Source author record

Vivien Bonvin

Vivien Bonvin appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

astro-ph.CO astro-ph.GA Artificial Intelligence Machine Learning astro-ph.HE astro-ph.IM

Catalog footprint

What is connected

8works

6topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

Generation and Simulation of Synthetic Datasets with Copulas

This paper proposes a new method to generate synthetic data sets based on copula models. Our goal is to produce surrogate data resembling real data in terms of marginal and joint distributions. We present a complete and reliable algorithm for generating a synthetic data set comprising numeric or categorical variables. Applying our methodology to two datasets shows better performance compared to other methods such as SMOTE and autoencoders.

preprint2022arXiv

Radial Autoencoders for Enhanced Anomaly Detection

In classification problems, supervised machine-learning methods outperform traditional algorithms, thanks to the ability of neural networks to learn complex patterns. However, in two-class classification tasks like anomaly or fraud detection, unsupervised methods could do even better, because their prediction is not limited to previously learned types of anomalies. An intuitive approach of anomaly detection can be based on the distances from the centers of mass of the two respective classes. Autoencoders, although trained without supervision, can also detect anomalies: considering the center of mass of the normal points, reconstructions have now radii, with largest radii most likely indicating anomalous points. Of course, radii-based classification were already possible without interposing an autoencoder. In any space, radial classification can be operated, to some extent. In order to outperform it, we proceed to radial deformations of data (i.e. centric compression or expansions of axes) and autoencoder training. Any autoencoder that makes use of a data center is here baptized a centric autoencoder (cAE). A special type is the cAE trained with a uniformly compressed dataset, named the centripetal autoencoder (cpAE). The new concept is studied here in relation with a schematic artificial dataset, and the derived methods show consistent score improvements. But tested on real banking data, our radial deformation supervised algorithms alone still perform better that cAEs, as expected from most supervised methods; nonetheless, in hybrid approaches, cAEs can be combined with a radial deformation of space, improving its classification score. We expect that centric autoencoders will become irreplaceable objects in anomaly live detection based on geometry, thanks to their ability to stem naturally on geometrical algorithms and to their native capability of detecting unknown anomaly types.

preprint2020arXiv

A Microlensing Accretion Disk Size Measurement in the Lensed Quasar WFI 2026-4536

We use thirteen seasons of R-band photometry from the 1.2m Leonard Euler Swiss Telescope at La Silla to examine microlensing variability in the quadruply-imaged lensed quasar WFI 2026-4536. The lightcurves exhibit ${\sim}\,0.2\,\text{mag}$ of uncorrelated variability across all epochs and a prominent single feature of ${\sim}\,0.1\,\text{mag}$ within a single season. We analyze this variability to constrain the size of the quasar's accretion disk. Adopting a nominal inclination of 60$^\text{o}$, we find an accretion disk scale radius of $\log(r_s/\text{cm}) = 15.74^{+0.34}_{-0.29}$ at a rest-frame wavelength of $2043\,\unicode{xC5}$, and we estimate a black hole mass of $\log(M_{\text{BH}}/M_{\odot}) = 9.18^{+0.39}_{-0.34}$, based on the CIV line in VLT spectra. This size measurement is fully consistent with the Quasar Accretion Disk Size - Black Hole Mass relation, providing another system in which the accretion disk is larger than predicted by thin disk theory.

preprint2020arXiv

Cosmic dissonance: new physics or systematics behind a short sound horizon?

Persistent tension between low-redshift observations and the Cosmic Microwave Background radiation (CMB), in terms of two fundamental distance scales set by the sound horizon $r_d$ and the Hubble constant $H_0$, suggests new physics beyond the Standard Model or residual systematics. We examine recently updated distance calibrations from Cepheids, gravitational lensing time-delay observations, and the Tip of the Red Giant Branch. Calibrating the Baryon Acoustic Oscillations (BAO) and Type Ia supernovae with combinations of the distance indicators, we obtain a joint and self-consistent measurement of $H_0$ and $r_d$ at low redshift, independent of cosmological models and CMB inference. In an attempt to alleviate the tension between late-time and CMB-based measurements, we consider four extensions of the standard $Λ$CDM model. The sound horizon from our different measurements is $r_d=(137\pm3^{stat.}\pm2^{syst.})$~Mpc. Depending on the adopted distance indicators, the $combined$ tension in $H_0$ and $r_d$ ranges between 2.3 and 5.1 $σ$. We find that modifications of $Λ$CDM that change the physics after recombination fail to solve the problem, for the reason that they only resolve the tension in $H_0$, while the tension in $r_d$ remains unchanged. Pre-recombination extensions (with early dark energy or the effective number of neutrinos $\rm{N}_{\rm{eff}}=3.24 \pm 0.16$) are allowed by the data, unless the calibration from Cepheids is included. Results from time-delay lenses are consistent with those from distance-ladder calibrations and point to a discrepancy between absolute distance scales measured from the CMB (assuming the standard cosmological model) and late-time observations. New proposals to resolve this tension should be examined with respect to reconciling not only the Hubble constant but also the sound horizon derived from the CMB and other cosmological probes.

preprint2019arXiv

H0LiCOW XII. Lens mass model of WFI2033-4723 and blind measurement of its time-delay distance and $H_0$

