-
How Intermodal Interaction Affects the Performance of Deep Multimodal Fusion for Mixed-Type Time Series
Authors:
Simon Dietz,
Thomas Altstidl,
Dario Zanca,
Björn Eskofier,
An Nguyen
Abstract:
Mixed-type time series (MTTS) is a bimodal data type that is common in many domains, such as healthcare, finance, environmental monitoring, and social media. It consists of regularly sampled continuous time series and irregularly sampled categorical event sequences. The integration of both modalities through multimodal fusion is a promising approach for processing MTTS. However, the question of ho…
▽ More
Mixed-type time series (MTTS) is a bimodal data type that is common in many domains, such as healthcare, finance, environmental monitoring, and social media. It consists of regularly sampled continuous time series and irregularly sampled categorical event sequences. The integration of both modalities through multimodal fusion is a promising approach for processing MTTS. However, the question of how to effectively fuse both modalities remains open. In this paper, we present a comprehensive evaluation of several deep multimodal fusion approaches for MTTS forecasting. Our comparison includes three fusion types (early, intermediate, and late) and five fusion methods (concatenation, weighted mean, weighted mean with correlation, gating, and feature sharing). We evaluate these fusion approaches on three distinct datasets, one of which was generated using a novel framework. This framework allows for the control of key data properties, such as the strength and direction of intermodal interactions, modality imbalance, and the degree of randomness in each modality, providing a more controlled environment for testing fusion approaches. Our findings show that the performance of different fusion approaches can be substantially influenced by the direction and strength of intermodal interactions. The study reveals that early and intermediate fusion approaches excel at capturing fine-grained and coarse-grained cross-modal features, respectively. These findings underscore the crucial role of intermodal interactions in determining the most effective fusion strategy for MTTS forecasting.
△ Less
Submitted 21 June, 2024;
originally announced June 2024.
-
Contrastive Language-Image Pretrained Models are Zero-Shot Human Scanpath Predictors
Authors:
Dario Zanca,
Andrea Zugarini,
Simon Dietz,
Thomas R. Altstidl,
Mark A. Turban Ndjeuha,
Leo Schwinn,
Bjoern Eskofier
Abstract:
Understanding the mechanisms underlying human attention is a fundamental challenge for both vision science and artificial intelligence. While numerous computational models of free-viewing have been proposed, less is known about the mechanisms underlying task-driven image exploration. To address this gap, we present CapMIT1003, a database of captions and click-contingent image explorations collecte…
▽ More
Understanding the mechanisms underlying human attention is a fundamental challenge for both vision science and artificial intelligence. While numerous computational models of free-viewing have been proposed, less is known about the mechanisms underlying task-driven image exploration. To address this gap, we present CapMIT1003, a database of captions and click-contingent image explorations collected during captioning tasks. CapMIT1003 is based on the same stimuli from the well-known MIT1003 benchmark, for which eye-tracking data under free-viewing conditions is available, which offers a promising opportunity to concurrently study human attention under both tasks. We make this dataset publicly available to facilitate future research in this field. In addition, we introduce NevaClip, a novel zero-shot method for predicting visual scanpaths that combines contrastive language-image pretrained (CLIP) models with biologically-inspired neural visual attention (NeVA) algorithms. NevaClip simulates human scanpaths by aligning the representation of the foveated visual stimulus and the representation of the associated caption, employing gradient-driven visual exploration to generate scanpaths. Our experimental results demonstrate that NevaClip outperforms existing unsupervised computational models of human visual attention in terms of scanpath plausibility, for both captioning and free-viewing tasks. Furthermore, we show that conditioning NevaClip with incorrect or misleading captions leads to random behavior, highlighting the significant impact of caption guidance in the decision-making process. These findings contribute to a better understanding of mechanisms that guide human attention and pave the way for more sophisticated computational approaches to scanpath prediction that can integrate direct top-down guidance of downstream tasks.
△ Less
Submitted 23 May, 2023; v1 submitted 21 May, 2023;
originally announced May 2023.
