Search

In this article, we describe the components and procedures for building a drone ready for networking experimentation. In particular, our drone design includes multiple technologies and elements such as 4G/5G connectivity for real-time data transmission, a 360-degree camera for immersive vision and AR/VR, precise GPS for navigation, and a powerful Linux-based system with GPU for computer vision experiments and applications. Component selection and assembly techniques are included, along with software integration for a smooth, seamless operation of advanced edge applications.

The observational detection of some metastable isomers in the interstellar medium with abundances comparable to those of the most stable isomer, or even when the stable isomer is not detected, highlights the importance of non-equilibrium chemistry. This challenges our understanding of the interstellar chemistry. We present a chemical study of isomers through the sulphur isomer pair HNCS and HSCN, since HSCN has been observed in regions where its stable isomer has not been detected, and the observed HNCS/HSCN ratio seems to significantly vary from cold to warm regions. We have used the Nautilus chemical code to model the formation and destruction paths of HNCS and HSCN in different astrochemical scenarios, and the time evolution of the HNCS/HSCN ratio. We have also analysed the influence of the environmental conditions on their chemical abundances. We present an observational detection of the metastable isomer HSCN in the Class I object B1-a, but not of the stable isomer HNCS, despite HNCS lying 3200 K lower in energy than HSCN. Our results show an HNCS/HSCN ratio sensitive to the gas temperature and the evolutionary time, with the highest values obtained at early stages (t<10^4 yr) and low (Tg<20 K) temperatures. The results suggest a different efficiency of the isomerisation processes depending on the source temperature. The progressive decrease of HNCS/HSCN with gas temperature at early evolutionary times indicates that this ratio may be used as a tracer of cold young objects. This work also demonstrates the key role of grain surface chemistry in the formation of the isomer pair HNCS and HSCN in cold regions, and the importance of the ions H2NCS+ and HNCSH+ in warm/hot regions. Since most of the interstellar regions where HSCN is detected are cold regions, a larger sample including sources characterised by high temperatures are needed to corroborate the theoretical results., Comment: 10 pages, 12 figures. Accepted for publication by A&A

The structure and kinematics of the Milky Way disk are largely inferred from the solar vicinity. To gain a comprehensive understanding, it is essential to find reliable tracers in less-explored regions like the bulge and the far side of the disk. Mira variables, which are well-studied and bright standard candles, offer an excellent opportunity to trace intermediate and old populations in these complex regions. We aim to isolate a clean sample of Miras in the Vista Variables in the V\'ia L\'actea survey using Gaussian process algorithms. This sample will be used to study intermediate and old age populations in the Galactic bulge and far disk. Near- and mid-infrared time-series photometry were processed using Gaussian Process algorithms to identify Mira variables and model their light curves. We calibrated selection criteria with a visually inspected sample to create a high-purity sample of Miras, integrating multi-band photometry and kinematic data from proper motions. We present a catalog of 3602 Mira variables. By analyzing photometry, we classify them by O-rich or C-rich surface chemistry and derive selective-to-total extinction ratios of $A_{K_{s}}/E(J - K_{s}) = 0.471 \pm 0.01$ and $A_{K_{s}}/E(H - K_{s}) = 1.320 \pm 0.020$. Using the Mira period-age relation, we find evidence supporting the inside-out formation of the Milky Way disk. The distribution of proper motions and distances aligns with the Galactic rotation curve and disk kinematics. We extend the rotation curve up to R$_{\rm GC} \sim 17 \ \rm{kpc}$ and find no strong evidence of the nuclear stellar disk in our Mira sample. This study constitutes the largest catalog of variable stars on the far side of the Galactic disk to date., Comment: 20 pages, 19 figures, Accepted in A&A

This work presents a low-rank tensor model for multi-dimensional Markov chains. A common approach to simplify the dynamical behavior of a Markov chain is to impose low-rankness on the transition probability matrix. Inspired by the success of these matrix techniques, we present low-rank tensors for representing transition probabilities on multi-dimensional state spaces. Through tensor decomposition, we provide a connection between our method and classical probabilistic models. Moreover, our proposed model yields a parsimonious representation with fewer parameters than matrix-based approaches. Unlike these methods, which impose low-rankness uniformly across all states, our tensor method accounts for the multi-dimensionality of the state space. We also propose an optimization-based approach to estimate a Markov model as a low-rank tensor. Our optimization problem can be solved by the alternating direction method of multipliers (ADMM), which enjoys convergence to a stationary solution. We empirically demonstrate that our tensor model estimates Markov chains more efficiently than conventional techniques, requiring both fewer samples and parameters. We perform numerical simulations for both a synthetic low-rank Markov chain and a real-world example with New York City taxi data, showcasing the advantages of multi-dimensionality for modeling state spaces.

Discontinuous Named Entity Recognition (DNER) presents a challenging problem where entities may be scattered across multiple non-adjacent tokens, making traditional sequence labelling approaches inadequate. Existing methods predominantly rely on custom tagging schemes to handle these discontinuous entities, resulting in models tightly coupled to specific tagging strategies and lacking generalisability across diverse datasets. To address these challenges, we propose TriG-NER, a novel Triplet-Grid Framework that introduces a generalisable approach to learning robust token-level representations for discontinuous entity extraction. Our framework applies triplet loss at the token level, where similarity is defined by word pairs existing within the same entity, effectively pulling together similar and pushing apart dissimilar ones. This approach enhances entity boundary detection and reduces the dependency on specific tagging schemes by focusing on word-pair relationships within a flexible grid structure. We evaluate TriG-NER on three benchmark DNER datasets and demonstrate significant improvements over existing grid-based architectures. These results underscore our framework's effectiveness in capturing complex entity structures and its adaptability to various tagging schemes, setting a new benchmark for discontinuous entity extraction., Comment: Code will be made available upon publication

We analyze a sample of 4500 photometrically-selected Ha emitter galaxies at redshifts z~2 and z~4-6 selected from James Webb Space Telescope (JWST) Near-Infrared Camera (NIRCam) medium-band images in the Great Observatories Origins Deep Survey South (GOODS-S). The bulk (80%) of the galaxies in our sample have stellar masses lower than 10^8 Msun, with a median stellar mass of ~10^7.3 Msun. We derive Ha rest-frame equivalent widths ( EW0(Ha) ), line fluxes, and star formation rates using a robust photometric excess technique tested against spectroscopic measurements, being sensitive to EW0(Ha) > 75 A. Both EW0(Ha) and sSFR(Ha) anti-correlate with stellar mass, and at fixed stellar mass, show a steep increasing trend with redshift sSFR(Ha) ~ (1+z)^2.55. By comparing the Ha and rest-frame UV-derived SFRs, we probe the star formation histories (SFHs) of our galaxies in the past 100 Myr. The fraction of low-mass galaxies (M < 10^8 Msun) with signs of bursty star formation from their SFR(Ha)/SFR(UV) is ~ 50 %. It quickly drops to ~ 25 % for M > 10^8 Msun. This is consistent with the results from sSFR(Ha), showing 80% and 17%, respectively. SFR(Ha)/SFR(UV) is a stricter criterion than those based on the galaxy sSFR(Ha), as it only identifies the strongest starbursts, the ones at the initial phases of a bursty star-formation episode., Comment: 20 pages, 10 figures. Submitted to ApJ, comments are welcome

