequence mission applications in national security, such as the prediction of yield from optical signals and malware detection. While random forests may provide accurate predictions, the complexity of the algorithm causes a lack of interpretability. Random forests are an ensemble of regression or decision trees. Individual regression and decision trees are interpretable, but ensembles are inherently difficult to interpret due to the compilation of many models. We aim to increase the interpretability of random forests by finding patterns in the ensemble of trees that can be used to “thin” (or remove) trees. As a starting point, in this report, we develop a new distance metric for quantifying the similarity between trees based on their topologies (i.e., shapes). We base the metric on a novel distance metric for graphs that is a proper mathematical distance, is invariant to transformations, has registration between graphs, and computes topological evolutions between graphs. We use the tree distance metric to compute tree statistics such as a “mean tree” and to identify clusters of trees. We apply the developed methodology to a toy dataset and a mission relevant product inspection dataset to demonstrate how the metric can provide insight into random forests. Furthermore, we discuss the limitations of the approach and ideas for future research into how the metric could be used as a thinning tool to develop less complex models.