arXiv subjects

Michael Regan

Publications and source records attributed to Michael Regan.

At least 19 recordsLinked to original sources

arXiv2024

The James Webb Space Telescope Absolute Flux Calibration. II. Mid-Infrared Instrument Imaging and Coronagraphy

The absolute flux calibration of the Mid-Infrared Instrument Imaging and Coronagraphy is based on observations of multiple stars taken during the first 2.5 years of JWST operations. The observations were designed to ensure that the flux calibration is valid for a range of flux densities, different subarrays, and different types of stars. The flux calibration was measured by combining observed aperture photometry corrected to infinite aperture with predictions based on previous observations and models of stellar atmospheres. A subset of these observations were combined with model point-spread-functions to measure the corrections to infinite aperture. Variations in the calibration factor with time, flux density, background level, type of star, subarray, integration time, rate, and well depth were investigated, and the only significant variations were with time and subarray. Observations of the same star taken approximately every month revealed a modest time-dependent response loss seen mainly at the longest wavelengths. This loss is well characterized by a decaying exponential with a time constant of ~200 days. After correcting for the response loss, the band-dependent scatter around the corrected average (aka repeatability) was found to range from 0.1 to 1.2%. Signals in observations taken with different subarrays can be lower by up to 3.4% compared to FULL frame. After correcting for the time and subarray dependencies, the scatter in the calibration factors measured for individual stars ranges from 1 to 4% depending on the band. The formal uncertainties on the flux calibration averaged for all observations are 0.3 to 1.0%, with longer-wavelength bands generally having larger uncertainties.

astro-ph.IM↗

arXiv2024

MASSIVE Multilingual Abstract Meaning Representation: A Dataset and Baselines for Hallucination Detection

Abstract Meaning Representation (AMR) is a semantic formalism that captures the core meaning of an utterance. There has been substantial work developing AMR corpora in English and more recently across languages, though the limited size of existing datasets and the cost of collecting more annotations are prohibitive. With both engineering and scientific questions in mind, we introduce MASSIVE-AMR, a dataset with more than 84,000 text-to-graph annotations, currently the largest and most diverse of its kind: AMR graphs for 1,685 information-seeking utterances mapped to 50+ typologically diverse languages. We describe how we built our resource and its unique features before reporting on experiments using large language models for multilingual AMR and SPARQL parsing as well as applying AMRs for hallucination detection in the context of knowledge base question answering, with results shedding light on persistent issues using LLMs for structured parsing.

cs.CL↗

arXiv2024

Linear Cross-document Event Coreference Resolution with X-AMR

Event Coreference Resolution (ECR) as a pairwise mention classification task is expensive both for automated systems and manual annotations. The task's quadratic difficulty is exacerbated when using Large Language Models (LLMs), making prompt engineering for ECR prohibitively costly. In this work, we propose a graphical representation of events, X-AMR, anchored around individual mentions using a \textbf{cross}-document version of \textbf{A}bstract \textbf{M}eaning \textbf{R}epresentation. We then linearize the ECR with a novel multi-hop coreference algorithm over the event graphs. The event graphs simplify ECR, making it a) LLM cost-effective, b) compositional and interpretable, and c) easily annotated. For a fair assessment, we first enrich an existing ECR benchmark dataset with these event graphs using an annotator-friendly tool we introduce. Then, we employ GPT-4, the newest LLM by OpenAI, for these annotations. Finally, using the ECR algorithm, we assess GPT-4 against humans and analyze its limitations. Through this research, we aim to advance the state-of-the-art for efficient ECR and shed light on the potential shortcomings of current LLMs at this task. Code and annotations: \url{https://github.com/ahmeshaf/gpt_coref}

cs.CL↗

arXiv2023

Why Autonomous Vehicles Are Not Ready Yet: A Multi-Disciplinary Review of Problems, Attempted Solutions, and Future Directions