Modeling the intrinsic alignment (IA) of galaxies poses a challenge to weak lensing analyses. The Dark Energy Survey is expected to be less impacted by IA when limited to blue, star-forming galaxies. The cosmological parameter constraints from this blue cosmic shear sample are stable to IA model choice, unlike passive galaxies in the full DES Y3 sample, the goodness-of-fit is improved and the $\Omega_{m}$ and $S_8$ better agree with the cosmic microwave background. Mitigating IA with sample selection, instead of flexible model choices, can reduce uncertainty in $S_8$ by a factor of 1.5., Comment: Data access available at https://jamiemccullough.github.io/data/blueshear/

After reionization, neutral hydrogen (HI) traces the large-scale structure (LSS) of the Universe, enabling HI intensity mapping (IM) to capture the LSS in 3D and constrain key cosmological parameters. We present a new framework utilizing higher-order cross-correlations to study HI clustering around galaxies, tested using real-space data from the IllustrisTNG300 simulation. This approach computes the joint distributions of $k$-nearest neighbor ($k$NN) optical galaxies and the HI brightness temperature field smoothed at relevant scales (the $k$NN-field framework), providing sensitivity to all higher-order cross-correlations, unlike two-point statistics. To simulate HI data from actual surveys, we add random thermal noise and apply a simple foreground cleaning model, filtering out Fourier modes of the brightness temperature field with $k_\parallel < k_{\rm min,\parallel}$. Under current levels of thermal noise and foreground cleaning, typical of a Canadian Hydrogen Intensity Mapping Experiment (CHIME)-like survey, the HI-galaxy cross-correlation signal in our simulations, using the $k$NN-field framework, is detectable at $>30\sigma$ across $r = [3,12] \, h^{-1}$Mpc. In contrast, the detectability of the standard two-point correlation function (2PCF) over the same scales depends strongly on the foreground filter: a sharp $k_\parallel$ filter can spuriously boost detection to $8\sigma$ due to position-space ringing, whereas a less sharp filter yields no detection. Nonetheless, we conclude that $k$NN-field cross-correlations are robustly detectable across a broad range of foreground filtering and thermal noise conditions, suggesting their potential for enhanced constraining power over 2PCFs., Comment: 14 pages, 9 figures (with 2 figures included in the appendix), 2 tables. Comments are welcome

We study the abundance, radial distribution, and orbits of luminous satellites in simulations of Milky Way-mass dark halos in the $\Lambda$CDM cosmology. We follow the evolution of a halo from the Aquarius project and the formation of satellites with the GALFORM semi-analytic model of galaxy formation, in which gas cools radiatively into halos before reionization and in halos that exceed a redshift-dependent ``critical'' virial mass after that. Subhalos are prone to disruption in the tidal field of the main halo, with the number of surviving self-bound subhalos increasing with resolution. Even in the highest resolution simulation (Aq-L1, with particle mass $m_{\rm p}\sim10^3\, M_\odot$), a substantial number of subhalos are disrupted but their galaxies may survive as ``orphans''. Whether or not a satellite becomes an orphan depends primarily on its time of infall. When orphans are included, the simulations yield a converged satellite stellar mass function across different resolution levels. The total number of luminous satellites is sensitive to the assumed redshift of reionization, but the shape of the satellite stellar mass function is robust, peaking at the stellar mass ($\sim 10^3\, M_\odot$) of a halo just above the critical threshold. Most orphans are found in the central regions of the main halo and make up roughly half of all satellites in Aq-L1. When orphans are taken into account there is no need to populate subhalos below the critical mass with satellites to fit the radial distribution of Milky Way satellites, as had been argued in recent work. Our model predicts that orphans dominate the ultra-faint population and that many more satellites with small apocentric radii should be detected in upcoming deep wide-field surveys., Comment: 16 pages, 10 figures, submitted to MNRAS

The Hamiltonian Mean-Field (HMF) model is a long-range interaction model that exhibits quasi-stationary states associated with a phase transition. Its quasi-stationary states with a lifetime diverging with the number of particles in the system. These states are characterized by homogeneous or non-homogeneous structures in phase-space. There exists a phase-transition between these states that have been traditionally characterized by the their mean magnetization. However, the magnetization also exhibits fluctuations in time around its mean value, that can be an indicator of the kind of quasi-stationary state. Thus, we want to characterize the quasi-stationary states of the HMF model through the time-series of the magnetization and its fluctuations through a measure of information, i.e. the permutation entropy and the complexity-entropy plane. Permutation entropy is a measure for characterizing chaotic time series, especially in the presence of dynamic and observational noise, as it is computationally and conceptually simple. For non-homogeneous states, the permutation entropy shows that the HMF model tends towards order, while the magnetizacion fluctuations reveal reduced structures in time. On the contrary, homogeneous states tend to disorder and the structures of the magnetization fluctuations increase as the initial magnetization is larger. In all the study cases of this thesis, the HMF model is characterized by low entropy values but the highest possible complexity value. Thus, the HMF model can be described as a chaotic, deterministic and intermitent system. This aligns with previous studies of the model in the phase space. The results demonstrate that the HMF model can be understood and interpreted from the fluctuations of magnetization using permutation entropy and the complexity-entropy plane., Comment: 15 pages, 6 figures, submitted to Physical Review E

