Search

The unique experimental connection to the QCD energy-momentum tensor offered by generalised parton distributions has been strongly highlighted in the past few years with attempts to extract the pressure and shear forces distributions within the nucleon. If, in principle, this can be performed in a model independent way from experimental data, in practice, the current limited precision and kinematic coverage make such an extraction very challenging. Moreover, the limitation to a leading-order description in the strong coupling of the data has provided only an indirect and weakly sensitive access to gluon degrees of freedoms, solely through their mixing to quarks via evolution. In this paper we address this issue by pro viding a next-to-leading order formalism allowing a reanalysis of global fits with genuine gluonic degrees of freedom. In addition, we provide an estimate of the reduction in uncertainty that could stem from the extended kinematic range relevant for the future Electron Ion Collider. Finally, we stress the connection between the analysis of the dispersion relation in terms of generalised parton distributions and the deconvolution problem., Comment: 21 pages, 14 figures

With the expanding integration of infrared instruments in astronomical missions, accurate per-pixel flux estimation for near-infrared hybrid detectors has become critical to the success of these missions. Based on CPPM's involvement in both SVOM/Colibri and Euclid missions, this study introduces universally applicable methods and framework for characterizing IR hybrid detectors and decorrelating their intrinsic properties. The characterization framework, applied to the ALFA detector and \Euclid's H2RG, not only validates the proposed methods but also points out subtle behaviors inherent to each detector.

Track finding can be considered as a complex optimization problem initially introduced in particle physics involving the reconstruction of particle trajectories. A track is typically composed of several consecutive segments (track segments) that resembles a smooth curve without bifurcations. In this paper various modeling approaches are explored in order to assess both their impact and their effectiveness in solving them using quantum and classical methods. We present implementations of three classical models using CPLEX, two quantum models running on actual D-Wave quantum computers, and one quantum model on a D-Wave simulator. To facilitate a fair comparative study and encourage future research in this area, we introduce a new set of benchmark instances, categorized into small, medium, and large scales. Our evaluation of these methods on the benchmark instances indicates that D-Wave methods offer an excellent balance between computation time and result quality, outperforming CPLEX in numerous cases.

Context. Transient sky astronomy is entering a new era with the advent of the SVOM mission (Space Variable Objects Monitor), which was successfully launched on the 26th of June, 2024. The primary goal of SVOM is to monitor the hard X-ray sky searching for gamma-ray bursts (GRBs). On top of its on-board follow-up capabilities, SVOM will be backed by its ground segment composed of several facilities, of which the near-infrared imager CAGIRE. Mounted on the robotic telescope COLIBRI, it will be a unique instrument, able to perform fast follow-up of GRB afterglows in J and H bands, an ideal combination to catch high-redshift (z>6) and/or obscured GRBs. Aims. This paper aims at estimating the performances of CAGIRE for GRB near-infrared afterglow detection based on the characteristics of the detector and the specificities of the COLIBRI telescope. Quickly fading GRB afterglows pose challenges that should be addressed by adapting observing strategies to the capabilities of CAGIRE. Methods. We use an end-to-end image simulator to produce realistic CAGIRE images, taking into account results from the characterization of the ALFA detector used by CAGIRE. We implemented a GRB afterglow generator that simulates infrared lightcurves and spectra based on published observation of distant GRBs (z>6). Results. We retrieved the photometry of 9 GRB afterglows in various scenarios covered by CAGIRE. Catching afterglows as early as two minutes after burst allows the identification of a nIR counterpart in the brightest 4 events. When artificially redshifted even further away, these events remain detectable by CAGIRE up to z=9.6 in J band, and z=13.3 in H band, indicating the potential of CAGIRE to be a pioneer in the identification of the most distant GRBs to date., Comment: 8 pages, 5 figures, 3 tables, accepted for publication in A&A

