Search

We present deep total intensity and polarization observations of the Coma cluster at 1.4 and 6.6 GHz performed with the Sardinia Radio Telescope. By combining the single-dish 1.4 GHz data with archival Very Large Array observations we obtain new images of the central radio halo and of the peripheral radio relic where we properly recover the brightness from the large scale structures. At 6.6 GHz we detect both the relic and the central part of the halo in total intensity and polarization. These are the highest frequency images available to date for these radio sources in this galaxy cluster. In the halo, we find a localized spot of polarized signal, with fractional polarization of about 45%. The polarized emission possibly extends along the north-east side of the diffuse emission. The relic is highly polarized, up to 55%, as usually found for these sources. We confirm the halo spectrum is curved, in agreement with previous single-dish results. The spectral index is alpha=1.48 +/- 0.07 at a reference frequency of 1 GHz and varies from alpha ~1.1, at 0.1 GHz, up to alpha ~ 1.8, at 10 GHz. We compare the Coma radio halo surface brightness profile at 1.4 GHz (central brightness and e-folding radius) with the same properties of the other halos, and we find that it has one of the lowest emissivities observed so far. Reanalyzing the relic's spectrum in the light of the new data, we obtain a refined radio Mach number of M=2.9 +/- 0.1., Comment: 27 pages, 18 figures, this article has been accepted for publication in MNRAS Published by Oxford University Press on behalf of the Royal Astronomical Society

One of the most important objectives of solar physics is the physical understanding of the solar atmosphere, the structure of which is also described in terms of the density (N) and temperature (T) distributions of the atmospheric matter. Several multi-frequency analyses show that the characteristics of these distributions are still debated, especially for the outer coronal emission. We aim to constrain the T and N distributions of the solar atmosphere through observations in the centimetric radio domain. We employ single-dish observations from two of the INAF radio telescopes at the K-band frequencies (18 - 26 GHz). We investigate the origin of the significant brightness temperature ($T_B$) level that we detected up to the upper corona ($\sim 800$ Mm of altitude with respect to the photospheric solar surface). To probe the physical origin of the atmospheric emission and to constrain instrumental biases, we reproduced the solar signal by convolving specific 2D antenna beam models. The analysis of the solar atmosphere is performed by adopting a physical model that assumes the thermal bremsstrahlung as the emission mechanism, with specific T and N distributions. The modelled $T_B$ profiles are compared with those observed by averaging solar maps obtained during the minimum of solar activity (2018 - 2020). The T and N distributions are compatible (within $25\%$ of uncertainty) with the model up to $\sim 60$ Mm and $\sim 100$ Mm of altitude, respectively. The analysis of the role of the antenna beam pattern on our solar maps proves the physical nature of the atmospheric emission in our images up to the coronal tails seen in our $T_B$ profiles. The challenging analysis of the coronal radio emission at higher altitudes, together with the data from satellite instruments will require further multi-frequency measurements., Comment: 12 pages, 8 figures, 5 tables, accepted by A&A; v1

Large Language Models (LLMs) have demonstrated remarkable proficiency in understanding and generating natural language. However, their capabilities wane in highly specialized domains underrepresented in the pretraining corpus, such as physical and biomedical sciences. This work explores how to repurpose general LLMs into effective task solvers for specialized domains. We introduce a novel, model-agnostic framework for learning custom input tags, which are parameterized as continuous vectors appended to the LLM's embedding layer, to condition the LLM. We design two types of input tags: domain tags are used to delimit specialized representations (e.g., chemical formulas) and provide domain-relevant context; function tags are used to represent specific functions (e.g., predicting molecular properties) and compress function-solving instructions. We develop a three-stage protocol to learn these tags using auxiliary data and domain knowledge. By explicitly disentangling task domains from task functions, our method enables zero-shot generalization to unseen problems through diverse combinations of the input tags. It also boosts LLM's performance in various specialized domains, such as predicting protein or chemical properties and modeling drug-target interactions, outperforming expert models tailored to these tasks., Comment: ICML 2024

The Sun is an extraordinary workbench, from which several fundamental astronomical parameters can be measured with high precision. Among these parameters, the solar radius $R_{\odot}$ plays an important role in several aspects, such as in evolutionary models. Despite the efforts in obtaining accurate measurements of $R_{\odot}$, the subject is still debated and measurements are puzzling and/or lacking in many frequency ranges. We aimed to determine the mean, equatorial, and polar radii of the Sun ($R_c$, $R_{eq}$, and $R_{pol}$) in the frequency range 18.1 - 26.1 GHz. We employed single-dish observations from the newly-appointed Medicina "Gavril Grueff" Radio Telescope and the Sardinia Radio Telescope (SRT) throughout 5 years, from 2018 to mid-2023, in the framework of the SunDish project for solar monitoring. Two methods to calculate the radius at radio frequencies are considered and compared. To assess the quality of our radius determinations, we also analysed the possible degrading effects of the antenna beam pattern on our solar maps, using two 2D-models. We carried out a correlation analysis with the evolution of the solar cycle through the calculation of Pearson's correlation coefficient $\rho$. We obtained several values for the solar radius - ranging between 959 and 994 arcsec - and $\rho$, with typical errors of a few arcsec. Our $R_{\odot}$ measurements, consistent with values reported in literature, suggest a weak prolatness of the solar limb ($R_{eq}$ > $R_{pol}$), although $R_{eq}$ and $R_{pol}$ are statistically compatible within 3$\sigma$ errors. The correlation analysis using the solar images from Grueff shows (1) a positive correlation between the solar activity and the temporal variation of $R_c$ (and $R_{eq}$) at all observing frequencies, and (2) a weak anti-correlation between the temporal variation of $R_{pol}$ and the solar activity at 25.8 GHz., Comment: 18 pages, 12 figures, 6 tables, accepted by A&A; v1

