Searcharxiv⌕ Search

arXiv subjects

Christian Wolf

Publications and source records attributed to Christian Wolf.

At least 145 records · Page 8Linked to original sources

On the computability of rotation sets and their entropies

Given a continuous dynamical system $f:X\to X$ on a compact metric space $X$ and an $m$-dimensional continuous potential $Φ:X\to \mathbb R^m$, the (generalized) rotation set ${\rm Rot}(Φ)$ is defined as the set of all $μ$-integrals of $Φ$, where $μ$ runs over all invariant probability measures. Analogous to the classical topological entropy, one can associate the localized entropy ${\mathcal H}(w)$ to each $w\in {\rm Rot}(Φ)$. In this paper, we study the computability of rotation sets and localized entropy functions by deriving conditions that imply their computability. We then apply our results to study to the case of subshifts of finite type. We prove that ${\rm Rot}(Φ)$ is computable and that ${\mathcal H}(w)$ is computable in the interior of the rotation set. Finally, we construct an explicit example that shows that, in general, ${\mathcal H}$ is not continuous on the boundary of the rotation set, when considered as a function of $Φ$ and $w$. This suggests that, in general, ${\mathcal H}$ is not computable at the boundary of rotation sets.

math.DS↗

OMEGA - OSIRIS Mapping of Emission-line Galaxies in A901/2: III. - Galaxy Properties Across Projected Phase Space in A901/2

We conduct a comprehensive projected phase-space analysis of the A901/2 multi-cluster system at $z\sim0.165$. Aggregating redshifts from spectroscopy, tunable-filter imaging, and prism techniques, we assemble a sample of 856 cluster galaxies reaching $10^{8.5}M_\odot$ in stellar mass. We look for variations in cluster galaxy properties between virialised and non-virialised regions of projected phase-space (PPS). Our main conclusions point to relatively gentle environmental effects, expressed mainly on galaxy gas reservoirs. (1) Stacking the four subclusters in A901/2, we find galaxies in the virialised region are more massive, redder, and have marginally higher Sèrsic indices, but their half-light radii and Hubble types are not significantly different. (2) After accounting for trends in stellar mass, there is a remaining change in rest-frame colour across PPS. Primarily, the colour difference is due to an absence in the virialised region of galaxies with rest-frame $B-V<0.7$ and moderate-to-high ($M_\star>10^{9.85}M_\odot$) stellar mass. (3) There is an infalling population of lower-mass ($M_\star\leq10^{9.85}M_\odot$), relatively blue ($B-V<0.7$) elliptical or spheroidal galaxies that is strikingly absent in the virialised region. (4) The number of bona-fide star-forming and AGN galaxies in the PPS regions is strongly dictated by stellar mass. However, there remains a reduced fraction of star-forming galaxies in the centres of the clusters at fixed stellar mass, consistent with the star-formation-density relation in galaxy clusters. (5) There is no change in specific H$α$-derived star-formation rates of star-forming galaxies at fixed mass across the cluster environment. This suggests that preprocessing of galaxies during infall plays a prominent role in quenching star formation.

astro-ph.GA↗

Full-Page Text Recognition: Learning Where to Start and When to Stop

Text line detection and localization is a crucial step for full page document analysis, but still suffers from heterogeneity of real life documents. In this paper, we present a new approach for full page text recognition. Localization of the text lines is based on regressions with Fully Convolutional Neural Networks and Multidimensional Long Short-Term Memory as contextual layers. In order to increase the efficiency of this localization method, only the position of the left side of the text lines are predicted. The text recognizer is then in charge of predicting the end of the text to recognize. This method has shown good results for full page text recognition on the highly heterogeneous Maurdor dataset.

cs.CV↗

OMEGA -- OSIRIS Mapping of Emission-line Galaxies in A901/2: II. -- Environmental influence on integrated star formation properties and AGN activity