-
Three-body resonances in pionless effective field theory
Authors:
Sebastian Dietz,
Hans-Werner Hammer,
Sebastian König,
Achim Schwenk
Abstract:
We investigate the appearance of resonances in three-body systems using pionless effective field theory at leading order. The Faddeev equation is analytically continued to the unphysical sheet adjacent to the positive real energy axis using a contour rotation. We consider both, the three-boson system and the three-neutron system. For the former, we calculate the trajectory of Borromean three-body…
▽ More
We investigate the appearance of resonances in three-body systems using pionless effective field theory at leading order. The Faddeev equation is analytically continued to the unphysical sheet adjacent to the positive real energy axis using a contour rotation. We consider both, the three-boson system and the three-neutron system. For the former, we calculate the trajectory of Borromean three-body Efimov states turning into resonances as they cross the three-body threshold. For the latter, we find no sign of three-body resonances or virtual states at leading order. This result is validated by exploring the level structure of three-body states in a finite volume approach.
△ Less
Submitted 23 July, 2022; v1 submitted 23 September, 2021;
originally announced September 2021.
-
The R-Process Alliance: Chemo-Dynamically Tagged Groups of Halo $r$-Process-Enhanced Stars Reveal a Shared Chemical-Evolution History
Authors:
Dmitrii Gudin,
Derek Shank,
Timothy C. Beers,
Zhen Yuan,
Guilherme Limberg,
Ian U. Roederer,
Vinicius Placco,
Erika M. Holmbeck,
Sarah Dietz,
Kaitlin C. Rasmussen,
Terese T. Hansen,
Charli M. Sakari,
Rana Ezzeddine,
Anna Frebel
Abstract:
We derive dynamical parameters for a large sample of 446 $r$-process-enhanced (RPE) metal-poor stars in the halo and disk systems of the Milky Way, based on data releases from the $R$-Process Alliance, supplemented by additional literature samples. This sample represents more than a ten-fold increase in size relative to that previously considered by Roederer et al., and, by design, covers a larger…
▽ More
We derive dynamical parameters for a large sample of 446 $r$-process-enhanced (RPE) metal-poor stars in the halo and disk systems of the Milky Way, based on data releases from the $R$-Process Alliance, supplemented by additional literature samples. This sample represents more than a ten-fold increase in size relative to that previously considered by Roederer et al., and, by design, covers a larger range of $r$-process-element enrichment levels. We test a number of clustering analysis methods on the derived orbital energies and other dynamical parameters for this sample, ultimately deciding on application of the HDBSCAN algorithm, which obtains 30 individual Chemo-Dynamically Tagged Groups (CDTGs); 21 contain between 3 and 5 stars, and 9 contain between 6 and 12 stars. Even though the clustering was performed solely on the basis of their dynamical properties, the stars in these CDTGs exhibit statistically significant similarities in their metallicity ([Fe/H]), carbonicity ([C/Fe]), and neutron-capture element ratios ([Sr/Fe], [Ba/Fe], and [Eu/Fe]). These results demonstrate that the RPE stars in these CDTGs have likely experienced common chemical-evolution histories, presumably in their parent satellite galaxies or globular clusters, prior to being disrupted into the Milky Way's halo. We also confirm the previous claim that the orbits of the RPE stars preferentially exhibit pericentric distances that are substantially lower than the present distances of surviving ultra-faint dwarf and canonical dwarf spheroidal galaxies, consistent with the disruption hypothesis. The derived dynamical parameters for several of our CDTGs indicate their association with previously known substructures, Dynamically Tagged Groups, and RPE Groups.
△ Less
Submitted 29 December, 2020; v1 submitted 26 December, 2020;
originally announced December 2020.