The baryonic physics shaping galaxy formation and evolution are complex, spanning a vast range of scales and making them challenging to model. Cosmological simulations rely on subgrid models that produce significantly different predictions. Understanding how models of stellar and active galactic nuclei (AGN) feedback affect baryon behavior across different halo masses and redshifts is essential. Using the SIMBA and IllustrisTNG suites from the Cosmology and Astrophysics with MachinE Learning Simulations (CAMELS) project, we explore the effect of parameters governing the subgrid implementation of stellar and AGN feedback. We find that while IllustrisTNG shows higher cumulative feedback energy across all halos, SIMBA demonstrates a greater spread of baryons, quantified by the closure radius and circumgalactic medium (CGM) gas fraction. This suggests that feedback in SIMBA couples more effectively to baryons and drives them more efficiently within the host halo. There is evidence that different feedback modes are highly interrelated in these subgrid models. Parameters controlling stellar feedback efficiency significantly impact AGN feedback, as seen in the suppression of black hole mass growth and delayed activation of AGN feedback to higher mass halos with increasing stellar feedback efficiency in both simulations. Additionally, AGN feedback efficiency parameters affect the CGM gas fraction at low halo masses in SIMBA, hinting at complex, non-linear interactions between AGN and SNe feedback modes. Overall, we demonstrate that stellar and AGN feedback are intimately interwoven, especially at low redshift, due to subgrid implementation, resulting in halo property effects that might initially seem counterintuitive., Comment: 28 pages, 8 figures, submitted to ApJ

In an era increasingly dominated by digital platforms, the spread of misinformation poses a significant challenge, highlighting the need for solutions capable of assessing information veracity. Our research contributes to the field of Explainable Artificial Antelligence (XAI) by developing transformer-based fact-checking models that contextualise and justify their decisions by generating human-accessible explanations. Importantly, we also develop models for automatic evaluation of explanations for fact-checking verdicts across different dimensions such as \texttt{(self)-contradiction}, \texttt{hallucination}, \texttt{convincingness} and \texttt{overall quality}. By introducing human-centred evaluation methods and developing specialised datasets, we emphasise the need for aligning Artificial Intelligence (AI)-generated explanations with human judgements. This approach not only advances theoretical knowledge in XAI but also holds practical implications by enhancing the transparency, reliability and users' trust in AI-driven fact-checking systems. Furthermore, the development of our metric learning models is a first step towards potentially increasing efficiency and reducing reliance on extensive manual assessment. Based on experimental results, our best performing generative model \textsc{ROUGE-1} score of 47.77, demonstrating superior performance in generating fact-checking explanations, particularly when provided with high-quality evidence. Additionally, the best performing metric learning model showed a moderately strong correlation with human judgements on objective dimensions such as \texttt{(self)-contradiction and \texttt{hallucination}, achieving a Matthews Correlation Coefficient (MCC) of around 0.7.}, Comment: Forthcoming in Engineering Applications of Artificial Intelligence

Understanding the relationships between geometry and topology is a central theme in Riemannian geometry. We establish two results on the fundamental groups of open (complete and noncompact) $n$-manifolds with nonnegative Ricci curvature and linear volume growth. First, we show that the fundamental group of such a manifold contains a subgroup $\mathbb{Z}^k$ of finite index, where $0\le k\le n-1$. Second, we prove that if the Ricci curvature is positive everywhere, then the fundamental group is finite. The proofs are based on an analysis of the equivariant asymptotic geometry of successive covering spaces and a plane/halfplane rigidity result for RCD spaces.

The Workflows Community Summit gathered 111 participants from 18 countries to discuss emerging trends and challenges in scientific workflows, focusing on six key areas: time-sensitive workflows, AI-HPC convergence, multi-facility workflows, heterogeneous HPC environments, user experience, and FAIR computational workflows. The integration of AI and exascale computing has revolutionized scientific workflows, enabling higher-fidelity models and complex, time-sensitive processes, while introducing challenges in managing heterogeneous environments and multi-facility data dependencies. The rise of large language models is driving computational demands to zettaflop scales, necessitating modular, adaptable systems and cloud-service models to optimize resource utilization and ensure reproducibility. Multi-facility workflows present challenges in data movement, curation, and overcoming institutional silos, while diverse hardware architectures require integrating workflow considerations into early system design and developing standardized resource management tools. The summit emphasized improving user experience in workflow systems and ensuring FAIR workflows to enhance collaboration and accelerate scientific discovery. Key recommendations include developing standardized metrics for time-sensitive workflows, creating frameworks for cloud-HPC integration, implementing distributed-by-design workflow modeling, establishing multi-facility authentication protocols, and accelerating AI integration in HPC workflow management. The summit also called for comprehensive workflow benchmarks, workflow-specific UX principles, and a FAIR workflow maturity model, highlighting the need for continued collaboration in addressing the complex challenges posed by the convergence of AI, HPC, and multi-facility research environments.

Thermal effects in a one-dimensional Su-Schrieffer-Hegger (SSH) topological insulator are studied. Particularly, we focus on quantum information processing (QIP) capacity for thermal ensembles. To evaluate QIP an optimized quantum Fisher information (OQFI) is introduced as a quantifier of entanglement and topological phases are calculated by a definition in real space for the electric polarization of mixture states. For the thermal ensemble, there is a relationship between the Fisher metric and the electric polarization in such a way that in the topological region, there is more entanglement, therefore, these generate more robustness and protection in the quantum information due to thermal effects. Also, long-range hopping effects are studied and it is found that in this case, the OQFI captures these topological phase transitions in the limit of low temperature by this formalism in real space.

Tactile perception is essential for human interaction with the environment and is becoming increasingly crucial in robotics. Tactile sensors like the BioTac mimic human fingertips and provide detailed interaction data. Despite its utility in applications like slip detection and object identification, this sensor is now deprecated, making many existing datasets obsolete. This article introduces a novel method for translating data between tactile sensors by exploiting sensor deformation information rather than output signals. We demonstrate the approach by translating BioTac signals into the DIGIT sensor. Our framework consists of three steps: first, converting signal data into corresponding 3D deformation meshes; second, translating these 3D deformation meshes from one sensor to another; and third, generating output images using the converted meshes. Our approach enables the continued use of valuable datasets., Comment: Extended Abstract. Accepted in ICRA@40 (40th Anniversary of the IEEE International Conference on Robotics and Automation) 23-26 September, 2024 Rotterdam, Netherlands