We present a study of the star formation and AGN activity for galaxies in the Abell 901/2 multi-cluster system at z~0.167 as part of the OMEGA survey. Using Tuneable Filter data obtained with the OSIRIS instrument at the GTC we produce spectra covering the Halpha and [N II] spectral lines for more than 400 galaxies. Using optical emission-line diagnostics, we identify a significant number of galaxies hosting AGN, which tend to have high masses and a broad range of morphologies. Moreover, within the environmental densities probed by our study, we find no environmental dependence on the fraction of galaxies hosting AGN. The analysis of the integrated Halpha emission shows that the specific star formation rates (SSFRs) of a majority of the cluster galaxies are below the field values for a given stellar mass. We interpret this result as evidence for a slow decrease in the star formation activity of star-forming galaxies as they fall into higher-density regions, contrary to some previous studies which suggested a rapid truncation of star formation. We find that most of the intermediate- and high-mass spiral galaxies go through a phase in which their star formation is suppressed but still retain significant star-formation activity. During this phase, these galaxies tend to retain their spiral morphology while their colours become redder. The presence of this type of galaxies in high density regions indicates that the physical mechanism responsible for suppressing star-formation affects mainly the gas component of the galaxies, suggesting that ram-pressure stripping or starvation are potentially responsible.

astro-ph.GA↗

The 2-degree Field Lensing Survey: photometric redshifts from a large new training sample to r<19.5

We present a new training set for estimating empirical photometric redshifts of galaxies, which was created as part of the 2dFLenS project. This training set is located in a 700 sq deg area of the KiDS South field and is randomly selected and nearly complete at r<19.5. We investigate the photometric redshift performance obtained with ugriz photometry from VST-ATLAS and W1/W2 from WISE, based on several empirical and template methods. The best redshift errors are obtained with kernel-density estimation, as are the lowest biases, which are consistent with zero within statistical noise. The 68th percentiles of the redshift scatter for magnitude-limited samples at r<(15.5, 17.5, 19.5) are (0.014, 0.017, 0.028). In this magnitude range, there are no known ambiguities in the colour-redshift map, consistent with a small rate of redshift outliers. In the fainter regime, the KDE method produces p(z) estimates per galaxy that represent unbiased and accurate redshift frequency expectations. The p(z) sum over any subsample is consistent with the true redshift frequency plus Poisson noise. Further improvements in redshift precision at r<20 would mostly be expected from filter sets with narrower passbands to increase the sensitivity of colours to small changes in redshift.

astro-ph.CO↗

A search for long-timescale, low-frequency radio transients

We present a search for transient and highly variable sources at low radio frequencies (150-200 MHz) that explores long timescales of 1-3 years. We conducted this search by comparing the TIFR GMRT Sky Survey Alternative Data Release 1 (TGSS ADR1) and the GaLactic and Extragalactic All-sky Murchison Widefield Array (GLEAM) survey catalogues. To account for the different completeness thresholds in the individual surveys, we searched for compact GLEAM sources above a flux density limit of 100 mJy that were not present in the TGSS ADR1; and also for compact TGSS ADR1 sources above a flux density limit of 200 mJy that had no counterpart in GLEAM. From a total sample of 234 333 GLEAM sources and 275 612 TGSS ADR1 sources in the overlap region between the two surveys, there were 99658 GLEAM sources and 38 978 TGSS ADR sources that passed our flux density cutoff and compactness criteria. Analysis of these sources resulted in three candidate transient sources. Further analysis ruled out two candidates as imaging artefacts. We analyse the third candidate and show it is likely to be real, with a flux density of 182 +/- 26 mJy at 147.5 MHz. This gives a transient surface density of rho = (6.2 +/- 6) x 10-5 deg-2 . We present initial follow-up observations and discuss possible causes for this candidate. The small number of spurious sources from this search demonstrates the high reliability of these two new low-frequency radio catalogues.

astro-ph.GA↗

2dFLenS and KiDS: Determining source redshift distributions with cross-correlations

