Search

We study the long-time behaviour of a stochastic Allen-Cahn-Navier-Stokes system modelling the dynamics of binary mixtures of immiscible fluids. The model features two stochastic forcings, one on the velocity in the Navier-Stokes equation and one on the phase variable in the Allen-Cahn equation, and includes the thermodynamically-relevant Flory-Huggins logarithhmic potential. We first show existence of ergodic invariant measures and characterise their support by exploiting ad-hoc regularity estimates and suitable Feller-type and Markov properties. Secondly, we prove that if the noise acting in the Navier-Stokes equation is non-degenerate along a sufficiently large number of low modes, and the Allen-Cahn equation is highly dissipative, then the stochastic flow admits a unique invariant measure and is asymptotically stable with respect to a suitable Wasserstein metric.

We introduce model-based transition rates for controlled compartmental models in mathematical epidemiology, with a focus on the effects of control strategies applied to interacting multi-agent systems describing contact formation dynamics. In the framework of kinetic control problems, we compare two prototypical control protocols: one additive control directly influencing the dynamics and another targeting the interaction strength between agents. The emerging controlled macroscopic models are derived for an SIR compartmentalization to illustrate their impact on epidemic progression and contact interaction dynamics. Numerical results show the effectiveness of this approach in steering the dynamics and controlling epidemic trends, even in scenarios where contact distributions exhibit an overpopulated tail.

Online test-time adaptation (OTTA) of vision-language models (VLMs) has recently garnered increased attention to take advantage of data observed along a stream to improve future predictions. Unfortunately, existing methods rely on dataset-specific hyperparameters, significantly limiting their adaptability to unseen tasks. In response, we propose Online Gaussian Adaptation (OGA), a novel method that models the likelihoods of visual features using Gaussian distributions and incorporates zero-shot priors into an interpretable Maximum A Posteriori (MAP) estimation framework with fixed hyper-parameters across all datasets. We demonstrate that OGA outperforms state-of-the-art methods on most datasets and runs. Additionally, we show that combining OTTA with popular few-shot techniques (a practical yet overlooked setting in prior research) is highly beneficial. Furthermore, our experimental study reveals that common OTTA evaluation protocols, which average performance over at most three runs per dataset, are inadequate due to the substantial variability observed across runs for all OTTA methods. Therefore, we advocate for more rigorous evaluation practices, including increasing the number of runs and considering additional quantitative metrics, such as our proposed Expected Tail Accuracy (ETA), calculated as the average accuracy in the worst 10% of runs. We hope these contributions will encourage more rigorous and diverse evaluation practices in the OTTA community. Code is available at https://github.com/cfuchs2023/OGA .

The zero-shot capabilities of Vision-Language Models (VLMs) have been widely leveraged to improve predictive performance. However, previous works on transductive or test-time adaptation (TTA) often make strong assumptions about the data distribution, such as the presence of all classes. Our work challenges these favorable deployment scenarios, and introduces a more realistic evaluation framework, including: (i) a variable number of effective classes for adaptation within a single batch, and (ii) non-i.i.d. batches of test samples in online adaptation settings. We provide comprehensive evaluations, comparisons, and ablation studies that demonstrate how current transductive or TTA methods for VLMs systematically compromise the models' initial zero-shot robustness across various realistic scenarios, favoring performance gains under advantageous assumptions about the test samples' distributions. Furthermore, we introduce StatA, a versatile method that could handle a wide range of deployment scenarios, including those with a variable number of effective classes at test time. Our approach incorporates a novel regularization term designed specifically for VLMs, which acts as a statistical anchor preserving the initial text-encoder knowledge, particularly in low-data regimes. Code available at https://github.com/MaxZanella/StatA.

We review some basic results on existence and uniqueness of the invariant measure for the two-dimensional stochastic Navier-Stokes equations. A large part of the literature concerns the additive noise case; after revising these models, we consider our recent result, arXiv:2307.03483, with a multiplicative noise.

Recent research is revealing data-sonification as a promising complementary approach to vision, benefiting both data perception and interpretation. We present herakoi, a novel open-source software that uses machine learning to allow real-time image sonification, with a focus on astronomical data. By tracking hand movements via a webcam and mapping them to image coordinates, herakoi translates visual properties into sound, enabling users to "hear" images. Its swift responsiveness allows users to access information in astronomical images with short training, demonstrating high reliability and effectiveness. The software has shown promise in educational and outreach settings, making complex astronomical concepts more engaging and accessible to diverse audiences, including blind and visually impaired individuals. We also discuss future developments, such as the integration of large language and vision models to create a more interactive experience in interpreting astronomical data., Comment: to be published in the proceedings of "Various Innovative Technological Experiences - VITE II" by MemSAIt

In this article we extend a recently introduced kinetic model for consensus-based segmentation of images. In particular, we will interpret the set of pixels of a 2D image as an interacting particle system which evolves in time in view of a consensus-type process obtained by interactions between pixels and external noise. Thanks to a kinetic formulation of the introduced model we derive the large time solution of the model. We will show that the choice of parameters defining the segmentation task can be chosen from a plurality of loss functions characterising the evaluation metrics.