Personal autonomous vehicles are cars, trucks and bikes capable of sensing their surrounding environment, planning their route, and driving with little or no involvement of human drivers. Despite the impressive technological achievements made by the industry in recent times and the hopeful announcements made by leading entrepreneurs, to date no personal vehicle is approved for road circulation in a 'fully' or 'semi' autonomous mode (autonomy levels 4 and 5) and it is still unclear when such vehicles will eventually be mature enough to receive this kind of approval. The present review adopts an integrative and multidisciplinary approach to investigate the major challenges faced by the automative sector, with the aim to identify the problems that still trouble and delay the commercialization of autonomous vehicles. The review examines the limitations and risks associated with current technologies and the most promising solutions devised by the researchers. This negative assessment methodology is not motivated by pessimism, but by the aspiration to raise critical awareness about the technology's state-of-the-art, the industry's quality standards, and the society's demands and expectations. While the survey primarily focuses on the applications of artificial intelligence for perception and navigation, it also aims to offer an enlarged picture that links the purely technological aspects with the relevant human-centric aspects, including, cultural attitudes, conceptual assumptions, and normative (ethico-legal) frameworks. Examining the broader context serves to highlight problems that have a cross-disciplinary scope and identify solutions that may benefit from a holistic consideration.

cs.CV↗

arXiv2023

How Good is the Model in Model-in-the-loop Event Coreference Resolution Annotation?

Annotating cross-document event coreference links is a time-consuming and cognitively demanding task that can compromise annotation quality and efficiency. To address this, we propose a model-in-the-loop annotation approach for event coreference resolution, where a machine learning model suggests likely corefering event pairs only. We evaluate the effectiveness of this approach by first simulating the annotation process and then, using a novel annotator-centric Recall-Annotation effort trade-off metric, we compare the results of various underlying models and datasets. We finally present a method for obtaining 97\% recall while substantially reducing the workload required by a fully manual annotation process. Code and data can be found at https://github.com/ahmeshaf/model_in_coref

cs.CL↗

arXiv2023

Causal schema induction for knowledge discovery

Making sense of familiar yet new situations typically involves making generalizations about causal schemas, stories that help humans reason about event sequences. Reasoning about events includes identifying cause and effect relations shared across event instances, a process we refer to as causal schema induction. Statistical schema induction systems may leverage structural knowledge encoded in discourse or the causal graphs associated with event meaning, however resources to study such causal structure are few in number and limited in size. In this work, we investigate how to apply schema induction models to the task of knowledge discovery for enhanced search of English-language news texts. To tackle the problem of data scarcity, we present Torquestra, a manually curated dataset of text-graph-schema units integrating temporal, event, and causal structures. We benchmark our dataset on three knowledge discovery tasks, building and evaluating models for each. Results show that systems that harness causal structure are effective at identifying texts sharing similar causal meaning components rather than relying on lexical cues alone. We make our dataset and models available for research purposes.

cs.CL↗

arXiv2023

The Brighter-Fatter Effect in the JWST MIRI Si:As IBC detectors I. Observations, impact on science, and modelling

The Mid-Infrared Instrument (MIRI) on board the James Webb Space Telescope (JWST) uses three Si:As impurity band conduction (IBC) detector arrays. The output voltage level of each MIRI detector pixel is digitally recorded by sampling-up-the-ramp. For uniform or low-contrast illumination, the pixel ramps become non-linear in a predictable way, but in areas of high contrast, the non-linearity curve becomes much more complex. The origin of the effect is poorly understood and currently not calibrated. We provide observational evidence of the Brighter-Fatter Effect (BFE) in MIRI conventional and high-contrast coronographic imaging, low-resolution spectroscopy, and medium-resolution spectroscopy data and investigate the physical mechanism that gives rise to the effect on the MIRI detector pixel raw voltage integration ramps. We use public data from the JWST MIRI commissioning and Cycle 1 phase. We also develop a numerical electrostatic model of the MIRI detectors using a modified version of the public Poisson_CCD code. We find that the physical mechanism behind the BFE manifesting in MIRI data is fundamentally different to that of CCDs and photodiode arrays such as the Hawaii-XRG (HXRG) near-infrared detectors used by the NIRISS, NIRCam, and NIRSpec instruments on board JWST. Observationally, the BFE makes the JWST MIRI data yield 10-25 % larger point sources and spectral line profiles as a function of the relative level of debiasing of neighboring detector pixels. This broadening impacts the MIRI absolute flux calibration, time-series observations of faint companions, and PSF modeling and subtraction. We also find that the intra-pixel 2D profile of the shrinking Si:As IBC detector depletion region directly impacts the accuracy of the pixel ramp non-linearity calibration model.