-
Two Populations of Carbon-Enhanced Metal-Poor Stars in the Disk System of the Milky Way
Authors:
Sarah E. Dietz,
**mi Yoon,
Timothy C. Beers,
Vinicius Placco,
Young Sun Lee
Abstract:
We present a chemo-dynamical analysis of low-resolution ($R \sim 1300$) spectroscopy of stars from the AAOmega Evolution of Galactic Structure (AEGIS) survey, focusing on two key populations of carbon-enhanced metal-poor (CEMP) stars within the disk system of the Milky Way: a mildly prograde population ($L_z < 1000\,$kpc$\,$km$\,$s$^{-1}$) and a strongly prograde ($L_z > 1000\,$kpc$\,$km$\,$s…
▽ More
We present a chemo-dynamical analysis of low-resolution ($R \sim 1300$) spectroscopy of stars from the AAOmega Evolution of Galactic Structure (AEGIS) survey, focusing on two key populations of carbon-enhanced metal-poor (CEMP) stars within the disk system of the Milky Way: a mildly prograde population ($L_z < 1000\,$kpc$\,$km$\,$s$^{-1}$) and a strongly prograde ($L_z > 1000\,$kpc$\,$km$\,$s$^{-1}$) population. Based on their chemical and kinematic characteristics, and on comparisons with similar populations found in the recent literature, we tentatively associate the former with an ex-situ inner-halo population originating from either the $Gaia$ Sausage or $Gaia$-Enceladus. The latter population is linked to the metal-weak thick-disk (MWTD). We discuss their implications in the context of the formation history of the Milky Way.
△ Less
Submitted 11 December, 2020; v1 submitted 7 December, 2020;
originally announced December 2020.
-
Metal-Poor Stars Observed with the Southern African Large Telescope
Authors:
Kaitlin C. Rasmussen,
Joseph Zepeda,
Timothy C. Beers,
Vinicius M. Placco,
Eric Depagne,
Anna Frebel,
Sarah Dietz,
Tilman Hartwig
Abstract:
We present the first release of a large-scale study of relatively bright (V < 13.5) metal-poor stars observed with the Southern African Large Telescope (SALT), based on high-resolution spectra of 50 stars with a resolving power of R ~ 40,000 and S/N ~ 20 per pixel at 4300 Angstrom. The elemental abundances of C, Sr, Ba, and Eu are reported, as well as several alpha-elements (Mg, Ca, Sc, Ti, V) and…
▽ More
We present the first release of a large-scale study of relatively bright (V < 13.5) metal-poor stars observed with the Southern African Large Telescope (SALT), based on high-resolution spectra of 50 stars with a resolving power of R ~ 40,000 and S/N ~ 20 per pixel at 4300 Angstrom. The elemental abundances of C, Sr, Ba, and Eu are reported, as well as several alpha-elements (Mg, Ca, Sc, Ti, V) and iron-peak elements (Mn, Co, Ni, Zn). We find a diverse array of abundance patterns, including several consistent with the signatures of carbon-enhanced metal-poor CEMP-i and CEMP-r stars. We find that 15 of 50 (30%) are carbon enhanced (with [C/Fe] >+ 0.70), and that a large fraction (26 of 50, 52%) are enhanced in r-process elements, among the r-process-enhanced stars, five are strongly enhanced r-II ([Eu/Fe] >+ 1.0) stars (two of which are newly discovered) and 21 are newly discovered moderately enhanced r-I (+0.3 <= [Eu/Fe] <=+ 1..0) stars. There are eight stars in our sample that, on the basis of their abundances and kinematics, are possible members of the metal-weak thick-disk population. We also compare our measured abundances to progenitor-enrichment models, and find that the abundance patterns for the majority of our stars can be attributed to a single (rather than multiple) enrichment event.
△ Less
Submitted 8 October, 2020;
originally announced October 2020.