Experimental setups for charge transport measurements are typically not compatible with the ultra-high vacuum conditions for chemical doping, limiting the charge carrier density that can be investigated by transport methods. Field-effect methods, including dielectric gating and ionic liquid gating, achieve too low a carrier density to induce electronic phase transitions. To bridge this gap, we developed an integrated flip-chip method to dope graphene by alkali vapour in the diffusive regime, suitable for charge transport measurements at ultra-high charge carrier density. We introduce a cesium droplet into a sealed cavity filled with inert gas to dope a monolayer graphene sample by the process of cesium atom diffusion, adsorption and ionization at the graphene surface, with doping beyond an electron density of $4.7\times10^{14}~\mathrm{cm}^{-2}$ monitored by operando Hall measurement. The sealed assembly is stable against oxidation, enabling measurement of charge transport versus temperature and magnetic field. Cyclotron mass inversion is observed via the Hall effect, indicative of the change of Fermi surface geometry associated with the Liftshitz transition at the hyperbolic $M$ point of monolayer graphene. The transparent quartz substrate also functions as an optical window, enabling non-resonant Raman scattering. Our findings show that chemical doping, hitherto restricted to ultra-high vacuum, can be applied in a diffusive regime at ambient pressure in an inert gas environment and thus enable charge transport studies in standard cryogenic environments

We present a lattice determination of the hadronic susceptibilities that, thanks to unitarity and analyticity, constrain the form factors entering the semileptonic $b\rightarrow c$ transitions. We evaluate the transverse and longitudinal susceptibilities of the vector and axial polarization functions at zero momentum transfer from the moments of appropriate two-point correlation functions. The latter are obtained on the lattice employing gauge ensembles of the Extended Twisted Mass Collaboration (ETMC) with $N_f$=2+1+1 flavors of Wilson-clover twisted-mass quarks with masses of all the dynamical quark flavors tuned close to their physical values. The simulations are carried out at four values of the lattice spacing, $a \simeq (0.057,0.068, 0.080, 0.091)$ fm, with spatial lattice sizes up to $L \simeq 7.6$ fm. Heavy-quark masses up to $\approx 3.5$ times the physical charm mass are employed, allowing for a smooth extrapolation to the physical b-quark mass., Comment: 7 pages, 4 figures, proceeding of Lattice 2023

We present high-spectral resolution M-band spectra from iSHELL on NASA's Infrared Telescope Facility (IRTF) along the line of sight to the debris disk host star HD 32297. We also present a Gemini Planet Imager (GPI) H-band polarimetric image of the HD 131488 debris disk. We search for fundamental CO absorption lines in the iSHELL spectra of HD 32297 but do not detect any. We place an upper limit on the CO column density of $\sim$6$\times10^{15}$ cm$^{-2}$. By combining the column density upper limit, the CO mass measured with ALMA, and the geometrical properties of the disk, we estimate the scale height of the CO to be $\lesssim$ 2 au across the radial extent of the disk ($\sim$80-120 au). We use the same method to estimate the CO scale height of three other edge-on, CO-rich debris disks that all have CO observed in absorption with HST as well as in emission with ALMA: $\beta$ Pictoris, HD 110058, and HD 131488. We compare our estimated CO scale heights of these four systems to the millimeter dust scale heights and find that, under the assumption of hydrostatic equilibrium, there is a potential correlation between the CO and millimeter dust scale heights. There are multiple factors that affect the gas vertical structure such as turbulence, photodissociation with weak vertical mixing, as well as where the gas originates. One possible explanation for the potential correlation could be that the gas and dust are of a similar secondary origin in these four systems., Comment: Accepted for publication in The Astrophysical Journal. 11 figures, 19 pages

We present a comprehensive overview of a volume-complete sample of white dwarfs located within 40 pc of the Sun, a significant proportion of which were detected in Gaia Data Release 3 (DR3). Our DR3 sample contains 1076 spectroscopically confirmed white dwarfs, with just five candidates within the volume remaining unconfirmed (more than 99 per cent spectroscopic completeness). Additionally, 28 white dwarfs were not in our initial selection from Gaia DR3, most of which are in unresolved binaries. We use Gaia DR3 photometry and astrometry to determine a uniform set of white dwarf parameters, including mass, effective temperature, and cooling age. We assess the demographics of the 40 pc sample, specifically magnetic fields, binarity, space density and mass distributions., Comment: Accepted for publication in MNRAS. 24 pages, 25 figures

