arXiv subjects

Donald Loveland

Publications and source records attributed to Donald Loveland.

28 records · Page 2Linked to original sources

arXiv2022

Zeroth-Order SciML: Non-intrusive Integration of Scientific Software with Deep Learning

Using deep learning (DL) to accelerate and/or improve scientific workflows can yield discoveries that are otherwise impossible. Unfortunately, DL models have yielded limited success in complex scientific domains due to large data requirements. In this work, we propose to overcome this issue by integrating the abundance of scientific knowledge sources (SKS) with the DL training process. Existing knowledge integration approaches are limited to using differentiable knowledge source to be compatible with first-order DL training paradigm. In contrast, our proposed approach treats knowledge source as a black-box in turn allowing to integrate virtually any knowledge source. To enable an end-to-end training of SKS-coupled-DL, we propose to use zeroth-order optimization (ZOO) based gradient-free training schemes, which is non-intrusive, i.e., does not require making any changes to the SKS. We evaluate the performance of our ZOO training scheme on two real-world material science applications. We show that proposed scheme is able to effectively integrate scientific knowledge with DL training and is able to outperform purely data-driven model for data-limited scientific applications. We also discuss some limitations of the proposed method and mention potentially worthwhile future directions.

cs.LG↗

arXiv2022

The Lick AGN Monitoring Project 2016: Dynamical Modeling of Velocity-Resolved H\b{eta} Lags in Luminous Seyfert Galaxies

We have modeled the velocity-resolved reverberation response of the H\b{eta} broad emission line in nine Seyfert 1 galaxies from the Lick Active Galactic Nucleus (AGN) Monitioring Project 2016 sample, drawing inferences on the geometry and structure of the low-ionization broad-line region (BLR) and the mass of the central supermassive black hole. Overall, we find that the H\b{eta} BLR is generally a thick disk viewed at low to moderate inclination angles. We combine our sample with prior studies and investigate line-profile shape dependence, such as log10(FWHM/σ), on BLR structure and kinematics and search for any BLR luminosity-dependent trends. We find marginal evidence for an anticorrelation between the profile shape of the broad H\b{eta} emission line and the Eddington ratio, when using the root-mean-square spectrum. However, we do not find any luminosity-dependent trends, and conclude that AGNs have diverse BLR structure and kinematics, consistent with the hypothesis of transient AGN/BLR conditions rather than systematic trends.

astro-ph.GA↗

arXiv2022

FairEdit: Preserving Fairness in Graph Neural Networks through Greedy Graph Editing

Graph Neural Networks (GNNs) have proven to excel in predictive modeling tasks where the underlying data is a graph. However, as GNNs are extensively used in human-centered applications, the issue of fairness has arisen. While edge deletion is a common method used to promote fairness in GNNs, it fails to consider when data is inherently missing fair connections. In this work we consider the unexplored method of edge addition, accompanied by deletion, to promote fairness. We propose two model-agnostic algorithms to perform edge editing: a brute force approach and a continuous approximation approach, FairEdit. FairEdit performs efficient edge editing by leveraging gradient information of a fairness loss to find edges that improve fairness. We find that FairEdit outperforms standard training for many data sets and GNN methods, while performing comparably to many state-of-the-art methods, demonstrating FairEdit's ability to improve fairness across many domains and models.

cs.LG↗

arXiv2021

The Lick AGN Monitoring Project 2016: Velocity-Resolved Hβ Lags in Luminous Seyfert Galaxies

We carried out spectroscopic monitoring of 21 low-redshift Seyfert 1 galaxies using the Kast double spectrograph on the 3-m Shane telescope at Lick Observatory from April 2016 to May 2017. Targeting active galactic nuclei (AGN) with luminosities of λLλ (5100 Å) = 10^44 erg/s and predicted Hβ lags of 20-30 days or black hole masses of 10^7-10^8.5 Msun, our campaign probes luminosity-dependent trends in broad-line region (BLR) structure and dynamics as well as to improve calibrations for single-epoch estimates of quasar black hole masses. Here we present the first results from the campaign, including Hβ emission-line light curves, integrated Hβ lag times (8-30 days) measured against V-band continuum light curves, velocity-resolved reverberation lags, line widths of the broad Hβ components, and virial black hole mass estimates (10^7.1-10^8.1 Msun). Our results add significantly to the number of existing velocity-resolved lag measurements and reveal a diversity of BLR gas kinematics at moderately high AGN luminosities. AGN continuum luminosity appears not to be correlated with the type of kinematics that its BLR gas may exhibit. Follow-up direct modeling of this dataset will elucidate the detailed kinematics and provide robust dynamical black hole masses for several objects in this sample.

astro-ph.GA↗

arXiv2021

Reliable Graph Neural Network Explanations Through Adversarial Training

Graph neural network (GNN) explanations have largely been facilitated through post-hoc introspection. While this has been deemed successful, many post-hoc explanation methods have been shown to fail in capturing a model's learned representation. Due to this problem, it is worthwhile to consider how one might train a model so that it is more amenable to post-hoc analysis. Given the success of adversarial training in the computer vision domain to train models with more reliable representations, we propose a similar training paradigm for GNNs and analyze the respective impact on a model's explanations. In instances without ground truth labels, we also determine how well an explanation method is utilizing a model's learned representation through a new metric and demonstrate adversarial training can help better extract domain-relevant insights in chemistry.