We present the lens mass model of the quadruply-imaged gravitationally lensed quasar WFI2033-4723, and perform a blind cosmographical analysis based on this system. Our analysis combines (1) time-delay measurements from 14 years of data obtained by the COSmological MOnitoring of GRAvItational Lenses (COSMOGRAIL) collaboration, (2) high-resolution $\textit{Hubble Space Telescope}$ imaging, (3) a measurement of the velocity dispersion of the lens galaxy based on ESO-MUSE data, and (4) multi-band, wide-field imaging and spectroscopy characterizing the lens environment. We account for all known sources of systematics, including the influence of nearby perturbers and complex line-of-sight structure, as well as the parametrization of the light and mass profiles of the lensing galaxy. After unblinding, we determine the effective time-delay distance to be $4784_{-248}^{+399}~\mathrm{Mpc}$, an average precision of $6.6\%$. This translates to a Hubble constant $H_{0} = 71.6_{-4.9}^{+3.8}~\mathrm{km~s^{-1}~Mpc^{-1}}$, assuming a flat $Λ$CDM cosmology with a uniform prior on $Ω_\mathrm{m}$ in the range [0.05, 0.5]. This work is part of the $H_0$ Lenses in COSMOGRAIL's Wellspring (H0LiCOW) collaboration, and the full time-delay cosmography results from a total of six strongly lensed systems are presented in a companion paper (H0LiCOW XIII).

preprint2019arXiv

H0LiCOW XIII. A 2.4% measurement of $H_{0}$ from lensed quasars: $5.3σ$ tension between early and late-Universe probes

We present a measurement of the Hubble constant ($H_{0}$) and other cosmological parameters from a joint analysis of six gravitationally lensed quasars with measured time delays. All lenses except the first are analyzed blindly with respect to the cosmological parameters. In a flat $Λ$CDM cosmology, we find $H_{0} = 73.3_{-1.8}^{+1.7}$, a 2.4% precision measurement, in agreement with local measurements of $H_{0}$ from type Ia supernovae calibrated by the distance ladder, but in $3.1σ$ tension with $Planck$ observations of the cosmic microwave background (CMB). This method is completely independent of both the supernovae and CMB analyses. A combination of time-delay cosmography and the distance ladder results is in $5.3σ$ tension with $Planck$ CMB determinations of $H_{0}$ in flat $Λ$CDM. We compute Bayes factors to verify that all lenses give statistically consistent results, showing that we are not underestimating our uncertainties and are able to control our systematics. We explore extensions to flat $Λ$CDM using constraints from time-delay cosmography alone, as well as combinations with other cosmological probes, including CMB observations from $Planck$, baryon acoustic oscillations, and type Ia supernovae. Time-delay cosmography improves the precision of the other probes, demonstrating the strong complementarity. Allowing for spatial curvature does not resolve the tension with $Planck$. Using the distance constraints from time-delay cosmography to anchor the type Ia supernova distance scale, we reduce the sensitivity of our $H_0$ inference to cosmological model assumptions. For six different cosmological models, our combined inference on $H_{0}$ ranges from $\sim73$-$78~\mathrm{km~s^{-1}~Mpc^{-1}}$, which is consistent with the local distance ladder constraints.

preprint2016arXiv

Computational issues in chemo-dynamical modelling of the formation and evolution of galaxies

Chemo-dynamical N-body simulations are an essential tool for understanding the formation and evolution of galaxies. As the number of observationally determined stellar abundances continues to climb, these simulations are able to provide new constraints on the early star formaton history and chemical evolution inside both the Milky Way and Local Group dwarf galaxies. Here, we aim to reproduce the low $α$-element scatter observed in metal-poor stars. We first demonstrate that as stellar particles inside simulations drop below a mass threshold, increases in the resolution produce an unacceptably large scatter as one particle is no longer a good approximation of an entire stellar population. This threshold occurs at around $10^3\,\rm{M_\odot}$, a mass limit easily reached in current (and future) simulations. By simulating the Sextans and Fornax dwarf spheroidal galaxies we show that this increase in scatter at high resolutions arises from stochastic supernovae explosions. In order to reduce this scatter down to the observed value, we show the necessity of introducing a metal mixing scheme into particle-based simulations. The impact of the method used to inject the metals into the surrounding gas is also discussed. We finally summarise the best approach for accurately reproducing the scatter in simulations of both Local Group dwarf galaxies and in the Milky Way.

preprint2014arXiv

Strong Lens Time Delay Challenge: II. Results of TDC1

We present the results of the first strong lens time delay challenge. The motivation, experimental design, and entry level challenge are described in a companion paper. This paper presents the main challenge, TDC1, which consisted of analyzing thousands of simulated light curves blindly. The observational properties of the light curves cover the range in quality obtained for current targeted efforts (e.g.,~COSMOGRAIL) and expected from future synoptic surveys (e.g.,~LSST), and include simulated systematic errors. \nteamsA\ teams participated in TDC1, submitting results from \nmethods\ different method variants. After a describing each method, we compute and analyze basic statistics measuring accuracy (or bias) $A$, goodness of fit $χ^2$, precision $P$, and success rate $f$. For some methods we identify outliers as an important issue. Other methods show that outliers can be controlled via visual inspection or conservative quality control. Several methods are competitive, i.e., give $|A|<0.03$, $P<0.03$, and $χ^2<1.5$, with some of the methods already reaching sub-percent accuracy. The fraction of light curves yielding a time delay measurement is typically in the range $f = $20--40\%. It depends strongly on the quality of the data: COSMOGRAIL-quality cadence and light curve lengths yield significantly higher $f$ than does sparser sampling. Taking the results of TDC1 at face value, we estimate that LSST should provide around 400 robust time-delay measurements, each with $P<0.03$ and $|A|<0.01$, comparable to current lens modeling uncertainties. In terms of observing strategies, we find that $A$ and $f$ depend mostly on season length, while P depends mostly on cadence and campaign duration.