Observations of planetary material polluting the atmospheres of white dwarfs are an important probe of the bulk composition of exoplanetary material. Medium- and high-resolution optical and ultraviolet spectroscopy of seven white dwarfs with known circumstellar dust and gas emission are presented. Detections or meaningful upper limits for photospheric absorption lines are measured for: C, O, Na, S, P, Mg, Al, Si, Ca, Ti, Cr, Fe, and Ni. For 16 white dwarfs with known observable gaseous emission discs (and measured photospheric abundances), there is no evidence that their accretion rates differ, on average, from those without detectable gaseous emission. This suggests that, typically, accretion is not enhanced by gas drag. At the effective temperature range of the white dwarfs in this sample (16,000-25,000K) the abundance ratios of elements are more consistent than absolute abundances when comparing abundances derived from spectroscopic white dwarf parameters versus photometric white dwarf parameters. Crucially, this highlights that the uncertainties on white dwarf parameters do not prevent white dwarfs from being utilised to study planetary composition. The abundances of oxygen and silicon for the three hydrogen-dominated white dwarfs in the sample with both optical and ultraviolet spectra differ by 0.62 dex depending on if they are derived from the optical or ultraviolet spectra. This optical/ultraviolet discrepancy may be related to differences in the atmospheric depth of line formation; further investigations into the white dwarf atmospheric modelling are needed to understand this discrepancy., Comment: Accepted for publication in MNRAS

We present an analysis of spectroscopic data of the cool, highly magnetic and polluted white dwarf 2MASS J0916-4215. The atmosphere of the white dwarf is dominated by hydrogen, but numerous spectral lines of magnesium, calcium, titanium, chromium, iron, strontium, along with Li I, Na I, Al I, and K I lines, are found in the incomplete Paschen-Back regime, most visibly, in the case of Ca II lines. Extensive new calculations of the Paschen-Back effect in several spectral lines are presented and results of the calculations are tabulated for the Ca II H&K doublet. The abundance pattern shows a large lithium and strontium excess, which may be viewed as a signature of planetary debris akin to Earth's continental crust accreted onto the star, although the scarcity of silicon indicates possible dilution in bulk Earth material. Accurate abundance measurements proved sensitive to the value of the broadening parameter due to collisions with neutral hydrogen (Gamma_HI), particularly in saturated lines such as the resonance lines of Ca I and Ca II. We found that Gamma_HI if formulated with values from the literature could be overestimated by a factor of 10 in most resonance lines., Comment: Accepted for publication in MNRAS

PKS 1830$-$211 is a $\gamma$-ray emitting, high-redshift (z $= 2.507 \pm 0.002$), lensed flat-spectrum radio quasar. During the period mid-February to mid-April 2019, this source underwent a series of strong $\gamma$-ray flares that were detected by both AGILE-GRID and Fermi-LAT, reaching a maximum $\gamma$-ray flux of $F_{\rm E>100 MeV}\approx 2.3\times10^{-5}$ ph cm$^{-2}$ s$^{-1}$. Here we report on a coordinated campaign from both on-ground (Medicina, OVRO, REM, SRT) and orbiting facilities (AGILE, Fermi, INTEGRAL, NuSTAR, Swift, Chandra), with the aim of investigating the multi-wavelength properties of PKS 1830$-$211 through nearly simultaneous observations presented here for the first time. We find a possible break in the radio spectra in different epochs above 15 GHz, and a clear maximum of the 15 GHz data approximately 110 days after the $\gamma$-ray main activity periods. The spectral energy distribution shows a very pronounced Compton dominance (> 200) which challenges the canonical one-component emission model. Therefore we propose that the cooled electrons of the first component are re-accelerated to a second component by, e.g., kink or tearing instability during the $\gamma$-ray flaring periods. We also note that PKS 1830$-$211 could be a promising candidate for future observations with both Compton satellites (e.g., e-ASTROGAM) and Cherenkov arrays (CTAO) which will help, thanks to their improved sensitivity, in extending the data availability in energy bands currently uncovered., Comment: Accepted for publication in MNRAS. 16 pages, 18 figures, 12 tables

Successfully handling context is essential for any dialog understanding task. This context maybe be conversational (relying on previous user queries or system responses), visual (relying on what the user sees, for example, on their screen), or background (based on signals such as a ringing alarm or playing music). In this work, we present an overview of MARRS, or Multimodal Reference Resolution System, an on-device framework within a Natural Language Understanding system, responsible for handling conversational, visual and background context. In particular, we present different machine learning models to enable handing contextual queries; specifically, one to enable reference resolution, and one to handle context via query rewriting. We also describe how these models complement each other to form a unified, coherent, lightweight system that can understand context while preserving user privacy., Comment: Sixth Workshop on Computational Models of Reference, Anaphora and Coreference (CRAC 2023)