The increased brightness temperature of young rocky protoplanets during their magma ocean epoch makes them potentially amenable to atmospheric characterization to distances from the solar system far greater than thermally equilibrated terrestrial exoplanets, offering observational opportunities for unique insights into the origin of secondary atmospheres and the near surface conditions of prebiotic environments. The Large Interferometer For Exoplanets (LIFE) mission will employ a space-based mid-infrared nulling interferometer to directly measure the thermal emission of terrestrial exoplanets. Here, we seek to assess the capabilities of various instrumental design choices of the LIFE mission concept for the detection of cooling protoplanets with transient high-temperature magma ocean atmospheres, in young stellar associations in particular. Using the LIFE mission instrument simulator (LIFEsim) we assess how specific instrumental parameters and design choices, such as wavelength coverage, aperture diameter, and photon throughput, facilitate or disadvantage the detection of protoplanets. We focus on the observational sensitivities of distance to the observed planetary system, protoplanet brightness temperature using a blackbody assumption, and orbital distance of the potential protoplanets around both G- and M-dwarf stars. Our simulations suggest that LIFE will be able to detect (S/N $\geq$ 7) hot protoplanets in young stellar associations up to distances of $\approx$100 pc from the solar system for reasonable integration times (up to $\sim$hours). Detection of an Earth-sized protoplanet orbiting a solar-sized host star at 1 AU requires less than 30 minutes of integration time. M-dwarfs generally need shorter integration times. The contribution from wavelength regions $<$6 $\mu$m is important for decreasing the detection threshold and discriminating emission temperatures., Comment: 18 pages, 19 figures; accepted for publication in A&A

Over the last decades, evolutionary population synthesis models have powered an unmatched leap forward in our understanding of galaxies. From dating the age of the first galaxies in the Universe to detailed measurements of the chemical composition of nearby galaxies, the success of this approach built upon simple stellar population (SSP) spectro-photometric models is unquestionable. However, the internal constraints inherent to the construction of SSP models may hinder our ability to analyze the integrated spectra of galaxies in situations where the SSP assumption does not sufficiently hold. Thus, here we revisit the possibilities of fitting galaxy spectra as a linear combination of stellar templates without assuming any a priori knowledge on stellar evolution. We showcase the sensitivity of this alternative approach to changes in the stellar population properties, in particular the direct connection to variations in the stellar initial mass function, as well as its advantages when dealing with non-canonical integrated populations and semi-resolved observations. Furthermore, our analysis demonstrates that the absorption spectra of galaxies can be used to independently constrain tellar evolution theory beyond the limited conditions of the solar neighborhood., Comment: 5 pages, 4 figures. Accepted for publication in Astronomy & Astrophysics Letters

Large Language Models~(LLMs) have demonstrated capabilities across various applications but face challenges such as hallucination, limited reasoning abilities, and factual inconsistencies, especially when tackling complex, domain-specific tasks like question answering~(QA). While Knowledge Graphs~(KGs) have been shown to help mitigate these issues, research on the integration of LLMs with background KGs remains limited. In particular, user accessibility and the flexibility of the underlying KG have not been thoroughly explored. We introduce AGENTiGraph (Adaptive Generative ENgine for Task-based Interaction and Graphical Representation), a platform for knowledge management through natural language interaction. It integrates knowledge extraction, integration, and real-time visualization. AGENTiGraph employs a multi-agent architecture to dynamically interpret user intents, manage tasks, and integrate new knowledge, ensuring adaptability to evolving user requirements and data contexts. Our approach demonstrates superior performance in knowledge graph interactions, particularly for complex domain-specific tasks. Experimental results on a dataset of 3,500 test cases show AGENTiGraph significantly outperforms state-of-the-art zero-shot baselines, achieving 95.12\% accuracy in task classification and 90.45\% success rate in task execution. User studies corroborate its effectiveness in real-world scenarios. To showcase versatility, we extended AGENTiGraph to legislation and healthcare domains, constructing specialized KGs capable of answering complex queries in legal and medical contexts., Comment: 30 pages, 7 figures; Submitted to COLING 2025 System Demonstrations Track

We present a study on the inference of cosmological and astrophysical parameters using stacked galaxy cluster profiles. Utilizing the CAMELS-zoomGZ simulations, we explore how various cluster properties--such as X-ray surface brightness, gas density, temperature, metallicity, and Compton-y profiles--can be used to predict parameters within the 28-dimensional parameter space of the IllustrisTNG model. Through neural networks, we achieve a high correlation coefficient of 0.97 or above for all cosmological parameters, including $\Omega_{\rm m}$, $H_0$, and $\sigma_8$, and over 0.90 for the remaining astrophysical parameters, showcasing the effectiveness of these profiles for parameter inference. We investigate the impact of different radial cuts, with bins ranging from $0.1R_{200c}$ to $0.7R_{200c}$, to simulate current observational constraints. Additionally, we perform a noise sensitivity analysis, adding up to 40\% Gaussian noise (corresponding to signal-to-noise ratios as low as 2.5), revealing that key parameters such as $\Omega_{\rm m}$, $H_0$, and the IMF slope remain robust even under extreme noise conditions. We also compare the performance of full radial profiles against integrated quantities, finding that profiles generally lead to more accurate parameter inferences. Our results demonstrate that stacked galaxy cluster profiles contain crucial information on both astrophysical processes within groups and clusters and the underlying cosmology of the universe. This underscores their significance for interpreting the complex data expected from next-generation surveys and reveals, for the first time, their potential as a powerful tool for parameter inference., Comment: Submitted to ApJ

The efficient release of magnetic energy in astrophysical plasmas, such as during solar flares, can in principle be achieved through magnetic diffusion, at a rate determined by the associated electric field. However, attempts at measuring electric fields in the solar atmosphere are scarce, and none exist for sites where the magnetic energy is presumably released. Here, we present observations of an energetic event using the National Science Foundation's Daniel K. Inouye Solar Telescope, where we detect the polarization signature of electric fields associated with magnetic diffusion. We measure the linear and circular polarization across the hydrogen H-epsilon Balmer line at 397 nm at the site of a brightening event in the solar chromosphere. Our spectro-polarimetric modeling demonstrates that the observed polarization signals can only be explained by the presence of electric fields, providing conclusive evidence of magnetic diffusion, and opening a new window for the quantitative study of this mechanism in space plasmas., Comment: 24 pages, 11 figures, published in Nature Communications

Rapid search and rescue is critical to maximizing survival rates following natural disasters. However, these efforts are challenged by the need to search large disaster zones, lack of reliability in the communications infrastructure, and a priori unknown numbers of objects of interest (OOIs), such as injured survivors. Aerial robots are increasingly being deployed for search and rescue due to their high mobility, but there remains a gap in deploying multi-robot autonomous aerial systems for methodical search of large environments. Prior works have relied on preprogrammed paths from human operators or are evaluated only in simulation. We bridge these gaps in the state of the art by developing and demonstrating a decentralized active search system, which biases its trajectories to take additional views of uncertain OOIs. The methodology leverages stochasticity for rapid coverage in communication denied scenarios. When communications are available, robots share poses, goals, and OOI information to accelerate the rate of search. Extensive simulations and hardware experiments in Bloomingdale, OH, are conducted to validate the approach. The results demonstrate the active search approach outperforms greedy coverage-based planning in communication-denied scenarios while maintaining comparable performance in communication-enabled scenarios.