Cytology slides are essential tools in diagnosing and staging cancer, but their analysis is time-consuming and costly. Foundation models have shown great potential to assist in these tasks. In this paper, we explore how existing foundation models can be applied to cytological classification. More particularly, we focus on low-rank adaptation, a parameter-efficient fine-tuning method suited to few-shot learning. We evaluated five foundation models across four cytological classification datasets. Our results demonstrate that fine-tuning the pre-trained backbones with LoRA significantly improves model performance compared to fine-tuning only the classifier head, achieving state-of-the-art results on both simple and complex classification tasks while requiring fewer data samples., Comment: 5 pages, 2 figures

Characterizing domains is essential for models analyzing dynamic environments, as it allows them to adapt to evolving conditions or to hand the task over to backup systems when facing conditions outside their operational domain. Existing solutions typically characterize a domain by solving a regression or classification problem, which limits their applicability as they only provide a limited summarized description of the domain. In this paper, we present a novel approach to domain characterization by characterizing domains as probability distributions. Particularly, we develop a method to predict the likelihood of different weather conditions from images captured by vehicle-mounted cameras by estimating distributions of physical parameters using normalizing flows. To validate our proposed approach, we conduct experiments within the context of autonomous vehicles, focusing on predicting the distribution of weather parameters to characterize the operational domain. This domain is characterized by physical parameters (absolute characterization) and arbitrarily predefined domains (relative characterization). Finally, we evaluate whether a system can safely operate in a target domain by comparing it to multiple source domains where safety has already been established. This approach holds significant potential, as accurate weather prediction and effective domain adaptation are crucial for autonomous systems to adjust to dynamic environmental conditions.

Systems where multiple sources at different redshifts are strongly lensed by the same deflector allow one to directly investigate the evolution of the angular diameter distances with redshift, and thus to learn about the geometry of the Universe. We present measurements of the values of the total matter density, $\Omega_m$, and of the dark energy equation of state parameter, $w$, through a strong lensing analysis of SDSSJ0100+1818, a group-scale system at $z=0.581$ with five lensed sources, from $z=1.698$ to $4.95$. We use new MUSE data to securely measure the redshift of 65 sources, including the five multiply imaged background sources (lensed into a total of 18 multiple images) and 19 galaxies on the deflector plane (the brightest group galaxy, BGG, and 18 fainter members), all employed to build robust strong lensing models with the software GLEE. We measure $\Omega_m = 0.14^{+0.16}_{-0.09}$ in a flat $\Lambda$ cold dark matter (CDM) model, and $\Omega_m = 0.19^{+0.17}_{-0.10}$ and $w=-1.27_{-0.48}^{+0.43}$ in a flat $w$CDM model. We quantify, through a multi-plane approach, the impact of different sources angularly close in projection on the inferred values of the cosmological parameters. We obtain consistent median values, with uncertainties for only $\Omega_m$ increasing by a factor of 1.5. We accurately measure a total mass of $(1.55 \pm 0.01) \times 10^{13}$ M$_\odot$ within 50 kpc and a stellar over total mass profile decreasing from $45.6^{+8.7}_{-8.3}\%$ at the BGG effective radius to $(6.6\pm 1.1)\%$ at $R\approx 77$ kpc. Our results confirm that SDSSJ0100+1818 is one of the most massive (lens) galaxies known at intermediate redshift and that group-scale systems that act as lenses for $\geq 3$ background sources at different redshifts enable to estimate the values of the cosmological parameters with an accuracy that is competitive with that obtained from lens galaxy clusters., Comment: Accepted for publication in A&A. 14 pages, 10 figures

Semantic communication is a new paradigm that considers the meaning of transmitted information to optimize communication. One possible application is the remote monitoring of a process under communication costs: scheduling updates based on semantic considerations can significantly reduce transmission frequency while maintaining high-quality tracking performance. However, semantic scheduling also opens a timing-based side-channel that an eavesdropper may exploit to obtain information about the state of the remote process, even if the content of updates is perfectly secure. In this work, we study an eavesdropping attack against pull-based semantic scheduling for the tracking of remote Markov processes. We provide a theoretical framework for defining the effectiveness of the attack and of possible countermeasures, as well as a practical heuristic that can provide a balance between the performance gains offered by semantic communication and the information leakage.

Artificial Intelligence (AI) is a key component of 6G networks, as it enables communication and computing services to adapt to end users' requirements and demand patterns. The management of Mobile Edge Computing (MEC) is a meaningful example of AI application: computational resources available at the network edge need to be carefully allocated to users, whose jobs may have different priorities and latency requirements. The research community has developed several AI algorithms to perform this resource allocation, but it has neglected a key aspect: learning is itself a computationally demanding task, and considering free training results in idealized conditions and performance in simulations. In this work, we consider a more realistic case in which the cost of learning is specifically accounted for, presenting a new algorithm to dynamically select when to train a Deep Reinforcement Learning (DRL) agent that allocates resources. Our method is highly general, as it can be directly applied to any scenario involving a training overhead, and it can approach the same performance as an ideal learning agent even under realistic training conditions.

Generalized linear mixed models (GLMMs) are a widely used tool in statistical analysis. The main bottleneck of many computational approaches lies in the inversion of the high dimensional precision matrices associated with the random effects. Such matrices are typically sparse; however, the sparsity pattern resembles a multi partite random graph, which does not lend itself well to default sparse linear algebra techniques. Notably, we show that, for typical GLMMs, the Cholesky factor is dense even when the original precision is sparse. We thus turn to approximate iterative techniques, in particular to the conjugate gradient (CG) method. We combine a detailed analysis of the spectrum of said precision matrices with results from random graph theory to show that CG-based methods applied to high-dimensional GLMMs typically achieve a fixed approximation error with a total cost that scales linearly with the number of parameters and observations. Numerical illustrations with both real and simulated data confirm the theoretical findings, while at the same time illustrating situations, such as nested structures, where CG-based methods struggle.