astro-ph.IM↗

arXiv2022

Spectroscopic Time-series Performance of JWST/NIRSpec from Commissioning Observations

We report on JWST commissioning observations of the transiting exoplanet HAT-P-14 b, obtained using the Bright Object Time Series (BOTS) mode of the NIRSpec instrument with the G395H/F290LP grating/filter combination ($3-5\mu$m). While the data were used primarily to verify that the NIRSpec BOTS mode is working as expected, and to enable it for general scientific use, they yield a precise transmission spectrum which we find is featureless down to the precision level of the instrument, consistent with expectations given HAT-P-14~b's small scale-height and hence expected atmospheric features. The exquisite quality and stability of the \emph{JWST/NIRSpec} transit spectrum -- almost devoid of any systematic effects -- allowed us to obtain median uncertainties of 50-60 ppm in this wavelength range at a resolution of $R=100$ in a single exposure, which is in excellent agreement with pre-flight expectations and close to the (or at the) photon-noise limit for a $J = 9.094$, F-type star like HAT-P-14. These observations showcase the ability of NIRSpec/BOTS to perform cutting-edge transiting exoplanet atmospheric science, setting the stage for observations and discoveries to be made in Cycle 1 and beyond.

astro-ph.EP↗

arXiv2019

A dataset for resolving referring expressions in spoken dialogue via contextual query rewrites (CQR)

We present Contextual Query Rewrite (CQR) a dataset for multi-domain task-oriented spoken dialogue systems that is an extension of the Stanford dialog corpus (Eric et al., 2017a). While previous approaches have addressed the issue of diverse schemas by learning candidate transformations (Naik et al., 2018), we instead model the reference resolution task as a user query reformulation task, where the dialog state is serialized into a natural language query that can be executed by the downstream spoken language understanding system. In this paper, we describe our methodology for creating the query reformulation extension to the dialog corpus, and present an initial set of experiments to establish a baseline for the CQR task. We have released the corpus to the public [1] to support further research in this area.

cs.CL↗

arXiv2015

The Spitzer Survey of Stellar Structure in Galaxies (S4G): Stellar Masses, Sizes and Radial Profiles for 2352 Nearby Galaxies

The Spitzer Survey of Stellar Structure in Galaxies (S4G) is a volume, magnitude, and size-limited survey of 2352 nearby galaxies with deep imaging at 3.6 and 4.5um. In this paper we describe our surface photometry pipeline and showcase the associated data products that we have released to the community. We also identify the physical mechanisms leading to different levels of central stellar mass concentration for galaxies with the same total stellar mass. Finally, we derive the local stellar mass-size relation at 3.6um for galaxies of different morphologies. Our radial profiles reach stellar mass surface densities below 1 Msun pc-2. Given the negligible impact of dust and the almost constant mass-to-light ratio at these wavelengths, these profiles constitute an accurate inventory of the radial distribution of stellar mass in nearby galaxies. From these profiles we have also derived global properties such as asymptotic magnitudes (and the corresponding stellar masses), isophotal sizes and shapes, and concentration indices. These and other data products from our various pipelines (science-ready mosaics, object masks, 2D image decompositions, and stellar mass maps), can be publicly accessed at IRSA (http://irsa.ipac.caltech.edu/data/SPITZER/S4G/).