The rising popularity of computational workflows is driven by the need for repetitive and scalable data processing, sharing of processing know-how, and transparent methods. As both combined records of analysis and descriptions of processing steps, workflows should be reproducible, reusable, adaptable, and available. Workflow sharing presents opportunities to reduce unnecessary reinvention, promote reuse, increase access to best practice analyses for non-experts, and increase productivity. In reality, workflows are scattered and difficult to find, in part due to the diversity of available workflow engines and ecosystems, and because workflow sharing is not yet part of research practice. WorkflowHub provides a unified registry for all computational workflows that links to community repositories, and supports both the workflow lifecycle and making workflows findable, accessible, interoperable, and reusable (FAIR). By interoperating with diverse platforms, services, and external registries, WorkflowHub adds value by supporting workflow sharing, explicitly assigning credit, enhancing FAIRness, and promoting workflows as scholarly artefacts. The registry has a global reach, with hundreds of research organisations involved, and more than 700 workflows registered., Comment: 30 pages, 4 figures

In this paper, we report the long-range and wireless interrogation of multiple chipless sensors from the isoline processing of three-dimensional polarimetric radar images. A Frequency-Modulated Continuous-Wave Radar operating at 24 GHz is used for the indoor interrogation of four sensors in the basement of a Laboratory. In such cluttered environment, the proposed radar image processing based on isolines computation allows the wireless measurement range of sensors up to 5.8m.

We address the problem of extending a pretrained large language model to a new domain that was not seen at training time, like adding a language for which the original model has seen no or little training data. Popular solutions like fine-tuning or low-rank adaptation are successful at domain adaptation, but formally they do not add any extra capacity and degrade the performance in the original domain. Our paper analyzes this extension problem under three angles: data, architecture and training procedure, which are advantageously considered jointly. In particular, we improve adapters and make it possible to learn an entire new language while ensuring that the output of the neural network is almost unchanged in the original domain. For this purpose, we modify the new residual blocks in a way that leads each new residual block to output near-zeros in the original domain. This solution of neutral residues, which borrows architectural components from mixture of experts, is effective: with only 20% extra learnable weights compared to an original model trained on English, we get results that are significantly better than concurrent approaches (fine-tuning, low-rank or vanilla adapters) in terms of the trade-off between learning a new language and not forgetting English.

This paper presents new methods that have the potential to improve privacy process efficiency with LLM and RAG. To reduce hallucination, we continually pre-train the base LLM model with a privacy-specific knowledge base and then augment it with a semantic RAG layer. Our evaluations demonstrate that this approach enhances the model performance (as much as doubled metrics compared to out-of-box LLM) in handling privacy-related queries, by grounding responses with factual information which reduces inaccuracies.

Foundation models such as the recently introduced Segment Anything Model (SAM) have achieved remarkable results in image segmentation tasks. However, these models typically require user interaction through handcrafted prompts such as bounding boxes, which limits their deployment to downstream tasks. Adapting these models to a specific task with fully labeled data also demands expensive prior user interaction to obtain ground-truth annotations. This work proposes to replace conditioning on input prompts with a lightweight module that directly learns a prompt embedding from the image embedding, both of which are subsequently used by the foundation model to output a segmentation mask. Our foundation models with learnable prompts can automatically segment any specific region by 1) modifying the input through a prompt embedding predicted by a simple module, and 2) using weak labels (tight bounding boxes) and few-shot supervision (10 samples). Our approach is validated on MedSAM, a version of SAM fine-tuned for medical images, with results on three medical datasets in MR and ultrasound imaging. Our code is available on https://github.com/Minimel/MedSAMWeakFewShotPromptAutomation., Comment: Accepted to MICCAI-MedAGI 2024 (LNCS Proceedings, Volume 15184), 10 pages