Star clusters stand at the crossroads between galaxies and single stars. Resolving the formation of star clusters in cosmological simulations represents an ambitious and challenging goal, since modelling their internal properties requires very high resolution. This paper is the third of a series within the SImulating the Environment where Globular clusters Emerged (SIEGE) project, where we conduct zoom-in cosmological simulations with sub-parsec resolution that include the feedback of individual stars, aimed to model the formation of star clusters in high-redshift proto-galaxies. We investigate the role of three fundamental quantities in shaping the intrinsic properties of star clusters, i. e., i) pre-supernova stellar feedback (continuous or instantaneous ejection of mass and energy through stellar winds); ii) star formation efficiency, defined as the fraction of gas converted into stars per freefall time, for which we test 2 different values (epsi_ff=0.1 and 1), and iii) stellar initial mass function (IMF, standard vs top-heavy). All our simulations are run down to z=10.5, which is sufficient for investigating some structural properties of the emerging clumps and clusters. [Abridged] The prescription for a continuous, low-intensity feedback, along with the adoption of epsi_ff=1, produces star clusters with maximum stellar density values up to 10^4 M_sun pc^(-2), in good agreement with the surface density-size relation observed in local young star clusters (YSCs). Therefore, a realistic stellar wind description and a high star formation effiency are the key ingredients that allow us to achieve realistic star clusters characterised by properties comparable to those of local YSCs. In contrast, the other models produce too diffuse clusters, in particular the one with a top-heavy IMF., Comment: 20 pages, 14 pages

We study the main properties of the solution of a Fokker-Planck equation characterized by a variable diffusion coefficient and a polynomial superlinear drift, modeling the formation of consensus in a large interacting system of individuals. The Fokker-Planck equation is derived from the kinetic description of the dynamics of a quantum particle system, and in presence of a high nonlinearity in the drift operator, mimicking the effects of the mass in the alignment forces, allows for steady states similar to a Bose-Einstein condensate. The analysis shows that the regularity of the solution is strongly linked to the degree of nonlinearity in the drift, and that finite-time blow-up of the solution can occur when the degree of nonlinearity is sufficiently high. However, the presence of diffusion prevents the solution from forming condensation after the blow-up time.

We study multiproposal Markov chain Monte Carlo algorithms, such as Multiple-try or generalised Metropolis-Hastings schemes, which have recently received renewed attention due to their amenability to parallel computing. First, we prove that no multiproposal scheme can speed-up convergence relative to the corresponding single proposal scheme by more than a factor of $K$, where $K$ denotes the number of proposals at each iteration. This result applies to arbitrary target distributions and it implies that serial multiproposal implementations are always less efficient than single proposal ones. Secondly, we consider log-concave distributions over Euclidean spaces, proving that, in this case, the speed-up is at most logarithmic in $K$, which implies that even parallel multiproposal implementations are fundamentally limited in the computational gain they can offer. Crucially, our results apply to arbitrary multiproposal schemes and purely rely on the two-step structure of the associated kernels (i.e. first generate $K$ candidate points, then select one among those). Our theoretical findings are validated through numerical simulations., Comment: 19 pages, 1 figure

In this letter, we propose a control scheme for rigid bodies designed to optimise transient behaviors. The search space for the optimal control input is parameterized to yield a passive, specifically lossless, nonlinear feedback controller. As a result, it can be combined with other stabilizing controllers without compromising the stability of the closed-loop system. The controller commands torques generating fictitious gyroscopic effects characteristics of 3D rotational rigid body motions, and as such does not inject nor extract kinetic energy from the system. We validate the controller in simulation using a model predictive control (MPC) scheme, successfully combining stability and performance in a stabilization task with obstacle avoidance constraints.

The logit transform is arguably the most widely-employed link function beyond linear settings. This transformation routinely appears in regression models for binary data and provides, either explicitly or implicitly, a core building-block within state-of-the-art methodologies for both classification and regression. Its widespread use, combined with the lack of analytical solutions for the optimization of general losses involving the logit transform, still motivates active research in computational statistics. Among the directions explored, a central one has focused on the design of tangent lower bounds for logistic log-likelihoods that can be tractably optimized, while providing a tight approximation of these log-likelihoods. Although progress along these lines has led to the development of effective minorize-maximize (MM) algorithms for point estimation and coordinate ascent variational inference schemes for approximate Bayesian inference under several logit models, the overarching focus in the literature has been on tangent quadratic minorizers. In fact, it is still unclear whether tangent lower bounds sharper than quadratic ones can be derived without undermining the tractability of the resulting minorizer. This article addresses such a challenging question through the design and study of a novel piece-wise quadratic lower bound that uniformly improves any tangent quadratic minorizer, including the sharpest ones, while admitting a direct interpretation in terms of the classical generalized lasso problem. As illustrated in a ridge logistic regression, this unique connection facilitates more effective implementations than those provided by available piece-wise bounds, while improving the convergence speed of quadratic ones.

We design and analyze unbiased Markov chain Monte Carlo (MCMC) schemes based on couplings of blocked Gibbs samplers (BGSs), whose total computational costs scale linearly with the number of parameters and data points. Our methodology is designed for and applicable to high-dimensional BGS with conditionally independent blocks, which are often encountered in Bayesian modeling. We provide bounds on the expected number of iterations needed for coalescence for Gaussian targets, which imply that practical two-step coupling strategies achieve coalescence times that match the relaxation times of the original BGS scheme up to a logarithmic factor. To illustrate the practical relevance of our methodology, we apply it to high-dimensional crossed random effect and probabilistic matrix factorization models, for which we develop a novel BGS scheme with improved convergence speed. Our methodology provides unbiased posterior estimates at linear cost (usually requiring only a few BGS iterations for problems with thousands of parameters), matching state-of-the-art procedures for both frequentist and Bayesian estimation of those models., Comment: 48 pages, 10 figures, 1 table

We present a systematic review of some basic results on the derivation of classical epidemiological models from simple agent-based dynamics. The evolution of large populations is coupled with the dynamics of the contact distribution, providing insights into how individual behaviors impact macroscopic epidemiological trends. The resulting set of equations incorporates local characteristics of the operator governing the emergence of a family of contact distributions. To validate the proposed approach, we provide several numerical results based on asymptotic preserving methods, demonstrating their effectiveness in capturing the multi-scale nature of the problem and ensuring robust performance across different regimes.

