arXiv subjects

Christian Wolf

Publications and source records attributed to Christian Wolf.

At least 91 records · Page 5Linked to original sources

arXiv2021

Beyond spectroscopy. I. Metallicities, distances, and age estimates for over twenty million stars from SMSS DR2 and Gaia EDR3

Accurate determinations of stellar parameters and distances for large complete samples of stars are keys for conducting detailed studies of the formation and evolution of our Galaxy. Here we present stellar atmospheric parameters ($T_{\rm eff}$, luminosity classifications, and [Fe/H]) estimates for some 24 million stars determined from the stellar colors of SMSS DR2 and Gaia EDR3, based on training datasets with available spectroscopic measurements from previous high/medium/low-resolution spectroscopic surveys. The number of stars with photometric-metallicity estimates is 4-5 times larger than that collected by the current largest spectroscopic survey to date - LAMOST - over the course of the past decade. External checks indicate that the precision of the photometric-metallicity estimates are quite high, comparable to or slightly better than that derived from spectroscopy, with typical values around 0.05-0.15dex for both dwarf and giant stars with [Fe/H]>$-$1.0, 0.10-0.20dex for giant stars with $-$2.0<[Fe/H]<$-$1.0. and 0.20-0.25dex for giant stars with [Fe/H]<$-$2.0, and include estimates for stars as metal-poor as [Fe/H]~$-$3.5, substantially lower than previous photometric techniques. Photometric-metallicity estimates are obtained for an unprecedented number of metal-poor stars, including a total of over three million metal-poor (MP; [Fe/H] <$-$1.0) stars, over half a million very metal-poor (VMP; [Fe/H]<$-$2.0) stars, and over 25,000 extremely metal-poor (EMP; [Fe/H]<$-$3.0) stars. Moreover, distances are determined for over 20 million stars in our sample. For the over 18 million sample stars with accurate Gaia parallaxes, stellar ages are estimated by comparing with theoretical isochrones. Astrometric information is provided for the stars in our catalog, along with radial velocities for ~10% of our sample stars, taken from completed/ongoing large-scale spectroscopic surveys.

astro-ph.SR↗

arXiv2021

Universal Domain Adaptation in Ordinal Regression

We address the problem of universal domain adaptation (UDA) in ordinal regression (OR), which attempts to solve classification problems in which labels are not independent, but follow a natural order. We show that the UDA techniques developed for classification and based on the clustering assumption, under-perform in OR settings. We propose a method that complements the OR classifier with an auxiliary task of order learning, which plays the double role of discriminating between common and private instances, and expanding class labels to the private target images via ranking. Combined with adversarial domain discrimination, our model is able to address the closed set, partial and open set configurations. We evaluate our method on three face age estimation datasets, and show that it outperforms the baseline methods.

cs.CV↗

arXiv2021

Underwater Acoustic Networks for Security Risk Assessment in Public Drinking Water Reservoirs

We have built a novel system for the surveillance of drinking water reservoirs using underwater sensor networks. We implement an innovative AI-based approach to detect, classify and localize underwater events. In this paper, we describe the technology and cognitive AI architecture of the system based on one of the sensor networks, the hydrophone network. We discuss the challenges of installing and using the hydrophone network in a water reservoir where traffic, visitors, and variable water conditions create a complex, varying environment. Our AI solution uses an autoencoder for unsupervised learning of latent encodings for classification and anomaly detection, and time delay estimates for sound localization. Finally, we present the results of experiments carried out in a laboratory pool and the water reservoir and discuss the system's potential.

cs.AI↗

arXiv2021

VisQA: X-raying Vision and Language Reasoning in Transformers

Visual Question Answering systems target answering open-ended textual questions given input images. They are a testbed for learning high-level reasoning with a primary use in HCI, for instance assistance for the visually impaired. Recent research has shown that state-of-the-art models tend to produce answers exploiting biases and shortcuts in the training data, and sometimes do not even look at the input image, instead of performing the required reasoning steps. We present VisQA, a visual analytics tool that explores this question of reasoning vs. bias exploitation. It exposes the key element of state-of-the-art neural models -- attention maps in transformers. Our working hypothesis is that reasoning steps leading to model predictions are observable from attention distributions, which are particularly useful for visualization. The design process of VisQA was motivated by well-known bias examples from the fields of deep learning and vision-language reasoning and evaluated in two ways. First, as a result of a collaboration of three fields, machine learning, vision and language reasoning, and data analytics, the work lead to a better understanding of bias exploitation of neural models for VQA, which eventually resulted in an impact on its design and training through the proposition of a method for the transfer of reasoning patterns from an oracle model. Second, we also report on the design of VisQA, and a goal-oriented evaluation of VisQA targeting the analysis of a model decision process from multiple experts, providing evidence that it makes the inner workings of models accessible to users.