We develop a statistical estimator to infer the redshift probability distribution of a photometric sample of galaxies from its angular cross-correlation in redshift bins with an overlapping spectroscopic sample. This estimator is a minimum variance weighted quadratic function of the data: a quadratic estimator. This extends and modifies the methodology presented by McQuinn & White (2013). The derived source redshift distribution is degenerate with the source galaxy bias, which must be constrained via additional assumptions. We apply this estimator to constrain source galaxy redshift distributions in the Kilo-Degree imaging survey through cross-correlation with the spectroscopic 2-degree Field Lensing Survey, presenting results first as a binned step-wise distribution in the range z < 0.8, and then building a continuous distribution using a Gaussian process model. We demonstrate the robustness of our methodology using mock catalogues constructed from N-body simulations, and comparisons with other techniques for inferring the redshift distribution.

astro-ph.CO↗

Learning to detect and localize many objects from few examples

The current trend in object detection and localization is to learn predictions with high capacity deep neural networks trained on a very large amount of annotated data and using a high amount of processing power. In this work, we propose a new neural model which directly predicts bounding box coordinates. The particularity of our contribution lies in the local computations of predictions with a new form of local parameter sharing which keeps the overall amount of trainable parameters low. Key components of the model are spatial 2D-LSTM recurrent layers which convey contextual information between the regions of the image. We show that this model is more powerful than the state of the art in applications where training data is not as abundant as in the classical configuration of natural images and Imagenet/Pascal VOC tasks. We particularly target the detection of text in document images, but our method is not limited to this setting. The proposed model also facilitates the detection of many objects in a single image and can deal with inputs of variable sizes without resizing.

cs.CV↗

The 2-degree Field Lensing Survey: design and clustering measurements

We present the 2-degree Field Lensing Survey (2dFLenS), a new galaxy redshift survey performed at the Anglo-Australian Telescope. 2dFLenS is the first wide-area spectroscopic survey specifically targeting the area mapped by deep-imaging gravitational lensing fields, in this case the Kilo-Degree Survey. 2dFLenS obtained 70,079 redshifts in the range z < 0.9 over an area of 731 sq deg, and is designed to extend the datasets available for testing gravitational physics and promote the development of relevant algorithms for joint imaging and spectroscopic analysis. The redshift sample consists first of 40,531 Luminous Red Galaxies (LRGs), which enable analyses of galaxy-galaxy lensing, redshift-space distortion, and the overlapping source redshift distribution by cross-correlation. An additional 28,269 redshifts form a magnitude-limited (r < 19.5) nearly-complete sub-sample, allowing direct source classification and photometric-redshift calibration. In this paper, we describe the motivation, target selection, spectroscopic observations, and clustering analysis of 2dFLenS. We use power spectrum multipole measurements to fit the redshift-space distortion parameter of the LRG sample in two redshift ranges 0.15 < z < 0.43 and 0.43 < z < 0.7 as beta = 0.49 +/- 0.15 and beta = 0.26 +/- 0.09, respectively. These values are consistent with those obtained from LRGs in the Baryon Oscillation Spectroscopic Survey. 2dFLenS data products will be released via our website http://2dflens.swin.edu.au.

astro-ph.CO↗

Learning Human Identity from Motion Patterns

We present a large-scale study exploring the capability of temporal deep neural networks to interpret natural human kinematics and introduce the first method for active biometric authentication with mobile inertial sensors. At Google, we have created a first-of-its-kind dataset of human movements, passively collected by 1500 volunteers using their smartphones daily over several months. We (1) compare several neural architectures for efficient learning of temporal multi-modal data representations, (2) propose an optimized shift-invariant dense convolutional mechanism (DCWRNN), and (3) incorporate the discriminatively-trained dynamic features in a probabilistic generative framework taking into account temporal characteristics. Our results demonstrate that human kinematics convey important information about user identity and can serve as a valuable component of multi-modal authentication systems.