Large Language Models (LLMs) are rapidly becoming commodity components of larger software systems. This poses natural security and privacy problems: poisoned data retrieved from one component can change the model's behavior and compromise the entire system, including coercing the model to spread confidential data to untrusted components. One promising approach is to tackle this problem at the system level via dynamic information flow (aka taint) tracking. Unfortunately, the traditional approach of propagating the most restrictive input label to the output is too conservative for applications where LLMs operate on inputs retrieved from diverse sources. In this paper, we propose a novel, more permissive approach to propagate information flow labels through LLM queries. The key idea behind our approach is to propagate only the labels of the samples that were influential in generating the model output and to eliminate the labels of unnecessary input. We implement and investigate the effectiveness of two variations of this approach, based on (i) prompt-based retrieval augmentation, and (ii) a $k$-nearest-neighbors language model. We compare these with the baseline of an introspection-based influence estimator that directly asks the language model to predict the output label. The results obtained highlight the superiority of our prompt-based label propagator, which improves the label in more than 85% of the cases in an LLM agent setting. These findings underscore the practicality of permissive label propagation for retrieval augmentation., Comment: 16 pages, 11 figures

The Gibbs sampler (a.k.a. Glauber dynamics and heat-bath algorithm) is a popular Markov Chain Monte Carlo algorithm which iteratively samples from the conditional distributions of a probability measure $\pi$ of interest. Under the assumption that $\pi$ is strongly log-concave, we show that the random scan Gibbs sampler contracts in relative entropy and provide a sharp characterization of the associated contraction rate. Assuming that evaluating conditionals is cheap compared to evaluating the joint density, our results imply that the number of full evaluations of $\pi$ needed for the Gibbs sampler to mix grows linearly with the condition number and is independent of the dimension. If $\pi$ is non-strongly log-concave, the convergence rate in entropy degrades from exponential to polynomial. Our techniques are versatile and extend to Metropolis-within-Gibbs schemes and the Hit-and-Run algorithm. A comparison with gradient-based schemes and the connection with the optimization literature are also discussed.

After the recent COVID-19 outbreaks, it became increasingly evident that individuals' thoughts and beliefs can have a strong impact on disease transmission. It becomes therefore important to understand how information and opinions on protective measures evolve during epidemics. To this end, incorporating the impact of social media is essential to take into account the hierarchical structure of these platforms. In this context, we present a novel approach to take into account the interplay between infectious disease dynamics and socially-structured opinion dynamics. Our work extends a conventional compartmental framework including behavioral attitudes in shaping public opinion and promoting the adoption of protective measures under the influence of different degrees of connectivity. The proposed approach is capable to reproduce the emergence of epidemic waves. Specifically, it provides a clear link between the social influence of highly connected individuals and the epidemic dynamics. Through a heterogeneity of numerical tests we show how this comprehensive framework offers a more nuanced understanding of epidemic dynamics in the context of modern information dissemination and social behavior.

The degree distribution is a key statistical indicator in network theory, often used to understand how information spreads across connected nodes. In this paper, we focus on non-growing networks formed through a rewiring algorithm and develop kinetic Boltzmann-type models to capture the emergence of degree distributions that characterize both preferential attachment networks and random networks. Under a suitable mean-field scaling, these models reduce to a Fokker-Planck-type partial differential equation with an affine diffusion coefficient, that is consistent with a well-established master equation for discrete rewiring processes. We further analyze the convergence to equilibrium for this class of Fokker-Planck equations, demonstrating how different regimes -- ranging from exponential to algebraic rates -- depend on network parameters. Our results provide a unified framework for modeling degree distributions in non-growing networks and offer insights into the long-time behavior of such systems.

The development of vision-language models (VLMs) for histo-pathology has shown promising new usages and zero-shot performances. However, current approaches, which decompose large slides into smaller patches, focus solely on inductive classification, i.e., prediction for each patch is made independently of the other patches in the target test data. We extend the capability of these large models by introducing a transductive approach. By using text-based predictions and affinity relationships among patches, our approach leverages the strong zero-shot capabilities of these new VLMs without any additional labels. Our experiments cover four histopathology datasets and five different VLMs. Operating solely in the embedding space (i.e., in a black-box setting), our approach is highly efficient, processing $10^5$ patches in just a few seconds, and shows significant accuracy improvements over inductive zero-shot classification. Code available at https://github.com/FereshteShakeri/Histo-TransCLIP.

Vision-Language Models for remote sensing have shown promising uses thanks to their extensive pretraining. However, their conventional usage in zero-shot scene classification methods still involves dividing large images into patches and making independent predictions, i.e., inductive inference, thereby limiting their effectiveness by ignoring valuable contextual information. Our approach tackles this issue by utilizing initial predictions based on text prompting and patch affinity relationships from the image encoder to enhance zero-shot capabilities through transductive inference, all without the need for supervision and at a minor computational cost. Experiments on 10 remote sensing datasets with state-of-the-art Vision-Language Models demonstrate significant accuracy improvements over inductive zero-shot classification. Our source code is publicly available on Github: https://github.com/elkhouryk/RS-TransCLIP, Comment: Accepted at ICASSP 2025