The nearby radio galaxy M87 offers a unique opportunity to explore the connections between the central supermassive black hole and relativistic jets. Previous studies of the inner region of M87 revealed a wide opening angle for the jet originating near the black hole. The Event Horizon Telescope resolved the central radio source and found an asymmetric ring structure consistent with expectations from General Relativity. With a baseline of 17 years of observations, there was a shift in the jet's transverse position, possibly arising from an eight to ten-year quasi-periodicity. However, the origin of this sideways shift remains unclear. Here we report an analysis of radio observations over 22 years that suggests a period of about 11 years in the position angle variation of the jet. We infer that we are seeing a spinning black hole that induces the Lense-Thirring precession of a misaligned accretion disk. Similar jet precession may commonly occur in other active galactic nuclei but has been challenging to detect owing to the small magnitude and long period of the variation., Comment: 41 pages, 7 figures, 7 tables

We present ReMasker, a new method of imputing missing values in tabular data by extending the masked autoencoding framework. Compared with prior work, ReMasker is both simple -- besides the missing values (i.e., naturally masked), we randomly ``re-mask'' another set of values, optimize the autoencoder by reconstructing this re-masked set, and apply the trained model to predict the missing values; and effective -- with extensive evaluation on benchmark datasets, we show that ReMasker performs on par with or outperforms state-of-the-art methods in terms of both imputation fidelity and utility under various missingness settings, while its performance advantage often increases with the ratio of missing data. We further explore theoretical justification for its effectiveness, showing that ReMasker tends to learn missingness-invariant representations of tabular data. Our findings indicate that masked modeling represents a promising direction for further research on tabular data imputation. The code is publicly available.

There is a wealth of evidence to suggest that planetary systems can survive beyond the main sequence. Most commonly, white dwarfs are found to be accreting material from tidally disrupted asteroids, whose bulk compositions are reflected by the metals polluting the stellar photospheres. While many examples are known, most lack the deep, high-resolution data required to detect multiple elements, and thus characterise the planetesimals that orbit them. Here, spectra of seven DZ white dwarfs observed with Keck HIRES are analysed, where up to nine metals are measured per star. Their compositions are compared against those of solar system objects, working in a Bayesian framework to infer or marginalise over the accretion history. All of the stars have been accreting primitive material, similar to chondrites, with hints of a Mercury-like composition at one star. The most polluted star is observed several Myr after its last major accretion episode, in which a Moon-sized object met its demise., Comment: Accepted to MNRAS. 18 pages, 9 figures

We present a generalization of AEMD approach, routinely applied to estimate thermal conductivity, to the more general case in which Soret and Dufour effects determine a coupled heat-mass transfer. We show that, by starting from microscopical definitions of heat and mass currents, conservation laws dictates the form of the differential equations governing the time evolution. In particular, we focus to the well specific case in which a closed-form solution of the system is possible and derive the analytical form of time-evolution of temperature and concentration scalar fields in the case in which step-like initial profiles are imposed across a rectangular simulation cell. The validity of this new generalized expression is finally validated using as benchamrk system a two-component Lennard-Jones liquid system, for which generalized diffusivities are estimated in different reduced temperature and density region of phase diagram.

Two recently discovered white dwarfs, WDJ041246.84$+$754942.26 and WDJ165335.21$-$100116.33, exhibit H$\alpha$ and H$\beta$ Balmer line emission similar to stars in the emerging DAHe class, yet intriguingly have not been found to have detectable magnetic fields. These white dwarfs are assigned the spectral type DAe. We present detailed follow-up of the two known DAe stars using new time-domain spectroscopic observations and analysis of the latest photometric time-series data from TESS and ZTF. We measure the upper magnetic field strength limit of both stars as $B < 0.05$ MG. The DAe white dwarfs exhibit photometric and spectroscopic variability, where in the case of WDJ041246.84$+$754942.26 the strength of the H$\alpha$ and H$\beta$ emission cores varies in anti-phase with its photometric variability over the spin period, which is the same phase relationship seen in DAHe stars. The DAe white dwarfs closely cluster in one region of the Gaia Hertzsprung-Russell diagram together with the DAHe stars. We discuss current theories on non-magnetic and magnetic mechanisms which could explain the characteristics observed in DAe white dwarfs, but additional data are required to unambiguously determine the origin of these stars., Comment: 20 pages, 16 figures. Published by MNRAS