cs.LG↗

arXiv2020

Explainable Deep Learning for Uncovering Actionable Scientific Insights for Materials Discovery and Design

The scientific community has been increasingly interested in harnessing the power of deep learning to solve various domain challenges. However, despite the effectiveness in building predictive models, fundamental challenges exist in extracting actionable knowledge from deep neural networks due to their opaque nature. In this work, we propose techniques for exploring the behavior of deep learning models by injecting domain-specific actionable attributes as tunable "knobs" in the analysis pipeline. By incorporating the domain knowledge in a generative modeling framework, we are not only able to better understand the behavior of these black-box models, but also provide scientists with actionable insights that can potentially lead to fundamental discoveries.

cs.LG↗

arXiv2020

Actionable Attribution Maps for Scientific Machine Learning

The scientific community has been increasingly interested in harnessing the power of deep learning to solve various domain challenges. However, despite the effectiveness in building predictive models, fundamental challenges exist in extracting actionable knowledge from the deep neural network due to their opaque nature. In this work, we propose techniques for exploring the behavior of deep learning models by injecting domain-specific actionable concepts as tunable ``knobs'' in the analysis pipeline. By incorporating the domain knowledge with generative modeling, we are not only able to better understand the behavior of these black-box models, but also provide scientists with actionable insights that can potentially lead to fundamental discoveries.

cs.CV↗

arXiv2020

Predicting Compressive Strength of Consolidated Molecular Solids Using Computer Vision and Deep Learning

We explore the application of computer vision and machine learning (ML) techniques to predict material properties (e.g. compressive strength) based on SEM images. We show that it's possible to train ML models to predict materials performance based on SEM images alone, demonstrating this capability on the real-world problem of predicting uniaxially compressed peak stress of consolidated molecular solids samples. Our image-based ML approach reduces mean absolute percent error (MAPE) by an average of 24% over baselines representative of the current state-of-the-practice (i.e., domain-expert's analysis and correlation). We compared two complementary approaches to this problem: (1) a traditional ML approach, random forest (RF), using state-of-the-art computer vision features and (2) an end-to-end deep learning (DL) approach, where features are learned automatically from raw images. We demonstrate the complementarity of these approaches, showing that RF performs best in the "small data" regime in which many real-world scientific applications reside (up to 24% lower RMSE than DL), whereas DL outpaces RF in the "big data" regime, where abundant training samples are available (up to 24% lower RMSE than RF). Finally, we demonstrate that models trained using machine learning techniques are capable of discovering and utilizing informative crystal attributes previously underutilized by domain experts.

physics.comp-ph↗

arXiv2019

Generative Counterfactual Introspection for Explainable Deep Learning

In this work, we propose an introspection technique for deep neural networks that relies on a generative model to instigate salient editing of the input image for model interpretation. Such modification provides the fundamental interventional operation that allows us to obtain answers to counterfactual inquiries, i.e., what meaningful change can be made to the input image in order to alter the prediction. We demonstrate how to reveal interesting properties of the given classifiers by utilizing the proposed introspection approach on both the MNIST and the CelebA dataset.

cs.LG↗

arXiv2018

Studying the [OIII]$λ$5007A emission-line width in a sample of $\sim$80 local active galaxies: A surrogate for $σ_{\star}$?

For a sample of $\sim$80 local ($0.02 \leq z \leq 0.1$) Seyfert-1 galaxies with high-quality long-slit Keck spectra and spatially-resolved stellar-velocity dispersion ($σ_{\star}$) measurements, we study the profile of the [OIII]$λ$5007A emission line to test the validity of using its width as a surrogate for $σ_{\star}$. Such an approach has often been used in the literature, since it is difficult to measure $σ_{\star}$ for type-1 active galactic nuclei (AGNs) due to the AGN continuum outshining the stellar-absorption lines. Fitting the [OIII] line with a single Gaussian or Gauss-Hermite polynomials overestimates $σ_{\star}$ by 50-100%. When line asymmetries from non-gravitational gas motion are excluded in a double Gaussian fit, the average ratio between the core [OIII] width ($σ_{\rm {[OIII],D}}$) and $σ_{\star}$ is $\sim$1, but with individual data points off by up to a factor of two. The resulting black-hole-mass-$σ_{\rm {[OIII],D}}$ relation scatters around that of quiescent galaxies and reverberation-mapped AGNs. However, a direct comparison between $σ_{\star}$ and $σ_{\rm {[OIII],D}}$ shows no close correlation, only that both quantities have the same range, average and standard deviation, probably because they feel the same gravitational potential. The large scatter is likely due to the fact that line profiles are a luminosity-weighted average, dependent on the light distribution and underlying kinematic field. Within the range probed by our sample (80-260 km s$^{-1}$), our results strongly caution against the use of [OIII] width as a surrogate for $σ_{\star}$ on an individual basis. Even though our sample consists of radio-quiet AGNs, FIRST radio-detected objects have, on average, a $\sim$10% larger [OIII] core width.

astro-ph.GA↗