The first very long baseline interferometry (VLBI) detections at 870$\mu$m wavelength (345$\,$GHz frequency) are reported, achieving the highest diffraction-limited angular resolution yet obtained from the surface of the Earth, and the highest-frequency example of the VLBI technique to date. These include strong detections for multiple sources observed on inter-continental baselines between telescopes in Chile, Hawaii, and Spain, obtained during observations in October 2018. The longest-baseline detections approach 11$\,$G$\lambda$ corresponding to an angular resolution, or fringe spacing, of 19$\mu$as. The Allan deviation of the visibility phase at 870$\mu$m is comparable to that at 1.3$\,$mm on the relevant integration time scales between 2 and 100$\,$s. The detections confirm that the sensitivity and signal chain stability of stations in the Event Horizon Telescope (EHT) array are suitable for VLBI observations at 870$\mu$m. Operation at this short wavelength, combined with anticipated enhancements of the EHT, will lead to a unique high angular resolution instrument for black hole studies, capable of resolving the event horizons of supermassive black holes in both space and time., Comment: Corresponding author: S. Doeleman

We aim to trace the evolution of eight edge-on star-forming disc galaxies through the analysis of stellar population properties of their thin and thick discs. These galaxies have relatively low stellar masses (4 $\times$ 10$^9$ to 6 $\times$ 10$^{10}$ $M_{\odot}$). We use Multi-Unit Spectroscopic Explorer (MUSE) observations and full-spectrum fitting to produce spatially resolved maps of ages, metallicities and [Mg/Fe]-abundances and extract the star formation histories of stellar discs. Our maps show thick discs that are on average older, more metal-poor and more [Mg/Fe]-enhanced than thin discs. However, age differences between thin and thick discs are small (around 2 Gyr) and the thick discs are younger than previously observed in more massive and more quiescent galaxies. Both thin and thick discs show mostly sub-solar metallicities, and the vertical metallicity gradient is milder than previously observed in similar studies. [Mg/Fe] differences between thick and thin discs are not sharp. The star formation histories of thick discs extend down to recent times, although most of the mass in young stars was formed in thin discs. Our findings show thick discs that are different from old quiescent thick discs previously observed in galaxies of different morphologies and/or different masses. We propose that thick discs in these galaxies did not form quickly at high redshift, but slowly in an extended time. Also, the thin discs formed slowly, however, a larger mass fraction was created at very recent times., Comment: 20 pages, 9 figures, submitted to A&A

(Abridged) We explore the chemistry of the most abundant C, O, S, and N bearing species in molecular clouds, in the context of the IRAM 30 m Large Programme Gas phase Elemental abundances in Molecular Clouds (GEMS). In this work, we aim to assess the limitations introduced in the observational works when a uniform density is assumed along the line of sight for fitting the observations, developing a very simple numerical model of a turbulent box. We perform a MHD simulation in order to reproduce the turbulent steady-state of a turbulent box with properties typical of a molecular filament before collapse. We post-process the results of the MHD simulation with a chemical code to predict molecular abundances, and then post-process this cube with a radiative transfer code to create synthetic emission maps for a series of rotational transitions observed during the GEMS project. From the chemical point of view, we find that turbulence produces variations on the predicted abundances, but they are more or less critical depending on the chosen transition and the chemical age. When compared to real observations, the results from the turbulent simulation provides a better fit than when assuming a uniform gas distribution along the line of sight. In the view of our results, we conclude that taking into account turbulence when fitting observations might significantly improve the agreement with model predictions. This is especially important for sulfur bearing species that are very sensitive to the variations of density produced by turbulence at early times (0.1 Myr). The abundance of CO is also quite sensitive to turbulence when considering the evolution beyond a few 0.1 Myr., Comment: Published in Astronomy and Astrophysics

In the pursuit of alternative and less invasive medical treatments, magnetic nanoparticles (NPs) have gained significant relevance. Iron carbides NPs stand out for their higher saturation magnetizations compared to iron oxides, while maintaining a suitable biocompatibility. In this work, high control is achieved over the composition and morphology of Fe3C/C encapsulated core/shell nanoparticles through fine-tuning of the sol-gel synthesis parameters. Specifically, the impact of decreasing each surfactant concentration added, nt, the same both for oleylamine (ON) and oleic acid (OA), has been explored. A minimum value for such parameter denoted by nt,min. was required to produce pure Fe3C@C NP-composites. For nt < 4 mmol, some minor {\alpha}-Fe impurities arise, and the effective carburization becomes unstable due to insufficient carbon. The magnetic properties of the materials prepared were optimized by reducing the excess carbon from surfactants, resulting in saturation magnetization values of 86 emu/g. (for pure Fe3C at nt = 5 mmol) and 102 emu/g (for Fe3C and <2 % w.t. of {\alpha}-Fe impurity at nt = 4 mmol). In view of this, several cytotoxicity studies for different Fe3C@C samples were conducted, exhibiting excellent biocompatibility in cell-based assays, which could lead to potential application at the forefront of biomedical fields.

Social media images have proven to be a valuable source of information for understanding human interactions with important subjects such as cultural heritage, biodiversity, and nature, among others. The task of grouping such images into a number of semantically meaningful clusters without labels is challenging due to the high diversity and complex nature of the visual content in addition to their large volume. On the other hand, recent advances in Large Visual Models (LVMs), Large Language Models (LLMs), and Large Visual Language Models (LVLMs) provide an important opportunity to explore new productive and scalable solutions. This work proposes, analyzes, and compares various approaches based on one or more state-of-the-art LVM, LLM, and LVLM, for mapping social media images into a number of predefined classes. As a case study, we consider the problem of understanding the interactions between humans and nature, also known as Nature's Contribution to People or Cultural Ecosystem Services (CES). Our experiments show that the highest-performing approaches, with accuracy above 95%, still require the creation of a small labeled dataset. These include the fine-tuned LVM DINOv2 and the LVLM LLaVA-1.5 combined with a fine-tuned LLM. The top fully unsupervised approaches, achieving accuracy above 84%, are the LVLMs, specifically the proprietary GPT-4 model and the public LLaVA-1.5 model. Additionally, the LVM DINOv2, when applied in a 10-shot learning setup, delivered competitive results with an accuracy of 83.99%, closely matching the performance of the LVLM LLaVA-1.5., Comment: 17 pages, 9 figures