cs.CV↗

arXiv2021

Supervising the Transfer of Reasoning Patterns in VQA

Methods for Visual Question Anwering (VQA) are notorious for leveraging dataset biases rather than performing reasoning, hindering generalization. It has been recently shown that better reasoning patterns emerge in attention layers of a state-of-the-art VQA model when they are trained on perfect (oracle) visual inputs. This provides evidence that deep neural networks can learn to reason when training conditions are favorable enough. However, transferring this learned knowledge to deployable models is a challenge, as much of it is lost during the transfer. We propose a method for knowledge transfer based on a regularization term in our loss function, supervising the sequence of required reasoning operations. We provide a theoretical analysis based on PAC-learning, showing that such program prediction can lead to decreased sample complexity under mild hypotheses. We also demonstrate the effectiveness of this approach experimentally on the GQA dataset and show its complementarity to BERT-like self-supervised pre-training.

cs.CV↗

arXiv2021

Computability of topological pressure on compact shift spaces beyond finite type

We investigate the computability (in the sense of computable analysis) of the topological pressure $P_{\rm top}(ϕ)$ on compact shift spaces $X$ for continuous potentials $ϕ:X\to {\mathbb R}$. This question has recently been studied for subshifts of finite type (SFTs) and their factors (Sofic shifts). We develop a framework to address the computability of the topological pressure on general shift spaces and apply this framework to coded shifts. In particular, we prove the computability of the topological pressure for all continuous potentials on S-gap shifts, generalized gap shifts, and particular Beta-shifts. We also construct shift spaces which, depending on the potential, exhibit computability and non-computability of the topological pressure. We further prove that the generalized pressure function $(X,ϕ)\mapsto P_{\rm top}(X,ϕ\vert_{X})$ is not computable for a large set of shift spaces $X$ and potentials $ϕ$. In particular, the entropy map $X\mapsto h_{\rm top}(X)$ is computable at a shift space $X$ if and only if $X$ has zero topological entropy. Along the way of developing these computability results, we derive several ergodic-theoretical properties of coded shifts which are of independent interest beyond the realm of computability.

math.DS↗

arXiv2021

The impact and recovery of asteroid 2018 LA

The June 2, 2018, impact of asteroid 2018 LA over Botswana is only the second asteroid detected in space prior to impacting over land. Here, we report on the successful recovery of meteorites. Additional astrometric data refine the approach orbit and define the spin period and shape of the asteroid. Video observations of the fireball constrain the asteroid's position in its orbit and were used to triangulate the location of the fireball's main flare over the Central Kalahari Game Reserve. 23 meteorites were recovered. A consortium study of eight of these classifies Motopi Pan as a HED polymict breccia derived from howardite, cumulate and basaltic eucrite, and diogenite lithologies. Before impact, 2018 LA was a solid rock of about 156 cm diameter with high bulk density about 2.85 g/cm3, a relatively low albedo pV about 0.25, no significant opposition effect on the asteroid brightness, and an impact kinetic energy of about 0.2 kt. The orbit of 2018 LA is consistent with an origin at Vesta (or its Vestoids) and delivery into an Earth-impacting orbit via the nu_6 resonance. The impact that ejected 2018 LA in an orbit towards Earth occurred 22.8 +/- 3.8 Ma ago. Zircons record a concordant U-Pb age of 4563 +/- 11 Ma and a consistent 207Pb/206Pb age of 4563 +/- 6 Ma. A much younger Pb-Pb phosphate resetting age of 4234 +/- 41 Ma was found. From this impact chronology, we discuss what is the possible source crater of Motopi Pan and the age of Vesta's Veneneia impact basin.

astro-ph.EP↗

arXiv2021

Estimating semantic structure for the VQA answer space

Since its appearance, Visual Question Answering (VQA, i.e. answering a question posed over an image), has always been treated as a classification problem over a set of predefined answers. Despite its convenience, this classification approach poorly reflects the semantics of the problem limiting the answering to a choice between independent proposals, without taking into account the similarity between them (e.g. equally penalizing for answering cat or German shepherd instead of dog). We address this issue by proposing (1) two measures of proximity between VQA classes, and (2) a corresponding loss which takes into account the estimated proximity. This significantly improves the generalization of VQA models by reducing their language bias. In particular, we show that our approach is completely model-agnostic since it allows consistent improvements with three different VQA models. Finally, by combining our method with a language bias reduction approach, we report SOTA-level performance on the challenging VQAv2-CP dataset.

cs.CV↗

arXiv2021

How Transferable are Reasoning Patterns in VQA?