-
On the Absence of Triplet Exciton Loss Pathways in Non-Fullerene Acceptor based Organic Solar Cells
Authors:
Maria S. Kotova,
Giacomo Londi,
Johannes Junker,
Stefanie Dietz,
Alberto Privitera,
Kristofer Tvingstedt,
David Beljonne,
Andreas Sperlich,
Vladimir Dyakonov
Abstract:
We investigate the viability of highly efficient organic solar cells (OSCs) based on non-fullerene acceptors (NFA) by taking into consideration efficiency loss channels and stability issues caused by triplet excitons (TE) formation. OSCs based on a blend of the conjugated donor polymer PBDB-T and ITIC as acceptor were fabricated and investigated with electrical, optical and spin-sensitive methods.…
▽ More
We investigate the viability of highly efficient organic solar cells (OSCs) based on non-fullerene acceptors (NFA) by taking into consideration efficiency loss channels and stability issues caused by triplet excitons (TE) formation. OSCs based on a blend of the conjugated donor polymer PBDB-T and ITIC as acceptor were fabricated and investigated with electrical, optical and spin-sensitive methods. The spin-Hamiltonian parameters of molecular TEs and charge transfer TEs in ITIC e.g., zero-field splitting and charge distribution, were calculated by Density Functional Theory (DFT) modelling. In addition, the energetic model describing the photophysical processes in the donor-acceptor blend was derived. Spin-sensitive photoluminescence measurements prove the formation of charge transfer (CT) states in the blend and the formation of TEs in the pure materials and the blend. However, no molecular TE signal is observed in the completed devices under working conditions by spin-sensitive electrical measurements. The absence of a molecular triplet state population allows to eliminate a charge carrier loss channel and irreversible photooxidation facilitated by long-lived triplet states. These results correlate well with the high power conversion efficiency of the PBDB-T:ITIC-based OSCs and their high stability.
△ Less
Submitted 7 April, 2020; v1 submitted 18 February, 2020;
originally announced February 2020.
-
The Metallicity Gradient and Complex Formation History of the Outermost Halo of the Milky Way
Authors:
Sarah E. Dietz,
**mi Yoon,
Timothy C. Beers,
Vinicius M. Placco
Abstract:
We present an examination of the metallicity distribution function of the outermost stellar halo of the Galaxy based on an analysis of both local (within 4 kpc of the Sun, ~16,500 stars) and non-local (~21,700 stars) samples. These samples were compiled using spectroscopic metallicities from the Sloan Digital Sky Survey and photometric metallicities from the SkyMapper Southern Survey. We detect a…
▽ More
We present an examination of the metallicity distribution function of the outermost stellar halo of the Galaxy based on an analysis of both local (within 4 kpc of the Sun, ~16,500 stars) and non-local (~21,700 stars) samples. These samples were compiled using spectroscopic metallicities from the Sloan Digital Sky Survey and photometric metallicities from the SkyMapper Southern Survey. We detect a negative metallicity gradient in the outermost halo (r > 35 kpc from the Galactic center), and find that the frequency of very metal-poor ([Fe/H] < -2.0) stars in the outer-halo region reaches up to ~60% in our most distant sample, commensurate with previous theoretical predictions. This result provides clear evidence that the outer-halo formed hierarchically. The retrograde stars in the outermost halo exhibit a roughly constant metallicity, which may be linked to the accretion of the Sequoia progenitor. In contrast, prograde stars in the outermost halo exhibit a strong metallicity-distance dependence, indicating that they likely originated from the accretion of galaxies less massive than the Sequoia progenitor galaxy.
△ Less
Submitted 4 March, 2020; v1 submitted 25 November, 2019;
originally announced November 2019.