MICADO is a first light instrument for the Extremely Large Telescope (ELT), set to start operating later this decade. It will provide diffraction limited imaging, astrometry, high contrast imaging, and long slit spectroscopy at near-infrared wavelengths. During the initial phase operations, adaptive optics (AO) correction will be provided by its own natural guide star wavefront sensor. In its final configuration, that AO system will be retained and complemented by the laser guide star multi-conjugate adaptive optics module MORFEO (formerly known as MAORY). Among many other things, MICADO will study exoplanets, distant galaxies and stars, and investigate black holes, such as Sagittarius A* at the centre of the Milky Way. After their final design phase, most components of MICADO have moved on to the manufacturing and assembly phase. Here we summarize the final design of the instrument and provide an overview about its current manufacturing status and the timeline. Some lessons learned from the final design review process will be presented in order to help future instrumentation projects to cope with the challenges arising from the substantial differences between projects for 8-10m class telescopes (e.g. ESO-VLT) and the next generation Extremely Large Telescopes (e.g. ESO-ELT). Finally, the expected performance will be discussed in the context of the current landscape of astronomical observatories and instruments. For instance, MICADO will have similar sensitivity as the James Webb Space Telescope (JWST), but with six times the spatial resolution., Comment: Proceedings of the SPIE, Volume 13096, id. 1309611 11 pp. (2024)

In recent years, Generative Artificial Intelligence (GenAI) has undergone a profound transformation in addressing intricate tasks involving diverse modalities such as textual, auditory, visual, and pictorial generation. Within this spectrum, text-to-image (TTI) models have emerged as a formidable approach to generating varied and aesthetically appealing compositions, spanning applications from artistic creation to realistic facial synthesis, and demonstrating significant advancements in computer vision, image processing, and multimodal tasks. The advent of Latent Diffusion Models (LDMs) signifies a paradigm shift in the domain of AI capabilities. This article delves into the feasibility of employing the Stable Diffusion LDM to illustrate literary works. For this exploration, seven classic Brazilian books have been selected as case studies. The objective is to ascertain the practicality of this endeavor and to evaluate the potential of Stable Diffusion in producing illustrations that augment and enrich the reader's experience. We will outline the beneficial aspects, such as the capacity to generate distinctive and contextually pertinent images, as well as the drawbacks, including any shortcomings in faithfully capturing the essence of intricate literary depictions. Through this study, we aim to provide a comprehensive assessment of the viability and efficacy of utilizing AI-generated illustrations in literary contexts, elucidating both the prospects and challenges encountered in this pioneering application of technology., Comment: 7 pages, 2 figures

We present a detailed JWST/NIRSpec and NIRCam analysis of a gravitationally-lensed galaxy ($\rm \mu=17-21$) at redshift 6.14 magnified by the Hubble Frontier Field galaxy cluster MACS J0416. The target galaxy is overall a typical compact and UV-faint ($\rm M_{UV}=-17.8$) Lyman-$\alpha$ emitter; yet, the large magnification allows the detailed characterisation of structures on sub-galactic (down to few parsec) scales. Prominent optical $\rm H\alpha$, $\rm H\beta$ and [OIII]$\lambda\lambda4959,5007$ lines are spatially resolved with the high spectral resolution grating (G395H, R~2700), with large equivalent widths, EW($\rm H\beta$+[OIII])$\gtrsim1000$ \AA, and elevated ionising photon production efficiencies $\rm log(\xi_{ion}/erg^{-1}Hz)=25.2-25.7$. NIRCam deep imaging reveals the presence of compact rest-UV bright regions along with individual star clusters of sizes $\rm R_{eff}=3-8~pc$ and masses $\rm M\sim2\cdot10^5-5\cdot10^{6}~M_\odot$ These clusters are characterised by steep UV slopes, $\rm\beta_{UV}\lesssim-2.5$, in some cases associated with a dearth of line emission, indicating possible leaking of the ionising radiation, as also supported by a Lyman-$\rm \alpha$ emission peaking at $\rm \sim100~km~s^{-1}$ from the systemic redshift. While the entire system is characterised by low-metallicity, $\sim0.1~Z_\odot$, the NIRSpec-IFU map also reveals the presence of a low-luminosity, metal-poor region with $\rm Z\lesssim2\%~Z_\odot$, barely detected in NIRCam imaging; this region is displaced by $\rm >200~pc$ from one of the UV brightest structures of the system, and it would have been too faint to detect if not for the large magnification of the system., Comment: 19 pages (11 figures, 2 tables) + appendix (3 pages, 4 figures, 1 table). Submitted to A&A; comments are welcome

We report the serendipitous discovery of a faint (M_UV > -12.2), low-metallicity (Z ~ 0.02 Zsun), ionizing source (dubbed T2c) with a spectroscopic redshift of z=6.146. T2c is part of a larger structure amplified by the Hubble Frontier Field galaxy cluster MACSJ0416, and was observed with JWST/NIRSpec IFU. Stacking the short-wavelength NIRCam data reveals no stellar continuum detection down to a magnitude limit of m_UV ~ 31.0 (3 sigma). However, prominent Hb, [OIII]4959,5007, and Ha emissions are detected, with equivalent widths exceeding 200A, 800A, and 1300A (3 sigma), respectively. The corresponding intrinsic (magnification-corrected x23 +/- 3) ultraviolet and optical rest-frame magnitudes exceed 34.4 and 33.9 (corresponding to M_uv and M_opt fainter than -12.2 and -12.8, at lambda_rest ~ 2000A and ~5000A, respectively), suggesting a stellar mass lower than a few 10^4 Msun under an instantaneous burst scenario. The inferred ionizing photon production efficiency (xi_ion) is high, xi_ion >~ 26.08(25.86) 3(5)sigma, assuming no dust attenuation and no Lyman continuum leakage, indicating the presence of massive stars despite the low mass of the object. The very poor sampling of the initial mass function at such low mass star-forming complex suggests that the formation of very massive stars might be favored in very low metallicity environments. T2c is surrounded by Balmer and weak oxygen emission on a spatial scale of a few hundred parsecs after correcting for lensing effects. This system resembles an HII region potentially powered by currently undetected, extremely efficient, low-metallicity star complexes or clusters. We propose that massive O-type stars populate this low-mass and metallicity high-redshift satellites, likely caught in an early and short formation phase, contributing to the ionization of the surrounding medium., Comment: 9 pages, 5 figures, 1 table. Submitted to A&A. Comments are welcome