Since its inception, Visual Question Answering (VQA) is notoriously known as a task, where models are prone to exploit biases in datasets to find shortcuts instead of performing high-level reasoning. Classical methods address this by removing biases from training data, or adding branches to models to detect and remove biases. In this paper, we argue that uncertainty in vision is a dominating factor preventing the successful learning of reasoning in vision and language problems. We train a visual oracle and in a large scale study provide experimental evidence that it is much less prone to exploiting spurious dataset biases compared to standard models. We propose to study the attention mechanisms at work in the visual oracle and compare them with a SOTA Transformer-based model. We provide an in-depth analysis and visualizations of reasoning patterns obtained with an online visualization tool which we make publicly available (https://reasoningpatterns.github.io). We exploit these insights by transferring reasoning patterns from the oracle to a SOTA Transformer-based VQA model taking standard noisy visual inputs via fine-tuning. In experiments we report higher overall accuracy, as well as accuracy on infrequent answers for each question type, which provides evidence for improved generalization and a decrease of the dependency on dataset biases.

cs.CV↗

arXiv2021

Roses Are Red, Violets Are Blue... but Should Vqa Expect Them To?

Models for Visual Question Answering (VQA) are notorious for their tendency to rely on dataset biases, as the large and unbalanced diversity of questions and concepts involved and tends to prevent models from learning to reason, leading them to perform educated guesses instead. In this paper, we claim that the standard evaluation metric, which consists in measuring the overall in-domain accuracy, is misleading. Since questions and concepts are unbalanced, this tends to favor models which exploit subtle training set statistics. Alternatively, naively introducing artificial distribution shifts between train and test splits is also not completely satisfying. First, the shifts do not reflect real-world tendencies, resulting in unsuitable models; second, since the shifts are handcrafted, trained models are specifically designed for this particular setting, and do not generalize to other configurations. We propose the GQA-OOD benchmark designed to overcome these concerns: we measure and compare accuracy over both rare and frequent question-answer pairs, and argue that the former is better suited to the evaluation of reasoning abilities, which we experimentally validate with models trained to more or less exploit biases. In a large-scale study involving 7 VQA models and 3 bias reduction techniques, we also experimentally demonstrate that these models fail to address questions involving infrequent concepts and provide recommendations for future directions of research.

cs.CV↗

arXiv2021

SkyMapper Optical Follow-up of Gravitational Wave Triggers: Alert Science Data Pipeline and LIGO/Virgo O3 Run

We present an overview of the SkyMapper optical follow-up program for gravitational-wave event triggers from the LIGO/Virgo observatories, which aims at identifying early GW170817-like kilonovae out to $\sim 200$ Mpc distance. We describe our robotic facility for rapid transient follow-up, which can target most of the sky at $δ<+10°$ to a depth of $i_\mathrm{AB}\approx 20$ mag. We have implemented a new software pipeline to receive LIGO/Virgo alerts, schedule observations and examine the incoming real-time data stream for transient candidates. We adopt a real-bogus classifier using ensemble-based machine learning techniques, attaining high completeness ($\sim$98%) and purity ($\sim$91%) over our whole magnitude range. Applying further filtering to remove common image artefacts and known sources of transients, such as asteroids and variable stars, reduces the number of candidates by a factor of more than 10. We demonstrate the system performance with data obtained for GW190425, a binary neutron star merger detected during the LIGO/Virgo O3 observing campaign. In time for the LIGO/Virgo O4 run, we will have deeper reference images allowing transient detection to $i_\mathrm{AB}\approx $21 mag.

astro-ph.IM↗

arXiv2021

SSTVOS: Sparse Spatiotemporal Transformers for Video Object Segmentation

In this paper we introduce a Transformer-based approach to video object segmentation (VOS). To address compounding error and scalability issues of prior work, we propose a scalable, end-to-end method for VOS called Sparse Spatiotemporal Transformers (SST). SST extracts per-pixel representations for each object in a video using sparse attention over spatiotemporal features. Our attention-based formulation for VOS allows a model to learn to attend over a history of multiple frames and provides suitable inductive bias for performing correspondence-like computations necessary for solving motion segmentation. We demonstrate the effectiveness of attention-based over recurrent networks in the spatiotemporal domain. Our method achieves competitive results on YouTube-VOS and DAVIS 2017 with improved scalability and robustness to occlusions compared with the state of the art. Code is available at https://github.com/dukebw/SSTVOS.