The non-Hermitian skin effect is an intriguing physical phenomenon, in which all eigen-modes of a non-Hermitian lattice become localized at boundary regions. While such an exotic behavior has been demonstrated in various physical platforms, these realizations have been so far restricted to the linear regime. Here, we explore the interplay between nonlinearity and the non-Hermitian skin effect which can lead to extraordinary amplitude-driven skin localization dynamics that are unfeasible in linear systems. We show that nonlinear non-Hermitian systems can exhibit new kinds of non-Hermitian skin modes, capable of concentrating the energy of a source at any point of space, depending on the power level. This allows the realization of an energy guide that adapts spontaneously to the amount of energy previously stored in the system: it transfers the energy stored in bulk modes to a point only after a sufficient level of energy has been accumulated. Our theoretical, numerical and experimental findings open up exciting paths for a new generation of non-reciprocal systems, in which nonlinearities serve as a strategic tuning knob to manipulate the guiding process., Comment: 5 pages, 3 figures

We present a novel approach for benchmarking and validating quantitative phase tomography (QPT) systems using three-dimensional microphantoms. These microphantoms, crafted from biological and imaging data, replicate the optical and structural properties of multicellular biological samples. Their fabrication featuring refractive index modulation at sub-micrometer details is enabled by two-photon polymerization. We showcase the effectiveness of our technique via a round-robin test of healthy and tumoral liver organoid phantoms across three different QPT systems. This test reveals sample- and system-dependent errors in measuring dry mass and morphology. Tailored microphantoms establish the gold standard for the optimization of hardware setups, reconstruction strategies and error assessment, paving the way for novel non-invasive, label-free measurements of 3D biological samples.

End-to-end neural diarization models have usually relied on a multilabel-classification formulation of the speaker diarization problem. Recently, we proposed a powerset multiclass formulation that has beaten the state-of-the-art on multiple datasets. In this paper, we propose to study the calibration of a powerset speaker diarization model, and explore some of its uses. We study the calibration in-domain, as well as out-of-domain, and explore the data in low-confidence regions. The reliability of model confidence is then tested in practice: we use the confidence of the pretrained model to selectively create training and validation subsets out of unannotated data, and compare this to random selection. We find that top-label confidence can be used to reliably predict high-error regions. Moreover, training on low-confidence regions provides a better calibrated model, and validating on low-confidence regions can be more annotation-efficient than random regions.

Plant morphological traits, their observable characteristics, are fundamental to understand the role played by each species within their ecosystem. However, compiling trait information for even a moderate number of species is a demanding task that may take experts years to accomplish. At the same time, massive amounts of information about species descriptions is available online in the form of text, although the lack of structure makes this source of data impossible to use at scale. To overcome this, we propose to leverage recent advances in large language models (LLMs) and devise a mechanism for gathering and processing information on plant traits in the form of unstructured textual descriptions, without manual curation. We evaluate our approach by automatically replicating three manually created species-trait matrices. Our method managed to find values for over half of all species-trait pairs, with an F1-score of over 75%. Our results suggest that large-scale creation of structured trait databases from unstructured online text is currently feasible thanks to the information extraction capabilities of LLMs, being limited by the availability of textual descriptions covering all the traits of interest.

In a randomised clinical trial, when the result of the primary endpoint shows a significant benefit, the secondary endpoints are scrutinised to identify additional effects of the treatment. However, this approach entails a risk of concluding that there is a benefit for one of these endpoints when such benefit does not exist (inflation of type I error risk). There are mainly two methods used to control the risk of drawing erroneous conclusions for secondary endpoints. The first method consists of distributing the risk over several co-primary endpoints, so as to maintain an overall risk of 5%. The second is the hierarchical test procedure, which consists of first establishing a hierarchy of the endpoints, then evaluating each endpoint in succession according to this hierarchy while the endpoints continue to show statistical significance. This simple method makes it possible to show the additional advantages of treatments and to identify the factors that differentiate them.