Urban parks provide significant health benefits by offering spaces and facilities for various recreational and leisure activities. However, the capacity of specific park spaces and elements to foster health remains underexamined. Traditional studies have focused on parks' size, greenery, and accessibility, often overlooking their ability to facilitate specific health-promoting activities. To address this gap, we propose a taxonomy consisting of six categories of health-promoting activities in parks: physical, mind-body, nature appreciation, environmental, social, and cultural. We estimate the capacity of parks in 35 global cities to promote health by establishing a lexicon linking park spaces and elements with specific health-promoting activities from our taxonomy. Using this lexicon, we collected data on elements and spaces in all parks in 35 cities from OpenStreetMap. Our analysis covers 23,477 parks with a total of 827,038 elements and spaces. By first comparing similarly sized parks across cities, we found that North American parks offer more spaces for physical activities, while European parks focus more on nature appreciation. Second, by scoring parks based on both elements and spaces, we investigated the variability in their health-promoting potential. We found the most uniform provision across parks for physical activities and the highest disparities regarding social activities. Additionally, parks offering a variety of activities are usually located in city centers, while offerings diminish in parks towards the suburbs. Lastly, we identified significant inequalities in park standards across cities, regardless of their continental location: Tokyo and Paris offer the most uniform park standards, while Copenhagen and Rio de Janeiro exhibit the most pronounced disparities. Our study provides insights for making urban parks more equitable, engaging, and health-promoting., Comment: 29 pages main paper, 10 pages appendix

The highest scientific return, for adaptive optics (AO) observations, is achieved with a reliable reconstruction of the PSF. This is especially true for MICADO@ELT. In this presentation, we will focus on extending the MICADO PSF reconstruction (PSF-R) method to the off-axis case. Specifically, a novel approach based on temporal-based tomography of AO telemetry data has been recently implemented. Results from the PSF-R of both simulated and real data show that, at half isoplanatic angle distances, a precision of about 10-15% is achievable in both Strehl ratio and full-width at half maximum, paving the way to extend the MICADO PSF-R tool also to the multi-conjugated AO case., Comment: 8 pages, 4 figures. Proceeding of SPIE Astronomical Telescopes + Instrumentation 2024, Adaptive Optics Systems IX

MICADO is the first-light camera of the ESO ELT, allowing NIR imaging and long-slit spectroscopy assisted by adaptive optics. MICADO is now entering its construction phase, and the software for data reduction is reaching an adequate maturity level. The PSF Reconstruction (PSF-R) of MICADO is a software tool for the blind derivation of the PSF, only using adaptive optics telemetry data. An update of the status of the PSF-R service is provided here. The PSF-R prototype has been tested on ERIS@VLT data in order to check the reconstruction of on- and off-axis PSFs. The on-axis PSF-R is accurate at a few percent level on Strehl, FWHM, Encircled Energy, and half light radius, while for the off-axis case the match is within 10-15 percent at a distance of half isoplanatic angle. The first version of the workflow for the PSF-R pipeline has been developed and verified using the latest release of the ESO data processing system. A set of simulations has been implemented on the morphological analysis of distant galaxies, showing that the accuracy of the PSF-R matches the goals needed to study their morphology. In summary, the PSF-R team is on the right track towards the ELT first light., Comment: 5 pages, 3 figures, Proceedings for the SPIE Astronomical Telescopes and Instrumentation 2024, Adaptive Optics Systems IX, Paper No.13097-234

The gas in the interstellar medium (ISM) of galaxies is supersonically turbulent. Measurements of turbulence typically rely on cold gas emission lines for low-z galaxies and warm ionized gas observations for z>0 galaxies. Studies of warm gas kinematics at z>0 conclude that the turbulence strongly evolves as a function of redshift, due to the increasing impact of gas accretion and mergers in the early Universe. However, recent findings suggest potential biases in turbulence measurements derived from ionized gas at high-z, impacting our understanding of turbulence origin, ISM physics and disk formation. We investigate the evolution of turbulence using velocity dispersion ($\sigma$) measurements from cold gas tracers (i.e., CO, [CI], [CII]) derived from a sample of 57 galaxy disks spanning the redshift range z=0-5. This sample consists of main-sequence and starburst galaxies with stellar masses $\gtrsim 10^{10} M_{\odot}$. The comparison with current H$\alpha$ kinematic observations and existing models demonstrates that the velocity dispersion inferred from cold gas tracers differ by a factor of $\approx 3$ from those obtained using emission lines tracing warm gas. We show that stellar feedback is the main driver of turbulence measured from cold gas tracers. This is fundamentally different from the conclusions of studies based on warm gas, which had to consider additional turbulence drivers to explain the high values of $\sigma$. We present a model predicting the redshift evolution of turbulence in galaxy disks, attributing the increase of $\sigma$ with redshift to the higher energy injected by supernovae due to the elevated star-formation rate in high-z galaxies. This supernova-driven model suggests that turbulence is lower in galaxies with lower stellar mass compared to those with higher stellar mass. Additionally, it forecasts the evolution of $\sigma$ in Milky-Way like progenitors., Comment: Accepted for publication in A&A. The abstract has been modified to comply with arXiv's character limit

A pipeline to evaluate the evolution of viral dynamics based on a new model-driven approach has been developed in the present study. The proposed methods exploit real data and the multiscale structure of the infection dynamics to provide robust predictions of the epidemic dynamics. We focus on viral load kinetics whose dynamical features are typically available in the symptomatic stage of the infection. Hence, the epidemiological evolution is obtained by relying on a compartmental approach characterized by a varying infection rate to estimate early-stage viral load dynamics, of which few data are available. We test the proposed approach with real data of SARS-CoV-2 viral load kinetics collected from patients living in an Italian province. The considered database refers to early-phase infections, whose viral load kinetics are not affected by mass vaccination policies in Italy. Our contribution is devoted to provide an effective computational pipeline to evaluate in real time the evolution of infectivity. Comprehending the factors influencing the in-host viral dynamics represents a fundamental tool to provide robust public health strategies. This pilot study could be implemented in further investigations involving other respiratory viruses, to better clarify the process of viral dynamics as a preparatory action for future pandemics.