cs.CV↗

arXiv2021

KiDS-1000 Cosmology: constraints beyond flat $Λ$CDM

We present constraints on extensions to the flat $Λ$CDM cosmological model by varying the spatial curvature $Ω_K$, the sum of the neutrino masses $\sum m_ν$, the dark energy equation of state parameter $w$, and the Hu-Sawicki $f(R)$ gravity $f_{R0}$ parameter. With the combined $3\times2$pt measurements of cosmic shear from the Kilo-Degree Survey (KiDS-1000), galaxy clustering from the Baryon Oscillation Spectroscopic Survey (BOSS), and galaxy-galaxy lensing from the overlap between KiDS-1000, BOSS, and the spectroscopic 2-degree Field Lensing Survey (2dFLenS), we find results that are fully consistent with a flat $Λ$CDM model with $Ω_K=0.011^{+0.054}_{-0.057}$, $\sum m_ν<1.76$ eV (95% CL), and $w=-0.99^{+0.11}_{-0.13}$. The $f_{R0}$ parameter is unconstrained in our fully non-linear $f(R)$ cosmic shear analysis. Considering three different model selection criteria, we find no clear preference for either the fiducial flat $Λ$CDM model or any of the considered extensions. Besides extensions to the flat $Λ$CDM parameter space, we also explore restrictions to common subsets of the flat $Λ$CDM parameter space by fixing the amplitude of the primordial power spectrum to the Planck best-fit value, as well as adding external data from supernovae and lensing of the CMB. Neither the beyond-$Λ$CDM models nor the imposed restrictions explored in this analysis are able to resolve the $\sim 3σ$ tension in $S_8$ between the $3\times2$pt constraints and Planck, with the exception of $w$CDM, where the $S_8$ tension is resolved. The tension in the $w$CDM case persists, however, when considering the joint $S_8$-$w$ parameter space. The joint flat $Λ$CDM CMB lensing and $3\times2$pt analysis is found to yield tight constraints on $Ω_{\rm m}=0.307^{+0.008}_{-0.013}$, $σ_8=0.769^{+0.022}_{-0.010}$, and $S_8=0.779^{+0.013}_{-0.013}$.

astro-ph.CO↗

arXiv2021

A Systematic Survey for z < 0.04 Changing-Look AGNs

We have conducted a systematic survey for z $<$ 0.04 active Galactic nuclei (AGNs) that may have changed spectral class over the past decade. We use SkyMapper, Pan-STARRS and the Véron-Cetty & Véron (2010) catalogue to search the entire sky for these ``changing-look'' AGNs using a variety of selection methods, where Pan-STARRS has a coverage of 3$π$ steradians (sky north of Declination $-30^\circ$) and SkyMapper has coverage of $\sim$ 21,000$~\rm{deg^2}$ (sky south of Declination $0^\circ$). We use small aperture photometry to measure how colour and flux have changed over time, where a change may indicate a change in spectral type. Optical colour and flux are used as a proxy for changing H$α$ equivalent width, while WISE 3.4 $μ$m flux is used to look for changes in the hot dust component. We have identified four AGNs with varying spectra selected using our optical colour selection method. Three AGNs were confirmed from recent observations with WiFeS on the 2.3 m telescope at Siding Spring and the other was identified from archival spectra alone. From this, we identify two new changing look AGNs; NGC 1346 and 2MASX J20075129-1108346. We also recover Mrk 915 and Mrk 609, which are known to have varying spectra in the literature, but they do not meet our specific criteria for changing look AGNs.

astro-ph.GA↗

arXiv2020

KiDS-1000 Cosmology: Multi-probe weak gravitational lensing and spectroscopic galaxy clustering constraints