We present the discovery and validation of two TESS exoplanets orbiting nearby M dwarfs: TOI-2084b, and TOI-4184b. We characterized the host stars by combining spectra from Shane/Kast and Magellan/FIRE, SED (Spectral Energy Distribution) analysis, and stellar evolutionary models. In addition, we used Gemini-South/Zorro & -North/Alopeke high-resolution imaging, archival science images, and statistical validation packages to support the planetary interpretation. We performed a global analysis of multi-colour photometric data from TESS and ground-based facilities in order to derive the stellar and planetary physical parameters for each system. We find that TOI-2084b and TOI-4184b are sub-Neptune-sized planets with radii of Rp = 2.47 +/- 0.13R_Earth and Rp = 2.43 +/- 0.21R_Earth, respectively. TOI-2084b completes an orbit around its host star every 6.08 days, has an equilibrium temperature of T_eq = 527 +/- 8K and an irradiation of S_p = 12.8 +/- 0.8 S_Earth. Its host star is a dwarf of spectral M2.0 +/- 0.5 at a distance of 114pc with an effective temperature of T_eff = 3550 +/- 50 K, and has a wide, co-moving M8 companion at a projected separation of 1400 au. TOI-4184b orbits around an M5.0 +/- 0.5 type dwarf star (Kmag = 11.87) each 4.9 days, and has an equilibrium temperature of T_eq = 412 +/- 8 K and an irradiation of S_p = 4.8 +/- 0.4 S_Earth. TOI-4184 is a metal poor star ([Fe/H] = -0.27 +/- 0.09 dex) at a distance of 69 pc with an effective temperature of T_eff = 3225 +/- 75 K. Both planets are located at the edge of the sub-Jovian desert in the radius-period plane. The combination of the small size and the large infrared brightness of their host stars make these new planets promising targets for future atmospheric exploration with JWST., Comment: Accepted for publication in A&A

We present the first large sample of scintillation arcs in millisecond pulsars, analysing 12 sources observed with the Large European Array for Pulsars (LEAP), and the Effelsberg 100\,m telescope. We estimate the delays from multipath propagation, measuring significant correlated changes in scattering timescales over a 10-year timespan. Many sources show compact concentrations of power in the secondary spectrum, which in PSRs J0613$-$0200 and J1600$-$3053 can be tracked between observations, and are consistent with compact scattering at fixed angular positions. Other sources such as PSRs J1643$-$1224 and J0621+1002 show diffuse, asymmetric arcs which are likely related to phase-gradients across the scattering screen. PSR B1937+21 shows at least three distinct screens which dominate at different times and evidence of varying screen axes or multi-screen interactions. We model annual and orbital arc curvature variations in PSR J0613$-$0200, providing a measurement of the longitude of ascending node, resolving the sense of the orbital inclination, where our best fit model is of a screen with variable axis of anisotropy over time, corresponding to changes in the scattering of the source. Unmodeled variations of the screen's axis of anisotropy are likely to be a limiting factor in determining orbital parameters with scintillation, requiring careful consideration of variable screen properties, or independent VLBI measurements. Long-term scintillation studies such as this serve as a complementary tool to pulsar timing, to measure a source of correlated noise for pulsar timing arrays, solve pulsar orbits, and to understand the astrophysical origin of scattering screens., Comment: Submitted to MNRAS, first revision. 18 pages, 16 figures

The fine structure of solar prominences is made by thin threads that outline the magnetic field lines. Observations show that transverse waves of Alfv\'enic nature are ubiquitous in prominence threads. These waves are driven at the photosphere and propagate to prominences suspended in the corona. Heating due to Alfv\'en wave dissipation could be a relevant mechanism in the cool and partially ionized prominence plasma. We explore the construction of 1D equilibrium models of prominence thin threads that satisfy energy balance between radiative losses, thermal conduction, and Alfv\'en wave heating. We assume the presence of a broadband driver at the photosphere that launches Alfv\'en waves towards the prominence. An iterative method is implemented, in which the energy balance equation and the Alfv\'en wave equation are consecutively solved. From the energy balance equation and considering no wave heating initially, we compute the equilibrium profiles along the thread of the temperature, density, ionisation fraction. We use the Alfv\'en wave equation to compute the wave heating rate, which is then put back in the energy balance equation to obtain new equilibrium profiles. The process is repeated until convergence to a self-consistent thread model heated by Alfv\'en waves is achieved. We have obtained equilibrium models composed of a cold and dense thread, a extremely thin PCTR, and an extended coronal region. The length of the cold thread decreases with the temperature at the prominence core and increases with the Alfv\'en wave energy flux. Equilibrium models are not possible for sufficiently large wave energy fluxes when the wave heating rate inside the cold thread becomes larger than radiative losses. The maximum value of the wave energy flux that allows an equilibrium depends on the prominence core temperature. This constrains the existence of equilibria in realistic conditions., Comment: 12 pages, 11 figures

We present optical and infrared imaging and spectroscopy of the R Coronae Borealis-type (R Cor Bor) star IRAS 00450+7401. Optical spectra further confirm its classification as a cool R Cor Bor system, having a hydrogen-deficient carbon star spectral sub-class of HdC5 or later. Mid-infrared spectroscopy reveals the typical ~8 um ``hump'' seen in other R Cor Bor stars and no other features. A modern-epoch spectral energy distribution shows bright emission from hot dust having Tdust>600 K. Historical infrared data reveal generally cooler dust color temperatures combined with long-term fading trends, but provide no discernible correlation between flux level and temperature. Investigating the most mid-infrared variable R Cor Bor stars found in IRAS, AKARI, and WISE data reveals similar fading trends, bursts that can show a factor of up to 10 change in flux density between epochs, and blackbody-fit dust color temperatures that span 400-1300 K. While some R Cor Bor stars such as IRAS 00450+7401 appear to undergo fade/burst cycles in the mid-infrared, significant gaps in temporal coverage prevent conclusively identifying any preferred timescale for their mid-infrared variability and circumstellar dust temperature changes., Comment: AJ accepted, 15 pages, 6 figures, 5 tables, and an appendix