We test the robustness of simulation-based inference (SBI) in the context of cosmological parameter estimation from galaxy cluster counts and masses in simulated optical datasets. We construct ``simulations'' using analytical models for the galaxy cluster halo mass function (HMF) and for the observed richness (number of observed member galaxies) to train and test the SBI method. We compare the SBI parameter posterior samples to those from an MCMC analysis that uses the same analytical models to construct predictions of the observed data vector. The two methods exhibit comparable performance, with reliable constraints derived for the primary cosmological parameters, ($\Omega_m$ and $\sigma_8$), and richness-mass relation parameters. We also perform out-of-domain tests with observables constructed from galaxy cluster-sized halos in the Quijote simulations. Again, the SBI and MCMC results have comparable posteriors, with similar uncertainties and biases. Unsurprisingly, upon evaluating the SBI method on thousands of simulated data vectors that span the parameter space, SBI exhibits worsened posterior calibration metrics in the out-of-domain application. We note that such calibration tests with MCMC is less computationally feasible and highlight the potential use of SBI to stress-test limitations of analytical models, such as in the use for constructing models for inference with MCMC.

This paper proposes a variance-based measure of importance for coherent systems with dependent and heterogeneous components. The particular cases of independent components and homogeneous components are also considered. We model the dependence structure among the components by the concept of copula. The proposed measure allows us to provide the best estimation of the system lifetime, in terms of the mean squared error, under the assumption that the lifetime of one of its components is known. We include theoretical results that are useful to calculate a closed-form of our measure and to compare two components of a system. We also provide some procedures to approximate the importance measure by Monte Carlo simulation methods. Finally, we illustrate the main results with several examples.

The article is focused on studying how to predict the failure times of coherent systems from the early failure times of their components. Both the cases of independent and dependent components are considered by assuming that they are identically distributed (homogeneous components). The heterogeneous components' case can be addressed similarly but more complexly. The present study is for non-repairable systems, but the information obtained could be used to decide if a maintenance action should be carried out at time t. Different cases are considered regarding the information available at time t. We use quantile regression techniques to predict the system failure times and to provide prediction intervals. The theoretical results are applied to specific system structures in some illustrative examples.

The carbon isotope ratio is a powerful tool for studying the evolution of stellar systems. Recent detections of CO isotopologues in disks and exoplanet atmospheres pointed towards significant fractionation in these systems. In order to understand the evolution of this quantity, it is crucial to trace the isotope abundance from stellar nurseries to planetary systems. During the protostellar stage the multiple vibrational modes of CO$_2$ and CO ice provide a unique opportunity to examine the carbon isotope ratio in the solid state. Now with the sensitivity of the \textit{James Webb Space Telescope}, these absorption features have become accessible at high S/N in Solar-mass systems. We quantify the $^{12}$CO$_2$/$^{13}$CO$_2$ and the $^{12}$CO/$^{13}$CO isotope ratios in 17 class 0/I low mass protostars from the $^{12}$CO$_2$ combination modes (2.70 ${\mu}$m and 2.77 ${\mu}$m), the $^{12}$CO$_2$ stretching mode (4.27 ${\mu}$m), the $^{13}$CO$_2$ stretching mode (4.39 ${\mu}$m), the $^{12}$CO$_2$ bending mode (15.2 ${\mu}$m), the $^{12}$CO stretching mode (4.67 ${\mu}$m) and the $^{13}$CO stretching mode (4.78 ${\mu}$m) using JWST observations. We also report a detection of the $^{12}$CO overtone mode at 2.35 ${\mu}$m. The $^{12}$CO$_2$/$^{13}$CO$_2$ ratios are in agreement and we find mean ratios of 85 $\pm$ 23, 76 $\pm$ 12 and 97 $\pm$ 17 for the 2.70 ${\mu}$m, 4.27 ${\mu}$m and the 15.2 ${\mu}$m bands, respectively. The main source of uncertainty stem from the error on the band strengths. The $^{12}$CO/$^{13}$CO ratios derived from the 4.67 ${\mu}$m bands are consistent, albeit elevated with respect to the $^{12}$CO$_2$/$^{13}$CO$_2$ ratios and we find a mean ratio of 165 $\pm$ 52. These findings indicate that ices leave the pre-stellar stage with elevated carbon isotope ratios relative to the interstellar medium and that fractionation becomes significant during the later stages.

Device-independent quantum key distribution (DIQKD) provides the strongest form of quantum security, as it allows two honest users to establish secure communication channels even when using fully uncharacterized quantum devices. The security proof of DIQKD is derived from the violation of a Bell inequality, mitigating side-channel attacks by asserting the presence of nonlocality. This enhanced security comes at the cost of a challenging implementation, especially over long distances, as losses make Bell tests difficult to conduct successfully. Here, we propose a photonic realization of DIQKD, utilizing a heralded preparation of a single-photon path entangled state between the honest users. Being based on single-photon interference effects, the obtained secret key rate scales with the square root of the quantum channel transmittance. This leads to positive key rates over distances of up to hundreds of kilometers, making the proposed setup a promising candidate for securing long-distance communication in quantum networks.

The Seidov limit is a bound on the maximum latent heat that a presumed first-order phase transition of neutron-star matter can have before its excess energy density, not compensated by additional pressure, results in gravitational collapse. Because latent heat forces an apparent nonanalytic behaviour in plots correlating physical quantities (kinks in two-dimensional, ridges in three-dimensional ones), it can be constrained by data. As the onset of collapse depends on the intensity of gravity, testing for sudden derivative changes and, if they are large, breaching the Seidov limit would reward with two successive discoveries: such a phase transition (which could stem from hadron matter but also from a gravitational phase transition), and a modification of General Relativity (thus breaking the matter/gravity degeneracy). We illustrate the point with $f(R)=R+\alpha R^2$ metric gravity.