We present a joint cosmological analysis of weak gravitational lensing observations from the Kilo-Degree Survey (KiDS-1000), with redshift-space galaxy clustering observations from the Baryon Oscillation Spectroscopic Survey (BOSS), and galaxy-galaxy lensing observations from the overlap between KiDS-1000, BOSS and the spectroscopic 2-degree Field Lensing Survey (2dFLenS). This combination of large-scale structure probes breaks the degeneracies between cosmological parameters for individual observables, resulting in a constraint on the structure growth parameter $S_8=σ_8 \sqrt{Ω_{\rm m}/0.3} = 0.766^{+0.020}_{-0.014}$, that has the same overall precision as that reported by the full-sky cosmic microwave background observations from Planck. The recovered $S_8$ amplitude is low, however, by $8.3 \pm 2.6$ % relative to Planck. This result builds from a series of KiDS-1000 analyses where we validate our methodology with variable depth mock galaxy surveys, our lensing calibration with image simulations and null-tests, and our optical-to-near-infrared redshift calibration with multi-band mock catalogues and a spectroscopic-photometric clustering analysis. The systematic uncertainties identified by these analyses are folded through as nuisance parameters in our cosmological analysis. Inspecting the offset between the marginalised posterior distributions, we find that the $S_8$-difference with Planck is driven by a tension in the matter fluctuation amplitude parameter, $σ_8$. We quantify the level of agreement between the CMB and our large-scale structure constraints using a series of different metrics, finding differences with a significance ranging between $\sim\! 3\,σ$, when considering the offset in $S_{8}$, and $\sim\! 2\,σ$, when considering the full multi-dimensional parameter space.

astro-ph.CO↗

arXiv2020

Milky Way Tomography with the SkyMapper Southern Survey. II. Photometric Re-calibration of SMSS DR2

We apply the spectroscopy-based stellar-color regression (SCR) method to perform an accurate photometric re-calibration of the second data release from the SkyMapper Southern Survey (SMSS DR2). From comparison with a sample of over 200,000 dwarf stars with stellar atmospheric parameters taken from GALAH+ DR3 and with accurate, homogeneous photometry from $Gaia$ DR2, zero-point offsets are detected in the original photometric catalog of SMSS DR2, in particular for the gravity- and metallicity-sensitive $uv$ bands. For $uv$ bands, the zero-point offsets are close to zero at very low extinction, and then steadily increase with $E (B - V)$, reaching as large as 0.174 and 0.134 mag respectively, at $E (B - V) \sim 0.5$ mag. These offsets largely arise from the adopted dust term in the transformations used by SMSS DR2 to construct photometric calibrators from the ATLAS reference catalog. For the $gr$ bands, the zero-point offsets exhibit negligible variations with SFD $E(B - V )$, due to their tiny coefficients on the dust term in the transformation. Our study also reveals small, but significant, spatial variations of the zero-point offsets in all $uvgr$ bands. External checks using Strömgren photometry, WD loci and the SDSS Stripe 82 standard-star catalog independently confirm the zero-points found by our revised SCR method.

astro-ph.GA↗

arXiv2020

Entropy spectrum of rotation classes

In this note we study the entropy spectrum of rotation classes for collections of finitely many continuous potentials $φ_1,\dots,φ_m:X\to \mathbb{R}$ with respect to the set of invariant measures of an underlying dynamical system $f:X\to X$. We show for large classes of dynamical systems and potentials that these entropy spectra are maximal in the sense that every value between zero and the maximum is attained. We also provide criteria that imply the maximality of the ergodic entropy spectra. For $m$ being large, our results can be interpreted as a complimentary result to the classical Riesz representation theorem in the dynamical context.

math.DS↗

arXiv2020

SkyMapper colours of Seyfert galaxies and Changing-Look AGN

We study the utility of broad-band colours in the SkyMapper Southern Survey for selecting Seyfert galaxies at low luminosity. We find that the $u-v$ index, built from the ultraviolet $u$ and violet $v$ filters, separates normal galaxies, starburst galaxies and type-1 AGN. This $u-v$ index is not sensitive to age or metallicity in a stellar population but is instead a quenching-and-bursting indicator in galaxies and detects power-law continua in type-1 AGN. Using over 25,000 galaxies at $z<0.1$ from 6dFGS, we find a selection cut based on $u-v$ and central $u$ band brightness that identifies type-1 AGN. By eyeballing 6dFGS spectra we classify new Seyfert galaxies of type 1 to 1.8. Our sample includes eight known Changing-Look AGN, two of which show such strong variability that they move across the selection cut during the five years of SkyMapper observations in DR3, along mixing sequences of nuclear and host galaxy light. We identify 46 Changing-Look AGN candidates in our sample, one of which has been reported as a type-IIn supernova. We show that this transient persists for at least five years and marks a flare in a Seyfert-1 period of a new Changing-Look AGN.

astro-ph.GA↗