Data for pretraining machine learning models often consists of collections of heterogeneous datasets. Although training on their union is reasonable in agnostic settings, it might be suboptimal when the target domain -- where the model will ultimately be used -- is known in advance. In that case, one would ideally pretrain only on the dataset(s) most similar to the target one. Instead of limiting this choice to those datasets already present in the pretraining collection, here we explore extending this search to all datasets that can be synthesized as `combinations' of them. We define such combinations as multi-dataset interpolations, formalized through the notion of generalized geodesics from optimal transport (OT) theory. We compute these geodesics using a recent notion of distance between labeled datasets, and derive alternative interpolation schemes based on it: using either barycentric projections or optimal transport maps, the latter computed using recent neural OT methods. These methods are scalable, efficient, and -- notably -- can be used to interpolate even between datasets with distinct and unrelated label sets. Through various experiments in transfer learning in computer vision, we demonstrate this is a promising new approach for targeted on-demand dataset synthesis.

This paper studies federated linear contextual bandits under the notion of user-level differential privacy (DP). We first introduce a unified federated bandits framework that can accommodate various definitions of DP in the sequential decision-making setting. We then formally introduce user-level central DP (CDP) and local DP (LDP) in the federated bandits framework, and investigate the fundamental trade-offs between the learning regrets and the corresponding DP guarantees in a federated linear contextual bandits model. For CDP, we propose a federated algorithm termed as $\texttt{ROBIN}$ and show that it is near-optimal in terms of the number of clients $M$ and the privacy budget $\varepsilon$ by deriving nearly-matching upper and lower regret bounds when user-level DP is satisfied. For LDP, we obtain several lower bounds, indicating that learning under user-level $(\varepsilon,\delta)$-LDP must suffer a regret blow-up factor at least $\min\{1/\varepsilon,M\}$ or $\min\{1/\sqrt{\varepsilon},\sqrt{M}\}$ under different conditions., Comment: Accepted by ICML 2023

Privacy-Preserving machine learning (PPML) can help us train and deploy models that utilize private information. In particular, on-device machine learning allows us to avoid sharing raw data with a third-party server during inference. On-device models are typically less accurate when compared to their server counterparts due to the fact that (1) they typically only rely on a small set of on-device features and (2) they need to be small enough to run efficiently on end-user devices. Split Learning (SL) is a promising approach that can overcome these limitations. In SL, a large machine learning model is divided into two parts, with the bigger part residing on the server side and a smaller part executing on-device, aiming to incorporate the private features. However, end-to-end training of such models requires exchanging gradients at the cut layer, which might encode private features or labels. In this paper, we provide insights into potential privacy risks associated with SL. Furthermore, we also investigate the effectiveness of various mitigation strategies. Our results indicate that the gradients significantly improve the attackers' effectiveness in all tested datasets reaching almost perfect reconstruction accuracy for some features. However, a small amount of differential privacy (DP) can effectively mitigate this risk without causing significant training degradation., Comment: 10 pages

We report the first coexistence experiment of 1550 nm single-mode squeezed states of light with a 1310 nm classical telecom channel over a 10 km fiber channel while measuring squeezing using a locally generated local oscillator. This is achieved using real-time optical heterodyne phase locking, allowing us to measure up to 0.5 dB of squeezing with a phase noise of 2.2 degrees., Comment: This paper has been submitted to ECOC 2023

Counterfactual explanations are increasingly used as an Explainable Artificial Intelligence (XAI) technique to provide stakeholders of complex machine learning algorithms with explanations for data-driven decisions. The popularity of counterfactual explanations resulted in a boom in the algorithms generating them. However, not every algorithm creates uniform explanations for the same instance. Even though in some contexts multiple possible explanations are beneficial, there are circumstances where diversity amongst counterfactual explanations results in a potential disagreement problem among stakeholders. Ethical issues arise when for example, malicious agents use this diversity to fairwash an unfair machine learning model by hiding sensitive features. As legislators worldwide tend to start including the right to explanations for data-driven, high-stakes decisions in their policies, these ethical issues should be understood and addressed. Our literature review on the disagreement problem in XAI reveals that this problem has never been empirically assessed for counterfactual explanations. Therefore, in this work, we conduct a large-scale empirical analysis, on 40 datasets, using 12 explanation-generating methods, for two black-box models, yielding over 192.0000 explanations. Our study finds alarmingly high disagreement levels between the methods tested. A malicious user is able to both exclude and include desired features when multiple counterfactual explanations are available. This disagreement seems to be driven mainly by the dataset characteristics and the type of counterfactual algorithm. XAI centers on the transparency of algorithmic decision-making, but our analysis advocates for transparency about this self-proclaimed transparency

Multistate models (MSM) are well developed for continuous and discrete times under a first order Markov assumption. Motivated by a cohort of COVID-19 patients, an MSM was designed based on 14 transitions among 7 states of a patient. Since a preliminary analysis showed that the first order Markov condition was not met for some transitions, we have developed a second order Markov model where the future evolution not only depends on the current but also on the preceding state. Under a discrete time analysis, assuming homogeneity and that past information is restricted to 2 consecutive times, we expanded the transition probability matrix and proposed an extension of the Chapman- Kolmogorov equations.