We consider the nonlinear Schr\"odinger equation on the $d$-dimensional torus $\mathbb T^d$, with the nonlinearity of polynomial type $|u|^{2\sigma}u$. For any $\sigma \in \mathbb N$ and $s>\frac d2$ we prove that adding to this equation a suitable stochastic forcing term there exists a unique global solution for any initial data in $H^s(\mathbb T^d)$. The effect of the noise is to prevent blow-up in finite time, differently from the deterministic setting. Moreover we prove existence of invariant measures and their uniqueness under more restrictive assumptions on the noise term., Comment: 39 pages

Using a port-Hamiltonian formalism, we show the qualitative and quantitative effect of safety-critical control implemented with control barrier functions (CBFs) on the power balance of controlled physical systems. The presented results will provide novel tools to design CBFs inducing desired energetic behaviors of the closed-loop system, including nontrivial damping injection effects and non-passive control actions, effectively injecting energy in the system in a controlled manner. Simulations validate the stated results.

Large language model systems face important security risks from maliciously crafted messages that aim to overwrite the system's original instructions or leak private data. To study this problem, we organized a capture-the-flag competition at IEEE SaTML 2024, where the flag is a secret string in the LLM system prompt. The competition was organized in two phases. In the first phase, teams developed defenses to prevent the model from leaking the secret. During the second phase, teams were challenged to extract the secrets hidden for defenses proposed by the other teams. This report summarizes the main insights from the competition. Notably, we found that all defenses were bypassed at least once, highlighting the difficulty of designing a successful defense and the necessity for additional research to protect LLM systems. To foster future research in this direction, we compiled a dataset with over 137k multi-turn attack chats and open-sourced the platform.

The Coordinate Ascent Variational Inference scheme is a popular algorithm used to compute the mean-field approximation of a probability distribution of interest. We analyze its random scan version, under log-concavity assumptions on the target density. Our approach builds on the recent work of M. Arnese and D. Lacker, \emph{Convergence of coordinate ascent variational inference for log-concave measures via optimal transport} [arXiv:2404.08792] which studies the deterministic scan version of the algorithm, phrasing it as a block-coordinate descent algorithm in the space of probability distributions endowed with the geometry of optimal transport. We obtain tight rates for the random scan version, which imply that the total number of factor updates required to converge scales linearly with the condition number and the number of blocks of the target distribution. By contrast, available bounds for the deterministic scan case scale quadratically in the same quantities, which is analogue to what happens for optimization of convex functions in Euclidean spaces.

Transduction is a powerful paradigm that leverages the structure of unlabeled data to boost predictive accuracy. We present TransCLIP, a novel and computationally efficient transductive approach designed for Vision-Language Models (VLMs). TransCLIP is applicable as a plug-and-play module on top of popular inductive zero- and few-shot models, consistently improving their performances. Our new objective function can be viewed as a regularized maximum-likelihood estimation, constrained by a KL divergence penalty that integrates the text-encoder knowledge and guides the transductive learning process. We further derive an iterative Block Majorize-Minimize (BMM) procedure for optimizing our objective, with guaranteed convergence and decoupled sample-assignment updates, yielding computationally efficient transduction for large-scale datasets. We report comprehensive evaluations, comparisons, and ablation studies that demonstrate: (i) Transduction can greatly enhance the generalization capabilities of inductive pretrained zero- and few-shot VLMs; (ii) TransCLIP substantially outperforms standard transductive few-shot learning methods relying solely on vision features, notably due to the KL-based language constraint.

Science, Technology, Engineering, and Mathematics classes are mainly taught using visual supports. However, the advancement of technology and the increasing eforts to equip schools with digital instrumentation have opened up the possibility of exploring new teaching avenues, such as sonifcation. We explored the efcacy of sonifcation in education using a novel interactive tool, Edukoi, in the context of astronomy, which is predominantly disseminated through spectacular images, animations, and visuals. Edukoi is a motion-sensing sonifcation tool that converts images to sound in real-time for educational applications. Our study, conducted with nearly 150 middle-school students, included a preliminary questionnaire investigating the perception, engagement, and motivation of students towards science; two sessions dedicated to testing Edukoi and assessing the potentiality of the software for the recognition of the colour and the shape of real and sketchy images; and a fnal second administration of the questionnaire to capture a possible benefcial efect of the use of the tool in the engagement towards science. Results showed the efectiveness of Edukoi in colour recognition and reasonable efcacy in shape identifcation. Although the questionnaire did not reveal an increment in science engagement over the time of the study, oral feedback from the students was positive. Edukoi presents a possible alternative teaching aid, potentially benefting diverse learners, including the visually impaired. Further developments of the software are needed to enhance its efectiveness in conveying more complex features such as composite colours or shapes.