-
Galactic Archeology with the AEGIS Survey: The Evolution of Carbon and Iron in the Galactic Halo
Authors:
**mi Yoon,
Timothy C. Beers,
Sarah Dietz,
Young Sun Lee,
Vinicius M. Placco,
Gary Da Costa,
Stefan Keller,
Christopher I. Owen,
Mahavir Sharma
Abstract:
Understanding the evolution of carbon and iron in the Milky Way's halo is of importance because these two elements play crucial roles constraining star formation, Galactic assembly, and chemical evolution in the early Universe. Here, we explore the spatial distributions of carbonicity, [C/Fe], and metallicity, [Fe/H], of the halo system based on medium-resolution ($R \sim$ 1,300) spectroscopy of…
▽ More
Understanding the evolution of carbon and iron in the Milky Way's halo is of importance because these two elements play crucial roles constraining star formation, Galactic assembly, and chemical evolution in the early Universe. Here, we explore the spatial distributions of carbonicity, [C/Fe], and metallicity, [Fe/H], of the halo system based on medium-resolution ($R \sim$ 1,300) spectroscopy of $\sim$58,000 stars in the Southern Hemisphere from the AAOmega Evolution of Galactic Structure (AEGIS) survey. The AEGIS carbonicity map exhibits a positive gradient with distance, as similarly found for the Sloan Digital Sky Survey (SDSS) carbonicity map of Lee et al. The metallicity map confirms that [Fe/H] decreases with distance, from the inner halo to the outer halo. We also explore the formation and chemical-evolution history of the halo by considering the populations of carbon-enhanced metal-poor (CEMP) stars present in the AEGIS sample. The cumulative and differential frequencies of CEMP-no stars (as classified by their characteristically lower levels of absolute carbon abundance, $A$(C) $\leq$ 7.1 for sub-giants and giants) increases with decreasing metallicity, and is \textit{substantially higher than previous determinations} for CEMP stars as a whole. In contrast, that of CEMP-$s$ stars (with higher $A$(C)), remains almost flat, at a value $\sim$10\%, in the range $-\,4.0 \lesssim$ [Fe/H] $\lesssim-$2.0. The distinctly different behaviors of the CEMP-no and CEMP-$s$ stars relieve the tension with population-synthesis models assuming a binary mass-transfer origin, which previously struggled to account for the higher reported frequencies of CEMP stars, taken as a whole, at low metallicity.
△ Less
Submitted 12 June, 2018;
originally announced June 2018.
-
Bright Metal-Poor Stars from the Hamburg/ESO Survey. II. A Chemodynamical Analysis
Authors:
Timothy C. Beers,
Vinicius M. Placco,
Daniela Carollo,
Silvia Rossi,
Young Sun Lee,
Anna Frebel,
John E. Norris,
Sarah Dietz,
Thomas Masseron
Abstract:
We obtain estimates of stellar atmospheric parameters for a previously published sample of 1777 relatively bright (9 < B < 14) metal-poor candidates from the Hamburg/ESO Survey. The original Frebel et al. analysis of these stars was only able to derive estimates of [Fe/H] and [C/Fe] for a subset of the sample, due to limitations in the methodology then available. A new spectroscopic analysis pipel…
▽ More
We obtain estimates of stellar atmospheric parameters for a previously published sample of 1777 relatively bright (9 < B < 14) metal-poor candidates from the Hamburg/ESO Survey. The original Frebel et al. analysis of these stars was only able to derive estimates of [Fe/H] and [C/Fe] for a subset of the sample, due to limitations in the methodology then available. A new spectroscopic analysis pipeline has been used to obtain estimates of Teff, log g, [Fe/H], and [C/Fe] for almost the entire dataset. This sample is very local - about 90% of the stars are located within 0.5 kpc of the Sun. We consider the chemodynamical properties of these stars in concert with a similarly local sample of stars from a recent analysis of the Bidelman & MacConnell 'weak-metal' candidates by Beers et al. We use this combined sample to identify possible members of the suggested halo stream of stars by Helmi et al. and Chiba & Beers, as well as stars that may be associated with stripped debris from the putative parent dwarf of the globular cluster Omega Centauri, suggested to exist by previous authors. We identify a clear increase in the cumulative frequency of carbon-enhanced metal-poor (CEMP) stars with declining metallicity, as well as an increase in the fraction of CEMP stars with distance from the Galactic plane, consistent with previous results. We also identify a relatively large number of CEMP stars with kinematics consistent with the metal-weak thick-disk population, with possible implications for its origin.
△ Less
Submitted 15 November, 2016; v1 submitted 11 November, 2016;
originally announced November 2016.