cs.LG↗

Ground States and Zero-Temperature Measures at the Boundary of Rotation Sets

We consider a continuous dynamical system $f:X\to X$ on a compact metric space $X$ equipped with an $m$-dimensional continuous potential $Φ=(ϕ_1,\cdots,ϕ_m):X\to \bR^m$. We study the set of ground states $ GS(α)$ of the potential $α\cdot Φ$ as a function of the direction vector $α\in S^{m-1}$. %We also study the corresponding rotation vectors $\rv(GS(α))$. We show that the structure of the ground state sets is naturally related to the geometry of the generalized rotation set of $Φ$. In particular, for each $α$ the set of rotation vectors of $ GS(α)$ forms a non-empty, compact and connected subset of a face $F_α(Φ)$ of the rotation set associated with $α$. Moreover, every ground state maximizes entropy among all invariant measures with rotation vectors in $F_α(Φ)$. We further establish the occurrence of several quite unexpected phenomena. Namely, we construct for any $m\in\bN$ examples with an exposed boundary point (i.e. $F_α(Φ)$ being a singleton) without a unique ground state. Further, we establish the possibility of a line segment face $F_α(Φ)$ with a unique but non-ergodic ground state. Finally, we establish the possibility that the set of rotation vectors of $GS(α)$ is a non-trivial line segment.

math.DS↗

Interstellar Extinction Curve Variations Toward the Inner Milky Way: A Challenge to Observational Cosmology

We investigate interstellar extinction curve variations toward $\sim$4 deg$^{2}$ of the inner Milky Way in $VIJK_{s}$ photometry from the OGLE-III and $VVV$ surveys, with supporting evidence from diffuse interstellar bands and $F435W,F625W$ photometry. We obtain independent measurements toward $\sim$2,000 sightlines of $A_{I}$, $E(V-I)$, $E(I-J)$, and $E(J-K_{s})$, with median precision and accuracy of 2%. We find that the variations in the extinction ratios $A_{I}/E(V-I)$, $E(I-J)/E(V-I)$ and $E(J-K_{s})/E(V-I)$ are large (exceeding 20%), significant, and positively correlated, as expected. However, both the mean values and the trends in these extinction ratios are drastically shifted from the predictions of Cardelli and Fitzpatrick, regardless of how $R_{V}$ is varied. Furthermore, we demonstrate that variations in the shape of the extinction curve has at least two degrees of freedom, and not one (e.g. $R_{V}$), which we conform with a principal component analysis. We derive a median value of $ =13.44$, which is $\sim$60% higher than the "standard" value. We show that the Wesenheit magnitude $W_{I}=I-1.61(I-J)$ is relatively impervious to extinction curve variations. Given that these extinction curves are linchpins of observational cosmology, and that it is generally assumed that $R_{V}$ variations correctly capture variations in the extinction curve, we argue that systematic errors in the distance ladder from studies of type Ia supernovae and Cepheids may have been underestimated. Moreover, the reddening maps from the Planck experiment are shown to systematically overestimate dust extinction by $\sim$100%, and lack sensitivity to extinction curve variations.

astro-ph.SR↗

Linking the Structural Properties of Galaxies and their Star Formation Histories with STAGES

We study the links between star formation history and structure for a large mass-selected galaxy sample at 0.05 < z_phot < 0.30. The galaxies inhabit a very broad range of environments, from cluster cores to the field. Using HST images, we quantify their structure following Hoyos et al. (2012), and divide them into disturbed and undisturbed. We also visually identify mergers. Additionally, we provide a quantitative measure of the degree of disturbance for each galaxy ("roughness"). The majority of elliptical and lenticular galaxies have relaxed structure, showing no signs of ongoing star formation. Structurally-disturbed galaxies, which tend to avoid the lowest-density regions, have higher star-formation activity and younger stellar populations than undisturbed systems. Cluster spirals with reduced/quenched star formation have somewhat less disturbed morphologies than spirals with "normal" star-formation activity, suggesting that these "passive" spirals have started their morphological transformation into S0s. Visually identified mergers and galaxies not identified as mergers but with similar roughness have similar specific star formation rates and stellar ages. The degree of enhanced star formation is thus linked to the degree of structural disturbance, regardless of whether it is caused by major mergers or not. This suggests that merging galaxies are not special in terms of their higher-than-normal star-formation activity. Any physical process that produces "roughness", or regions of enhanced luminosity density, will increase the star-formation activity in a galaxy with similar efficiency. An alternative explanation is that star formation episodes increase the galaxies' roughness similarly, regardless of whether they are merger-induced or not.