We use the James Webb Space Telescope (JWST) and its Mid-Infrared Instrument (MIRI) (5-28 um), to study the embedded HH 211 flow. We map a 0.95'x0.22' region, covering the full extent of the blue-shifted lobe, the central protostellar region, and a small portion of the red-shifted lobe. The jet driving source is not detected even at the longest mid-IR wavelengths. The overall morphology of the flow consists of a highly collimated jet, mostly molecular (H2, HD) with an inner atomic ([FeI], [FeII], [SI], [NiII]) structure. The jet shocks the ambient medium, producing several large bow-shocks, rich in forbidden atomic and molecular lines, and is driving an H2 molecular outflow, mostly traced by low-J, v=0 transitions. Moreover, 0-0 S(1) uncollimated emission is also detected down to 2"-3" (~650-1000 au) from the source, tracing a cold (T=200-400 K), less dense and poorly collimated molecular wind. The atomic jet ([FeII] at 26 um) is detected down to ~130 au from source, whereas the lack of H2 emission close to the source is likely due to the large visual extinction. Dust continuum-emission is detected at the terminal bow-shocks, and in the blue- and red-shifted jet, being likely dust lifted from the disk. The jet shows an onion-like structure, with layers of different size, velocity, temperature, and chemical composition. Moreover, moving from the inner jet to the outer bow-shocks, different physical, kinematic and excitation conditions for both molecular and atomic gas are observed. The jet mass-flux rate, momentum, and momentum flux of the warm H2 component are up to one order of magnitude higher than those inferred from the atomic jet component. Our findings indicate that the warm H2 component is the primary mover of the outflow, namely it is the most significant dynamical component of the jet, in contrast to jets from more evolved YSOs, where the atomic component is dominant., Comment: Paper accepted in A&A for publication

The $r$-index represented a breakthrough in compressed indexing of repetitive text collections, outperforming its alternatives by orders of magnitude in query time. Its space usage, $O(r)$ where $r$ is the number of runs in the Burrows--Wheeler Transform of the text, is however higher than Lempel--Ziv (LZ) and grammar-based indexes, and makes it uninteresting in various real-life scenarios of milder repetitiveness. We introduce the $sr$-index, a variant that limits the space to $O(\min(r,n/s))$ for a text of length $n$ and a given parameter $s$, at the expense of multiplying by $s$ the time per occurrence reported. The $sr$-index is obtained subsampling the text positions indexed by the $r$-index, being still able to support pattern matching with guaranteed performance. Our experiments show that the theoretical analysis falls short in describing the practical advantages of the $sr$-index, because it performs much better on real texts than on synthetic ones: the $sr$-index retains the performance of the $r$-index while using 1.5--4.0 times less space, sharply outperforming {\em virtually every other} compressed index on repetitive texts in both time and space. Only a particular LZ-based index uses less space than the $sr$-index, but it is an order of magnitude slower. Our second contribution are the $r$-csa and $sr$-csa indexes. Just like the $r$-index adapts the well-known FM-Index to repetitive texts, the $r$-csa adapts Sadakane's Compressed Suffix Array (CSA) to this case. We show that the principles used on the $r$-index turn out to fit naturally and efficiently in the CSA framework. The $sr$-csa is the corresponding subsampled version of the $r$-csa. While the CSA performs better than the FM-Index on classic texts with alphabets larger than DNA, we show that the $sr$-csa outperforms the $sr$-index on repetitive texts over those larger alphabets and some DNA texts as well., Comment: arXiv admin note: text overlap with arXiv:2103.15329

We analyse high-resolution ESPaDOnS/CFHT spectra of 20 very metal-poor stars ([Fe/H]~$<-2.0$) in the solar neighbourhood (within $\sim2$ kpc) selected to be on planar orbits (with a maximum height of $\lesssim4$ kpc). Targets include 11 prograde and 9 retrograde stars, spanning a wide range of eccentricities ($0.20-0.95$). Their chemical abundances are consistent with those observed in the Galactic halo but show a smaller spread, with no notable difference between progrades and retrogrades. This suggests a common chemical evolution and likely a shared formation site (except for one star). In this case, chemical evolution models indicate that the formation site would have had a baryonic mass of $\sim1.4\times10^9\msun$, similar to classical dwarf galaxies. High-energy supernovae and hypernovae are needed to reproduce the [X/Fe] up to the Fe-peak, while fast-rotating massive stars and neutron star merger events explain the [X/Fe] of the neutron-capture elements. The absence of Type Ia supernova signatures suggests a star formation duration of $\lesssim1$~Gyr. Cosmological zoom-in simulations support the scenario that an in-plane infall of a single system could disperse stars over a wide range of angular momenta during the early Galactic assembly. We propose that these stars originated in a proto-Galactic building block, which we name Loki. Less likely, if progrades and retrogrades formed in two different systems, their chemical evolution must have been very similar, with a combined baryonic mass twice that of a single system. Forthcoming surveys will provide a large and homogeneous dataset to investigate whether Loki is associated with any of the known detected structures. A comparison (primarily [$\alpha$/Fe]) with other VMPs moving in planar orbits suggests multiple systems contributed to the Galactic planar population, presenting some differences in their kinematical parameters., Comment: REference list updated; discussion on GSE updated

Relic galaxies, the oldest ultra-compact massive galaxies (UCMGs), contain almost exclusively "pristine" stars formed during an intense star formation (SF) burst at high redshift. As such, they allow us to study in detail the early mechanism of galaxy assembly in the Universe. Using the largest catalogue of spectroscopically confirmed UCMGs for which a degree of relicness (DoR) had been estimated, the INSPIRE catalogue, we investigate whether or not relics prefer dense environments. The objective of this study is to determine if the DoR, which measures how extreme the SF history was, and the surrounding environment are correlated. In order to achieve this goal, we employ the AMICO galaxy cluster catalogue to compute the probability for a galaxy to be a member of a cluster, and measure the local density around each UCMG using machine learning-based photometric redshifts. We find that UCMGs can reside both in clusters and in the field, but objects with very low DoR (< 0.3, i.e., a relatively extended SF history) prefer under-dense environments. We additionally report a correlation between the DoR and the distance from the cluster centre: more extreme relics, when located in clusters, tend to occupy the more central regions of them. We finally outline potential evolution scenarios for UCMGs at different DoR to reconcile their presence in both clusters and field environments, Comment: Accepted for publication on MNRAS, 11 pages, 8 figure, 1 table

Verification and validation of autonomous driving (AD) systems and components is of increasing importance, as such technology increases in real-world prevalence. Safety-critical scenario generation is a key approach to robustify AD policies through closed-loop training. However, existing approaches for scenario generation rely on simplistic objectives, resulting in overly-aggressive or non-reactive adversarial behaviors. To generate diverse adversarial yet realistic scenarios, we propose SEAL, a scenario perturbation approach which leverages learned scoring functions and adversarial, human-like skills. SEAL-perturbed scenarios are more realistic than SOTA baselines, leading to improved ego task success across real-world, in-distribution, and out-of-distribution scenarios, of more than 20%. To facilitate future research, we release our code and tools: https://github.com/cmubig/SEAL, Comment: 8 pages, 4 figures, 2 tables