Correspondence analysis, multiple correspondence analysis and their discriminant counterparts (i.e., discriminant simple correspondence analysis and discriminant multiple correspondence analysis) are methods of choice for analyzing multivariate categorical data. In these methods, variables are integrated into optimal components computed as linear combinations whose weights are obtained from a generalized singular value decomposition (GSVD) that integrates specific metric constraints on the rows and columns of the original data matrix. The weights of the linear combinations are, in turn, used to interpret the components, and this interpretation is facilitated when components are 1) pairwise orthogonal and 2) when the values of the weights are either large or small but not intermediate-a pattern called a simple or a sparse structure. To obtain such simple configurations, the optimization problem solved by the GSVD is extended to include new constraints that implement component orthogonality and sparse weights. Because multiple correspondence analysis represents qualitative variables by a set of binary variables, an additional group constraint is added to the optimization problem in order to sparsify the whole set representing one qualitative variable. This new algorithm-called group-sparse GSVD (gsGSVD)-integrates these constraints via an iterative projection scheme onto the intersection of subspaces where each subspace implements a specific constraint. In this paper, we expose this new algorithm and show how it can be adapted to the sparsification of simple and multiple correspondence analysis, and illustrate its applications with the analysis of four different data sets-each illustrating the sparsification of a particular CA-based analysis.

We introduce Moshi, a speech-text foundation model and full-duplex spoken dialogue framework. Current systems for spoken dialogue rely on pipelines of independent components, namely voice activity detection, speech recognition, textual dialogue and text-to-speech. Such frameworks cannot emulate the experience of real conversations. First, their complexity induces a latency of several seconds between interactions. Second, text being the intermediate modality for dialogue, non-linguistic information that modifies meaning -- such as emotion or non-speech sounds -- is lost in the interaction. Finally, they rely on a segmentation into speaker turns, which does not take into account overlapping speech, interruptions and interjections. Moshi solves these independent issues altogether by casting spoken dialogue as speech-to-speech generation. Starting from a text language model backbone, Moshi generates speech as tokens from the residual quantizer of a neural audio codec, while modeling separately its own speech and that of the user into parallel streams. This allows for the removal of explicit speaker turns, and the modeling of arbitrary conversational dynamics. We moreover extend the hierarchical semantic-to-acoustic token generation of previous work to first predict time-aligned text tokens as a prefix to audio tokens. Not only this "Inner Monologue" method significantly improves the linguistic quality of generated speech, but we also illustrate how it can provide streaming speech recognition and text-to-speech. Our resulting model is the first real-time full-duplex spoken large language model, with a theoretical latency of 160ms, 200ms in practice, and is available at https://github.com/kyutai-labs/moshi.

Hidden Markov Models (HMM) model a sequence of observations that are dependent on a hidden (or latent) state that follow a Markov chain. These models are widely used in diverse fields including ecology, speech recognition, and genetics.Parameter estimation in HMM is typically performed using the Baum-Welch algorithm, a special case of the Expectation-Maximisation (EM) algorithm. While this method guarantee the convergence to a local maximum, its convergence rates is usually slow.Alternative methods, such as the direct maximisation of the likelihood using quasi-Newton methods (such as L-BFGS-B) can offer faster convergence but can be more complicated to implement due to challenges to deal with the presence of bounds on the space of parameters.We propose a novel hybrid algorithm, QNEM, that combines the Baum-Welch and the quasi-Newton algorithms. QNEM aims to leverage the strength of both algorithms by switching from one method to the other based on the convexity of the likelihood function.We conducted a comparative analysis between QNEM, the Baum-Welch algorithm, an EM acceleration algorithm called SQUAREM (Varadhan, 2008, Scand J Statist), and the L-BFGS-B quasi-Newton method by applying these algorithms to four examples built on different models. We estimated the parameters of each model using the different algorithms and evaluated their performances.Our results show that the best-performing algorithm depends on the model considered. QNEM performs well overall, always being faster or equivalent to L-BFGS-B. The Baum-Welch and SQUAREM algorithms are faster than the quasi-Newton and QNEM algorithms in certain scenarios with multiple optimum. In conclusion, QNEM offers a promising alternative to existing algorithms.