astro-ph.GA↗

Measuring nickel masses in Type Ia supernovae using cobalt emission in nebular phase spectra

The light curves of Type Ia supernovae (SNe Ia) are powered by the radioactive decay of $^{56}$Ni to $^{56}$Co at early times, and the decay of $^{56}$Co to $^{56}$Fe from ~60 days after explosion. We examine the evolution of the [Co III] 5892 A emission complex during the nebular phase for SNe Ia with multiple nebular spectra and show that the line flux follows the square of the mass of $^{56}$Co as a function of time. This result indicates both efficient local energy deposition from positrons produced in $^{56}$Co decay, and long-term stability of the ionization state of the nebula. We compile 77 nebular spectra of 25 SN Ia from the literature and present 17 new nebular spectra of 7 SNe Ia, including SN2014J. From these we measure the flux in the [Co III] 5892 A line and remove its well-behaved time dependence to infer the initial mass of $^{56}$Ni ($M_{Ni}$) produced in the explosion. We then examine $^{56}$Ni yields for different SN Ia ejected masses ($M_{ej}$ - calculated using the relation between light curve width and ejected mass) and find the $^{56}$Ni masses of SNe Ia fall into two regimes: for narrow light curves (low stretch s~0.7-0.9), $M_{Ni}$ is clustered near $M_{Ni}$ ~ 0.4$M_\odot$ and shows a shallow increase as $M_{ej}$ increases from ~1-1.4$M_\odot$; at high stretch, $M_{ej}$ clusters at the Chandrasekhar mass (1.4$M_\odot$) while $M_{Ni}$ spans a broad range from 0.6-1.2$M_\odot$. This could constitute evidence for two distinct SN Ia explosion mechanisms.

astro-ph.HE↗

ModDrop: adaptive multi-modal gesture recognition

We present a method for gesture detection and localisation based on multi-scale and multi-modal deep learning. Each visual modality captures spatial information at a particular spatial scale (such as motion of the upper body or a hand), and the whole system operates at three temporal scales. Key to our technique is a training strategy which exploits: i) careful initialization of individual modalities; and ii) gradual fusion involving random dropping of separate channels (dubbed ModDrop) for learning cross-modality correlations while preserving uniqueness of each modality-specific representation. We present experiments on the ChaLearn 2014 Looking at People Challenge gesture recognition track, in which we placed first out of 17 teams. Fusing multiple modalities at several spatial and temporal scales leads to a significant increase in recognition rates, allowing the model to compensate for errors of the individual classifiers as well as noise in the separate channels. Futhermore, the proposed ModDrop training technique ensures robustness of the classifier to missing signals in one or several channels to produce meaningful predictions from any number of available modalities. In addition, we demonstrate the applicability of the proposed fusion scheme to modalities of arbitrary nature by experiments on the same dataset augmented with audio.