astro-ph.GA↗

arXiv2014

Reconstructing the stellar mass distributions of galaxies using S4G IRAC 3.6 and 4.5 micron images: II. The conversion from light to mass

We present a new approach for estimating the 3.6 micron stellar mass-to-light ratio in terms of the [3.6]-[4.5] colors of old stellar populations. Our approach avoids several of the largest sources of uncertainty in existing techniques. By focusing on mid-IR wavelengths, we gain a virtually dust extinction-free tracer of the old stars, avoiding the need to adopt a dust model to correctly interpret optical or optical/NIR colors normally leveraged to assign M/L. By calibrating a new relation between NIR and mid-IR colors of GLIMPSE giant stars we also avoid discrepancies in model predictions for the [3.6]-[4.5] colors of old stellar populations due to uncertainties in molecular line opacities. We find that the [3.6]-[4.5] color, which is driven primarily by metallicity, provides a tight constraint on M/L_3.6, which varies intrinsically less than at optical wavelengths. The uncertainty on M/L_3.6 of ~0.07 dex due to unconstrained age variations marks a significant improvement on existing techniques for estimating the stellar M/L with shorter wavelength data. A single M/L_3.6=0.6 (assuming a Chabrier IMF), independent of [3.6]-[4.5] color, is also feasible as it can be applied simultaneously to old, metal-rich and young, metal-poor populations, and still with comparable (or better) accuracy (~0.1 dex) as alternatives. We expect our M/L_3.6 to be optimal for mapping the stellar mass distributions in S4G galaxies, for which we have developed an Independent Component Analysis technique to first isolate the old stellar light at 3.6 micron from non-stellar emission (e.g. hot dust and the 3.3 PAH feature). Our estimate can also be used to determine the fractional contribution of non-stellar emission to global (rest-frame) 3.6 micron fluxes, e.g. in WISE imaging, and establishes a reliable basis for exploring variations in the stellar IMF.

astro-ph.GA↗

arXiv2014

Understanding Persistence: A 3D Trap Map of an H2RG Imaging Sensor

Several theories exist to explain persistence, most of which revolve around the distribution of traps. We aim to simulate persistence and illustrate this complex issue with a 3D trap map. For this experiment, we vary the detector voltage bias to simulate the change in the depletion region that occurs when the detector is exposed to light. This allows us to measure the distribution of traps in the depletion region. This paper will explore the results from this experiment and discuss the implications.

astro-ph.IM↗

arXiv2013

X-ray nuclear activity in S4G barred galaxies: No link between bar strength and co-occurrent supermassive black hole fueling

Stellar bars can lead to gas inflow toward the center of a galaxy and stimulate nuclear star formation. However, there is no compelling evidence on whether they also feed a central supermassive black hole: by measuring the fractions of barred active and inactive galaxies, previous studies have yielded conflicting results. In this paper, we aim to understand the lack of observational evidence for bar-driven active galactic nucleus (AGN) activity by studying a sample of 41 nearby (d < 35 Mpc) barred galaxies from the Spitzer Survey for Stellar Structure in Galaxies. We use Chandra observations to measure nuclear 2--10 keV X-ray luminosities and estimate Eddington ratios, together with Spitzer 3.6um imaging to quantify the strength of the stellar bar in two independent ways: (1) from its structure, as traced by its ellipticity and boxiness, and (2) from its gravitational torque Q_b, taken as the maximum ratio of the tangential force to the mean background radial force. In this way, rather than discretizing the presence of both stellar bars and nuclear activity, we are able to account for the continuum of bar strengths and degrees of AGN activity. We find nuclear X-ray sources in 31 out of 41 galaxies with median X-ray luminosity and Eddington ratio of L_X=4.3x10^{38} erg/s and L_bol/L_Edd=6.9x10^{-6} respectively, consistent with low-luminosity AGN activity. Including upper limits for those galaxies without nuclear detections, we find no significant correlation between any of the bar strength indicators and the degree of nuclear activity, irrespective of galaxy luminosity, stellar mass, Hubble type, or bulge size. Strong bars do not favor brighter or more efficient nuclear activity, implying that at least for the low-luminosity regime, supermassive black hole fueling is not closely connected to large scale features.