Sprinting is a determinant ability, especially in team sports. The kinematics of the sprint have been studied in the past using different methods specially developed considering human biomechanics and, among those methods, markerless systems stand out as very cost-effective. On the other hand, we have now multiple general methods for pixel and body tracking based on recent machine learning breakthroughs with excellent performance in body tracking, but these excellent trackers do not generally consider realistic human biomechanics. This investigation first adapts two of these general trackers (MoveNet and CoTracker) for realistic biomechanical analysis and then evaluate them in comparison to manual tracking (with key points manually marked using the software Kinovea). Our best resulting markerless body tracker particularly adapted for sprint biomechanics is termed VideoRun2D. The experimental development and assessment of VideoRun2D is reported on forty sprints recorded with a video camera from 5 different subjects, focusing our analysis in 3 key angles in sprint biomechanics: inclination of the trunk, flex extension of the hip and the knee. The CoTracker method showed huge differences compared to the manual labeling approach. However, the angle curves were correctly estimated by the MoveNet method, finding errors between 3.2{\deg} and 5.5{\deg}. In conclusion, our proposed VideoRun2D based on MoveNet core seems to be a helpful tool for evaluating sprint kinematics in some scenarios. On the other hand, the observed precision of this first version of VideoRun2D as a markerless sprint analysis system may not be yet enough for highly demanding applications. Future research lines towards that purpose are also discussed at the end: better tracking post-processing and user- and time-dependent adaptation., Comment: Preprint of the paper presented to the Workshop on IAPR International Conference on Pattern Recognition (ICPR) 2024

To maximize the amount of information extracted from cosmological datasets, simulations that accurately represent these observations are necessary. However, traditional simulations that evolve particles under gravity by estimating particle-particle interactions (N-body simulations) are computationally expensive and prohibitive to scale to the large volumes and resolutions necessary for the upcoming datasets. Moreover, modeling the distribution of galaxies typically involves identifying virialized dark matter halos, which is also a time- and memory-consuming process for large N-body simulations, further exacerbating the computational cost. In this study, we introduce CHARM, a novel method for creating mock halo catalogs by matching the spatial, mass, and velocity statistics of halos directly from the large-scale distribution of the dark matter density field. We develop multi-stage neural spline flow-based networks to learn this mapping at redshift z=0.5 directly with computationally cheaper low-resolution particle mesh simulations instead of relying on the high-resolution N-body simulations. We show that the mock halo catalogs and painted galaxy catalogs have the same statistical properties as obtained from $N$-body simulations in both real space and redshift space. Finally, we use these mock catalogs for cosmological inference using redshift-space galaxy power spectrum, bispectrum, and wavelet-based statistics using simulation-based inference, performing the first inference with accelerated forward model simulations and finding unbiased cosmological constraints with well-calibrated posteriors. The code was developed as part of the Simons Collaboration on Learning the Universe and is publicly available at \url{https://github.com/shivampcosmo/CHARM}., Comment: 12 pages and 8 figures. This is a Learning the Universe Publication

Graph learning is the fundamental task of estimating unknown graph connectivity from available data. Typical approaches assume that not only is all information available simultaneously but also that all nodes can be observed. However, in many real-world scenarios, data can neither be known completely nor obtained all at once. We present a novel method for online graph estimation that accounts for the presence of hidden nodes. We consider signals that are stationary on the underlying graph, which provides a model for the unknown connections to hidden nodes. We then formulate a convex optimization problem for graph learning from streaming, incomplete graph signals. We solve the proposed problem through an efficient proximal gradient algorithm that can run in real-time as data arrives sequentially. Additionally, we provide theoretical conditions under which our online algorithm is similar to batch-wise solutions. Through experimental results on synthetic and real-world data, we demonstrate the viability of our approach for online graph learning in the presence of missing observations.

Graph neural networks (GNNs) have become a workhorse approach for learning from data defined over irregular domains, typically by implicitly assuming that the data structure is represented by a homophilic graph. However, recent works have revealed that many relevant applications involve heterophilic data where the performance of GNNs can be notably compromised. To address this challenge, we present a simple yet effective architecture designed to mitigate the limitations of the homophily assumption. The proposed architecture reinterprets the role of graph filters in convolutional GNNs, resulting in a more general architecture while incorporating a stronger inductive bias than GNNs based on filter banks. The proposed convolutional layer enhances the expressive capacity of the architecture enabling it to learn from both homophilic and heterophilic data and preventing the issue of oversmoothing. From a theoretical standpoint, we show that the proposed architecture is permutation equivariant. Finally, we show that the proposed GNNs compares favorably relative to several state-of-the-art baselines in both homophilic and heterophilic datasets, showcasing its promising potential.

Covariance-based data processing is widespread across signal processing and machine learning applications due to its ability to model data interconnectivities and dependencies. However, harmful biases in the data may become encoded in the sample covariance matrix and cause data-driven methods to treat different subpopulations unfairly. Existing works such as fair principal component analysis (PCA) mitigate these effects, but remain unstable in low sample regimes, which in turn may jeopardize the fairness goal. To address both biases and instability, we propose Fair coVariance Neural Networks (FVNNs), which perform graph convolutions on the covariance matrix for both fair and accurate predictions. Our FVNNs provide a flexible model compatible with several existing bias mitigation techniques. In particular, FVNNs allow for mitigating the bias in two ways: first, they operate on fair covariance estimates that remove biases from their principal components; second, they are trained in an end-to-end fashion via a fairness regularizer in the loss function so that the model parameters are tailored to solve the task directly in a fair manner. We prove that FVNNs are intrinsically fairer than analogous PCA approaches thanks to their stability in low sample regimes. We validate the robustness and fairness of our model on synthetic and real-world data, showcasing the flexibility of FVNNs along with the tradeoff between fair and accurate performance.

Humans perceive the world through multisensory integration, blending the information of different modalities to adapt their behavior. Contrastive learning offers an appealing solution for multimodal self-supervised learning. Indeed, by considering each modality as a different view of the same entity, it learns to align features of different modalities in a shared representation space. However, this approach is intrinsically limited as it only learns shared or redundant information between modalities, while multimodal interactions can arise in other ways. In this work, we introduce CoMM, a Contrastive MultiModal learning strategy that enables the communication between modalities in a single multimodal space. Instead of imposing cross- or intra- modality constraints, we propose to align multimodal representations by maximizing the mutual information between augmented versions of these multimodal features. Our theoretical analysis shows that shared, synergistic and unique terms of information naturally emerge from this formulation, allowing us to estimate multimodal interactions beyond redundancy. We test CoMM both in a controlled and in a series of real-world settings: in the former, we demonstrate that CoMM effectively captures redundant, unique and synergistic information between modalities. In the latter, CoMM learns complex multimodal interactions and achieves state-of-the-art results on the six multimodal benchmarks., Comment: 22 pages