We present here the East Asia to Italy Nearly Global VLBI (EATING VLBI) project. How this project started and the evolution of the international collaboration between Korean, Japanese, and Italian researchers to study compact sources with VLBI observations is reported. Problems related to the synchronization of the very different arrays and technical details of the telescopes involved are presented and discussed. The relatively high observation frequency (22 and 43 GHz) and the long baselines between Italy and East Asia produced high-resolution images. We present example images to demonstrate the typical performance of the EATING VLBI array. The results attracted international researchers and the collaboration is growing, now including Chinese and Russian stations. New in progress projects are discussed and future possibilities with a larger number of telescopes and a better frequency coverage are briefly discussed herein., Comment: 15 pages and 6 figures. This article belongs to the Special Issue Challenges in Understanding Black Hole Powered Jets with VLBI

It is tempting to interpret the minuscule scale of neutrino masses as a symptom of its radiative origin. In light of the notable leap in precision expected at the Future Circular Collider, we explore areas of the parameter space that can simultaneously support the detectable Higgs-strahlung signal with parallel ones from forthcoming measurements in low-energy observables. We pinpoint the role that the extra fermions have in shaping a signal distinct from the pure Inert Doublet one. The details of the full one-loop computation and on-shell renormalization are presented. Both normal and inverted hierarchies for the radiatively generated neutrino masses and angles are investigated., Comment: 17 pages, 16 figures, 8 tables

Histopathology is critical for the diagnosis of many diseases, including cancer. These protocols typically require pathologists to manually evaluate slides under a microscope, which is time-consuming and subjective, leading to interest in machine learning to automate analysis. However, computational techniques are limited by batch effects, where technical factors like differences in preparation protocol or scanners can alter the appearance of slides, causing models trained on one institution to fail when generalizing to others. Here, we propose a domain adaptation method that improves the generalization of histopathological models to data from unseen institutions, without the need for labels or retraining in these new settings. Our approach introduces an optimal transport (OT) loss, that extends adversarial methods that penalize models if images from different institutions can be distinguished in their representation space. Unlike previous methods, which operate on single samples, our loss accounts for distributional differences between batches of images. We show that on the Camelyon17 dataset, while both methods can adapt to global differences in color distribution, only our OT loss can reliably classify a cancer phenotype unseen during training. Together, our results suggest that OT improves generalization on rare but critical phenotypes that may only make up a small fraction of the total tiles and variation in a slide.

We present observations and analyses of eight white dwarf stars that have accreted rocky material from their surrounding planetary systems. The spectra of these helium-atmosphere white dwarfs contain detectable optical lines of all four major rock-forming elements (O, Mg, Si, Fe). This work increases the sample of oxygen-bearing white dwarfs with parent body composition analyses by roughly thirty-three percent. To first order, the parent bodies that have been accreted by the eight white dwarfs are similar to those of chondritic meteorites in relative elemental abundances and oxidation states. Seventy-five percent of the white dwarfs in this study have observed oxygen excesses implying volatiles in the parent bodies with abundances similar to those of chondritic meteorites. Three white dwarfs have oxidation states that imply more reduced material than found in CI chondrites, indicating the possible detection of Mercury-like parent bodies, but are less constrained. These results contribute to the recurring conclusion that extrasolar rocky bodies closely resemble those in our solar system, and do not, as a whole, yield unusual or unique compositions., Comment: Accepted for publication in ApJ. 7 Figures, 7 Tables

Blind cleaning methods are currently the preferred strategy for handling foreground contamination in single-dish HI intensity mapping surveys. Despite the increasing sophistication of blind techniques, some signal loss will be inevitable across all scales. Constructing a corrective transfer function using mock signal injection into the contaminated data has been a practice relied on for HI intensity mapping experiments. However, assessing whether this approach is viable for future intensity mapping surveys where precision cosmology is the aim, remains unexplored. In this work, using simulations, we validate for the first time the use of a foreground transfer function to reconstruct power spectra of foreground-cleaned low-redshift intensity maps and look to expose any limitations. We reveal that even when aggressive foreground cleaning is required, which causes ${>}\,50\%$ negative bias on the largest scales, the power spectrum can be reconstructed using a transfer function to within sub-percent accuracy. We specifically outline the recipe for constructing an unbiased transfer function, highlighting the pitfalls if one deviates from this recipe, and also correctly identify how a transfer function should be applied in an auto-correlation power spectrum. We validate a method that utilises the transfer function variance for error estimation in foreground-cleaned power spectra. Finally, we demonstrate how incorrect fiducial parameter assumptions (up to ${\pm}100\%$ bias) in the generation of mocks, used in the construction of the transfer function, do not significantly bias signal reconstruction or parameter inference (inducing ${<}\,5\%$ bias in recovered values)., Comment: 25 pages, 20 figures. See Figure 4 for the main demonstration of the transfer function's performance for reconstructing signal loss from foreground cleaning. Accepted for publication in MNRAS