astro-ph.CO↗

arXiv2012

The S4G perspective on circumstellar dust extinction of AGB stars in M100

We examine the effect of circumstellar dust extinction on the near-IR contribution of asymptotic giant branch (AGB) stars in intermediate-age clusters throughout the disk of M100. For our sample of 17 AGB-dominated clusters we extract optical-to-mid-IR SEDs and find that NIR brightness is coupled to the mid-IR dust emission in such a way that a significant reduction of AGB light, of up to 1 mag in K-band, follows from extinction by the dust shell formed during this stage. Since the dust optical depth varies with AGB chemistry (C-rich or O-rich), our results suggest that the contribution of AGB stars to the flux from their host clusters will be closely linked to the metallicity and the progenitor mass of the AGB star, to which dust chemistry and mass-loss rate are sensitive. Our sample of clusters--each the analogue of a ~1 Gyr old post-starburst galaxy--has implications within the context of mass and age estimation via SED modelling at high z: we find that the average ~0.5 mag extinction estimated here may be sufficient to reduce the AGB contribution in (rest-frame) K-band from ~70%, as predicted in the latest generation of synthesis models, to ~35%. Our technique for selecting AGB-dominated clusters in nearby galaxies promises to be effective for discriminating the uncertainties associated with AGB stars in intermediate-age populations that plague age and mass estimation in high-z galaxies.

astro-ph.CO↗

arXiv2011

Reconstructing the stellar mass distributions of galaxies using S4G IRAC 3.6 and 4.5 micron images: I. Correcting for contamination by PAH, hot dust, and intermediate age stars

With the aim of constructing accurate 2D maps of the stellar mass distribution in nearby galaxies from S4G 3.6 and 4.5 micron images, we report on the separation of the light from old stars from the emission contributed by contaminants (e.g. hot dust and the 3.3 micron PAH feature). Results for a small sample of six disk galaxies (NGC 1566, NGC 2976, NGC 3031, NGC 3184, NGC 4321, and NGC 5194) with a range of morphological properties, dust contents and star formation histories are presented to demonstrate our approach. We use an Independent Component Analysis (ICA) technique designed to separate statistically independent source distributions, maximizing the distinction in the [3.6]-[4.5] colors of the sources. The technique also removes emission from intermediate-age evolved red objects with a low mass-to-light ratio, such as asymptotic giant branch (AGB) and red supergiant (RSG) stars, revealing maps of the underlying old distribution of light with [3.6]-[4.5] colors consistent with the colors of K and M giants. Contaminants are identified via comparison to the non-stellar emission imaged at 8 microns, which is dominated by the broad PAH feature. Using the measured 3.6/8 micron ratio to select the individual contaminants, we find that hot dust and PAH together contribute between ~5-15% to the integrated light at 3.6 microns, while light from regions dominated by intermediate-age stars accounts for only 1-5%. Locally, however, the contribution from either contaminant can reach much higher levels; dust contributes on average 22% to the emission in star-forming regions throughout the sample, while intermediate age-stars contribute upwards of 50% in localized knots. The removal of these contaminants with ICA leaves maps of the old stellar disk that retain a high degree of structural information and are ideally suited for tracing the stellar mass, as will be the focus in a companion paper.