Recent progress in the few-shot adaptation of Vision-Language Models (VLMs) has further pushed their generalization capabilities, at the expense of just a few labeled samples within the target downstream task. However, this promising, already quite abundant few-shot literature has focused principally on prompt learning and, to a lesser extent, on adapters, overlooking the recent advances in Parameter-Efficient Fine-Tuning (PEFT). Furthermore, existing few-shot learning methods for VLMs often rely on heavy training procedures and/or carefully chosen, task-specific hyper-parameters, which might impede their applicability. In response, we introduce Low-Rank Adaptation (LoRA) in few-shot learning for VLMs, and show its potential on 11 datasets, in comparison to current state-of-the-art prompt- and adapter-based approaches. Surprisingly, our simple CLIP-LoRA method exhibits substantial improvements, while reducing the training times and keeping the same hyper-parameters in all the target tasks, i.e., across all the datasets and numbers of shots. Certainly, our surprising results do not dismiss the potential of prompt-learning and adapter-based research. However, we believe that our strong baseline could be used to evaluate progress in these emergent subjects in few-shot VLMs.

In this paper, we consider a scenario with one UAV equipped with a ULA, which sends combined information and sensing signals to communicate with multiple GBS and, at the same time, senses potential targets placed within an interested area on the ground. We aim to jointly design the transmit beamforming with the GBS association to optimize communication performance while ensuring high sensing accuracy. We propose a predictive beamforming framework based on a dual DNN solution to solve the formulated nonconvex optimization problem. A first DNN is trained to produce the required beamforming matrix for any point of the UAV flying area in a reduced time compared to state-of-the-art beamforming optimizers. A second DNN is trained to learn the optimal mapping from the input features, power, and EIRP constraints to the GBS association decision. Finally, we provide an extensive simulation analysis to corroborate the proposed approach and show the benefits of EIRP, SINR performance and computational speed.

Pragmatic or goal-oriented communication can optimize communication decisions beyond the reliable transmission of data, instead aiming at directly affecting application performance with the minimum channel utilization. In this paper, we develop a general theoretical framework for the remote control of finite-state Markov processes, using pragmatic communication over a costly zero-delay communication channel. To that end, we model a cyber-physical system composed of an encoder, which observes and transmits the states of a process in real-time, and a decoder, which receives that information and controls the behavior of the process. The encoder and the decoder should cooperatively optimize the trade-off between the control performance (i.e., reward) and the communication cost (i.e., channel use). This scenario underscores a pragmatic (i.e., goal-oriented) communication problem, where the purpose is to convey only the data that is most valuable for the underlying task, taking into account the state of the decoder (hence, the pragmatic aspect). We investigate two different decision-making architectures: in pull-based remote control, the decoder is the only decision-maker, while in push-based remote control, the encoder and the decoder constitute two independent decision-makers, leading to a multi-agent scenario. We propose three algorithms to optimize our system (i.e., design the encoder and the decoder policies), discuss the optimality guarantees ofs the algorithms, and shed light on their computational complexity and fundamental limits., Comment: Submitted for publication in the IEEE Journal on Selected Areas in Communications

Stochastic Gradient (SG) Markov Chain Monte Carlo algorithms (MCMC) are popular algorithms for Bayesian sampling in the presence of large datasets. However, they come with little theoretical guarantees and assessing their empirical performances is non-trivial. In such context, it is crucial to develop algorithms that are robust to the choice of hyperparameters and to gradients heterogeneity since, in practice, both the choice of step-size and behaviour of target gradients induce hard-to-control biases in the invariant distribution. In this work we introduce the stochastic gradient Barker dynamics (SGBD) algorithm, extending the recently developed Barker MCMC scheme, a robust alternative to Langevin-based sampling algorithms, to the stochastic gradient framework. We characterize the impact of stochastic gradients on the Barker transition mechanism and develop a bias-corrected version that, under suitable assumptions, eliminates the error due to the gradient noise in the proposal. We illustrate the performance on a number of high-dimensional examples, showing that SGBD is more robust to hyperparameter tuning and to irregular behavior of the target gradients compared to the popular stochastic gradient Langevin dynamics algorithm.

In this work, we define a class of models to understand the impact of population size on opinion formation dynamics, a phenomenon usually related to group conformity. To this end, we introduce a new kinetic model in which the interaction frequency is weighted by the kinetic density. In the quasi-invariant regime, this model reduces to a Kaniadakis-Quarati-type equation with nonlinear drift, originally introduced for the dynamics of bosons in a spatially homogeneous setting. From the obtained PDE for the evolution of the opinion density, we determine the regime of parameters for which a critical mass exists and triggers blow-up of the solution. Therefore, the model is capable of describing strong conformity phenomena in cases where the total density of individuals holding a given opinion exceeds a fixed critical size. In the final part, several numerical experiments demonstrate the features of the introduced class of models and the related consensus effects.

The development of large vision-language models, notably CLIP, has catalyzed research into effective adaptation techniques, with a particular focus on soft prompt tuning. Conjointly, test-time augmentation, which utilizes multiple augmented views of a single image to enhance zero-shot generalization, is emerging as a significant area of interest. This has predominantly directed research efforts toward test-time prompt tuning. In contrast, we introduce a robust MeanShift for Test-time Augmentation (MTA), which surpasses prompt-based methods without requiring this intensive training procedure. This positions MTA as an ideal solution for both standalone and API-based applications. Additionally, our method does not rely on ad hoc rules (e.g., confidence threshold) used in some previous test-time augmentation techniques to filter the augmented views. Instead, MTA incorporates a quality assessment variable for each view directly into its optimization process, termed as the inlierness score. This score is jointly optimized with a density mode seeking process, leading to an efficient training- and hyperparameter-free approach. We extensively benchmark our method on 15 datasets and demonstrate MTA's superiority and computational efficiency. Deployed easily as plug-and-play module on top of zero-shot models and state-of-the-art few-shot methods, MTA shows systematic and consistent improvements.