In the era of precision measurements of the neutrino oscillation parameters, upcoming neutrino experiments will also be sensitive to physics beyond the Standard Model. KM3NeT/ORCA is a neutrino detector optimised for measuring atmospheric neutrinos from a few GeV to around 100 GeV. In this paper, the sensitivity of the KM3NeT/ORCA detector to neutrino decay has been explored. A three-flavour neutrino oscillation scenario, where the third neutrino mass state $\nu_3$ decays into an invisible state, e.g. a sterile neutrino, is considered. We find that KM3NeT/ORCA would be sensitive to invisible neutrino decays with $1/\alpha_3=\tau_3/m_3 < 180$~$\mathrm{ps/eV}$ at $90\%$ confidence level, assuming true normal ordering. Finally, the impact of neutrino decay on the precision of KM3NeT/ORCA measurements for $\theta_{23}$, $\Delta m^2_{31}$ and mass ordering have been studied. No significant effect of neutrino decay on the sensitivity to these measurements has been found., Comment: 27 pages, 14 figures, bibliography updated, typos corrected

We present and compare several methods to mitigate time-correlated (1/f) noise within the HI intensity mapping component of the MeerKAT Large Area Synoptic Survey (MeerKLASS). By simulating scan strategies, the HI signal, foreground emissions, white and correlated noise, we assess the ability of various data processing pipelines to recover the power spectrum of HI brightness temperature fluctuations. We use MeerKAT pilot data to assess the level of 1/f noise expected for the MeerKLASS survey and use these measurements to create realistic levels of time-correlated noise for our simulations. We find the time-correlated noise component within the pilot data to be between 10 and 20 times higher than the white noise level at the scale of k = 0.04 Mpc^-1. Having determined that the MeerKAT 1/f noise is partially correlated across all the frequency channels, we employ Singular Value Decomposition (SVD) as a technique to remove both the 1/f noise and Galactic foregrounds but find that over-cleaning results in the removal of HI power at large (angular and radial) scales; a power loss of 40 per cent is seen for a 3-mode SVD clean at the scale of k = 0.04 Mpc^-1. We compare the impact of map-making using weighting by the full noise covariance (i.e. including a 1/f component), as opposed to just a simple unweighted binning, finding that including the time-correlated noise information reduces the excess power added by 1/f noise by up to 30 per cent., Comment: 14 pages, 12 figures, accepted by MNRAS. This replacement features updated simulations which now include foreground emissions, telescope slew time and a receiver and elevation-dependant temperature contribution to each receiver which changes over frequency

We present new VLA observations, between 6.8mm and 66mm, of the edge-on Class~I disk IRAS04302+2247. Observations at 6.8mm and 9.2mm lead to the detection of thermal emission from the disk, while shallow observations at the other wavelengths are used to correct for emission from other processes. The disk radial brightness profile transitions from broadly extended in previous ALMA 0.9mm and 2.1mm observations to much more centrally brightened at 6.8mm and 9.2mm, which can be explained by optical depth effects. The radiative transfer modeling of the 0.9mm, 2.1mm, and 9.2mm data suggests that the grains are smaller than 1cm in the outer regions of the disk and allows us to obtain the first lower limit for the scale height of grains emitting at millimeter wavelengths in a protoplanetary disk. We find that the millimeter dust scale height is between 1au and 6au at a radius 100au from the central star, while the gas scale height is estimated to be about 7au, indicating a modest level of settling. The estimated dust height is intermediate between less evolved Class 0 sources, that are found to be vertically thick, and more evolved Class II sources, which show a significant level of settling. This suggests that we are witnessing an intermediate stage of dust settling., Comment: Accepted for publication in the Astrophysical Journal

Detecting a planetary companion in a short-period orbit through radio emission from the interaction with its host star is a new prospect in exoplanet science. Recently, a tantalising signal was found close to the low-mass stellar system GJ 1151 using LOFAR observations. We studied spectroscopic time-series data of GJ 1151 in order to search for planetary companions, investigate possible signatures of stellar magnetic activity, and to find possible explanations for the radio signal. We used the combined radial velocities measured from spectra acquired with the CARMENES, HARPS-N, and HPF instruments, extracted activity indices from those spectra in order to mitigate the impact of stellar magnetic activity on the data, and performed a detailed analysis of Gaia astrometry and all available photometric time series coming from the MEarth and ASAS-SN surveys. We found a M$>$10.6 M$_{\oplus}$ companion to GJ 1151 in a 390d orbit at a separation of 0.57 au. Evidence for a second modulation is also present; this could be due to long-term magnetic variability or a second (substellar) companion. The star shows episodes of elevated magnetic activity, one of which could be linked to the observed LOFAR radio emission. We show that it is highly unlikely that the detected GJ 1151 b, or any additional outer companion is the source of the detected signal. We cannot firmly rule out the suggested explanation of an undetected short-period planet that could be related to the radio emission, as we establish an upper limit of 1.2 M$_{\oplus}$ for the minimum mass., Comment: 18 pages, 11 figures. Accepted version, A&A (2023)