cs.CV↗

Activity recognition from videos with parallel hypergraph matching on GPUs

In this paper, we propose a method for activity recognition from videos based on sparse local features and hypergraph matching. We benefit from special properties of the temporal domain in the data to derive a sequential and fast graph matching algorithm for GPUs. Traditionally, graphs and hypergraphs are frequently used to recognize complex and often non-rigid patterns in computer vision, either through graph matching or point-set matching with graphs. Most formulations resort to the minimization of a difficult discrete energy function mixing geometric or structural terms with data attached terms involving appearance features. Traditional methods solve this minimization problem approximately, for instance with spectral techniques. In this work, instead of solving the problem approximatively, the exact solution for the optimal assignment is calculated in parallel on GPUs. The graphical structure is simplified and regularized, which allows to derive an efficient recursive minimization algorithm. The algorithm distributes subproblems over the calculation units of a GPU, which solves them in parallel, allowing the system to run faster than real-time on medium-end GPUs.

cs.CV↗

OMEGA -- OSIRIS Mapping of Emission-line Galaxies in A901/2: I.-- Survey description, data analysis, and star formation and AGN activity in the highest density regions

We present an overview of and first results from the OMEGA survey: the OSIRIS Mapping of Emission-line Galaxies in the multi-cluster system A901/2. The ultimate goal of this project is to study star formation and AGN activity across a broad range of environments at a single redshift. Using the tuneable-filter mode of the OSIRIS instrument on GTC, we target Halpha and [NII] emission lines over a ~0.5 X 0.5 deg2 region containing the z~0.167 multi-cluster system A901/2. In this paper we describe the design of the survey, the observations and the data analysis techniques developed. We then present early results from two OSIRIS pointings centred on the cores of the A901a and A902 clusters. AGN and star-forming (SF) objects are identified using the [NII]/Halpha vs. W_Halpha (WHAN) diagnostic diagram. The AGN hosts are brighter, more massive, and possess earlier-type morphologies than SF galaxies. Both populations tend to be located towards the outskirts of the high density regions we study. The typical Halpha luminosity of these sources is significantly lower than that of field galaxies at similar redshifts, but greater than that found for A1689, a rich cluster at z~0.2. The Halpha luminosities of our objects translate into star-formation rates (SFRs) between ~0.02 and 6 Msun/yr. Comparing the relationship between stellar mass and Halpha-derived SFR with that found in the field indicates a suppression of star formation in the cores of the clusters. These findings agree with previous investigations of this multi-cluster structure, based on other star formation indicators, and demonstrate the power of tuneable filters for this kind of study.

astro-ph.GA↗

The environmental dependence of the structure of galactic discs in STAGES S0 galaxies: implications for S0 formation

We present an analysis of V-band radial surface brightness μ(r) profiles for S0s in different environments using HST/ACS imaging and data from the Space Telescope A901/2 Galaxy Evolution Survey (STAGES). Using a sample of ~280 field and cluster S0s, we find that in both environments, ~25 per cent have a pure exponential disc (Type I) and ~50 per cent exhibit an up-bending disc break (antitruncation, Type III). However, we find hardly any (< 5 per cent) down-bending disc breaks (truncations, Type II) in our S0s and many cases (~20 per cent) where no exponential component was observed. We also find no evidence for an environmental dependence on the disc scalelength or break strength (outer-to-inner scalelength ratio), implying that the galaxy environment does not affect the stellar distribution in S0 stellar discs. Comparing disc structure between these S0s and the spirals from our previous studies, we find: i) no evidence for the Type I scalelength being dependent on morphology; and ii) some evidence suggesting the Type II/III break strength is smaller (weaker) in S0s compared to spirals. Taken together, these results suggest that the stellar distribution in S0s is not drastically affected by the galaxy environment. However, some process inherent to the morphological transformation of spirals into S0s does affect the stellar disc causing a weakening of μ(r) breaks and may even eliminate truncations from S0s. In further tests, we perform analytical bulge-disc decompositions on our S0s and compare the results to those for spirals from our previous studies. For Type III galaxies, we find that bulge light can account for the excess light at large radii in up to ~50 per cent of S0s but in only ~15 per cent of spirals. We propose that this result is consistent with a fading stellar disc (evolving bulge-to-disc ratio) being an inherent process in the transformation of spirals into S0s.

astro-ph.GA↗