astro-ph.CO↗

arXiv2011

The Evolution of Stellar Populations in the Outer Disks of Spiral Galaxies

We investigate recent star formation in the extended ultraviolet (XUV) disks of five nearby galaxies (NGC 0628, NGC 2090, NGC 2841, NGC 3621, and NGC 5055) using a long wavelength baseline comprised of ultraviolet and mid-infrared imaging from the Galaxy Evolution Explorer and the Spitzer Infrared Array Camera. We identify 229 unresolved stellar complexes across targeted portions of their XUV disks and utilize spectral energy distribution fitting to measure their stellar ages and masses through comparison with Starburst99 population synthesis models of instantaneous burst populations. We find that the median age of outer disk associations in our sample is ~100 Myr with a large dispersion that spans the entire range of our models (1 Myr-1 Gyr). This relatively evolved state for most associations addresses the observed dearth of Halpha emission in some outer disks, as Halpha can only be observed in star forming regions younger than ~10 Myr. The large age dispersion is robust against variations in extinction (in the range E(B-V)=0-0.3 mag) and variations in the upper end of the stellar Initial Mass Function (IMF). In particular, we demonstrate that the age dispersion is insensitive to steepening of the IMF, up to extreme slopes.

astro-ph.CO↗

arXiv2011

New Constraints on Mass-Dependent Disruption of Star Clusters in M51

We use UBVI,Ha images of the Whirlpool galaxy, M51, taken with the ACS and WFPC2 cameras on the Hubble Space Telescope (HST) to select star clusters, and to estimate their masses and ages by comparing their observed colors with predictions from population synthesis models. We construct the mass function of intermediate age (1-4x10^8 yr) clusters, and find that it is well described by a power law, psi(M) propto M^beta, with beta=-2.1 +/- 0.2, for clusters more massive than approximately 6x10^3 Msun. This extends the mass function of intermediate age clusters in M51 to masses lower by nearly a factor of five over previous determinations. The mass function does not show evidence for curvature at either the high or low mass end. This shape indicates that there is no evidence for the earlier disruption of lower mass clusters compared with their higher mass counterparts (i.e., no mass-dependent disruption) over the observed range of masses and ages, or for a physical upper mass limit Mc with which clusters in M51 can form. These conclusions differ from previous suggestions based on poorer-quality HST observations. We discuss their implications for the formation and disruption of the clusters. Ages of clusters in two "feathers," stellar features extending from the outer portion of a spiral arm, show that the feather with a larger pitch angle formed earlier, and over a longer period, than the other.

astro-ph.GA↗

arXiv2010

The Spitzer Survey of Stellar Structure in Galaxies (S^4G)

The Spitzer Survey of Stellar Structure in Galaxies S^4G is an Exploration Science Legacy Program approved for the Spitzer post-cryogenic mission. It is a volume-, magnitude-, and size-limited (d < 40 Mpc, |b| > 30 degrees, m_(Bcorr) < 15.5, D25>1') survey of 2,331 galaxies using IRAC at 3.6 and 4.5 microns. Each galaxy is observed for 240 s and mapped to > 1.5 x D25. The final mosaicked images have a typical 1 sigma rms noise level of 0.0072 and 0.0093 MJy / sr at 3.6 and 4.5 microns, respectively. Our azimuthally-averaged surface brightness profile typically traces isophotes at mu_3.6 (AB) (1 sigma) ~ 27 mag arcsec^-2, equivalent to a stellar mass surface density of ~ 1 Msun pc^-2. S^4G thus provides an unprecedented data set for the study of the distribution of mass and stellar structures in the local Universe. This paper introduces the survey, the data analysis pipeline and measurements for a first set of galaxies, observed in both the cryogenic and warm mission phase of Spitzer. For every galaxy we tabulate the galaxy diameter, position angle, axial ratio, inclination at mu_3.6 (AB) = 25.5 and 26.5 mag arcsec^-2 (equivalent to ~ mu_B (AB) =27.2 and 28.2 mag arcsec^-2, respectively). These measurements will form the initial S^4G catalog of galaxy properties. We also measure the total magnitude and the azimuthally-averaged radial profiles of ellipticity, position angle, surface brightness and color. Finally, we deconstruct each galaxy using GALFIT into its main constituent stellar components: the bulge/spheroid, disk, bar, and nuclear point source, where necessary. Together these data products will provide a comprehensive and definitive catalog of stellar structures, mass and properties of galaxies in the nearby Universe.

astro-ph.CO↗