Realizing tunable functional materials with built-in nanoscale heat flow directionality represents a significant challenge with the potential to enable novel thermal management strategies. Here we use spatiotemporally-resolved thermoreflectance to visualize lateral thermal transport anisotropy in self-assembled supercrystals of anisotropic Au nanocrystals. Correlative electron and thermoreflectance microscopy reveal that heat predominantly flows along the long-axis of the anisotropic nanocrystals, and does so across grain boundaries and curved assemblies while voids disrupt heat flow. We finely control the anisotropy via the aspect ratio of constituent nanorods, and it exceeds the aspect ratio for nano-bipyramid supercrystals and certain nanorod arrangements. Finite element simulations and effective medium modeling rationalize the emergent anisotropic behavior in terms of a simple series resistance model, further providing a framework for estimating thermal anisotropy as a function of material and structural parameters. Self-assembly of colloidal nanocrystals promises a novel route to direct heat flow in a wide range of applications that utilize this important class of materials.

Precise wave manipulation has undoubtedly forged the technological landscape we thrive in today. Although our understanding of wave phenomena has come a long way since the earliest observations of desert dunes or ocean waves, the unimpeded development of mathematics has enabled ever more complex and exotic physical phenomena to be comprehensively described. Here, we take wave manipulation a step further by introducing an unprecedented synthetic acoustic crystal capable of realizing simultaneous linear broadband conduction and nonlinear topological insulation, depicting a robust amplitude-dependent mode localized deep within - i.e. an amplitude-driven topological confinement of sound. The latter is achieved by means of an open acoustic waveguide lined with a chain of nonlocally and nonlinearly coupled active electroacoustic resonators. Starting from a comprehensive topological model for classical waves, we demonstrate that different topological regimes can be accessed by increasing driving amplitude and that topological robustness against coupling disorder is a direct consequence of symmetric and simultaneous response between coupled resonators. Theoretical predictions are validated by a fully programmable experimental apparatus capable of realizing the real-time manipulation of metacrystal properties. In all, our results provide a solid foundation for future research in the design and manipulation of classical waves in artificial materials involving nonlinearity, nonlocality, and non-hermiticity.

Flow cytometry is a cornerstone technique in medical and biological research, providing crucial information about cell size and granularity through forward scatter (FSC) and side scatter (SSC) signals. Despite its widespread use, the precise relationship between these scatter signals and corresponding microscopic images remains underexplored. Here, we investigate this intrinsic relationship by utilizing scattering theory and holotomography, a three-dimensional quantitative phase imaging (QPI) technique. We demonstrate the extraction of FSC and SSC signals from individual, unlabeled cells by analyzing their three-dimensional refractive index distributions obtained through holotomography. Additionally, we introduce a method for digitally windowing SSC signals to facilitate effective segmentation and morphology-based cell type classification. Our approach bridges the gap between flow cytometry and microscopic imaging, offering a new perspective on analyzing cellular characteristics with high accuracy and without the need for labeling.

Mathematical modeling in systems toxicology enables a comprehensive understanding of the effects of pharmaceutical substances on cardiac health. However, the complexity of these models limits their widespread application in early drug discovery. In this paper, we introduce a novel approach to solving parameterized models of cardiac action potentials by combining meta-learning techniques with Systems Biology-Informed Neural Networks (SBINNs). The proposed method, HyperSBINN, effectively addresses the challenge of predicting the effects of various compounds at different concentrations on cardiac action potentials, outperforming traditional differential equation solvers in speed. Our model efficiently handles scenarios with limited data and complex parameterized differential equations. The HyperSBINN model demonstrates robust performance in predicting APD90 values, indicating its potential as a reliable tool for modeling cardiac electrophysiology and aiding in preclinical drug development. This framework represents an advancement in computational modeling, offering a scalable and efficient solution for simulating and understanding complex biological systems.

In this paper, we detail the high-performance implementation of our spaceborne radar simulator for satellite oceanography. Our software simulates the sea surface and the signal to imitate, as far as possible, the measurement process, starting from its lowest level mechanisms. In this perspective, raw data are computed as the sum of many illuminated scatterers, whose time-evolving properties are related to the surface roughness, topography, and kinematics. To achieve efficient performance, we intensively use GPU computing. Moreover, we propose a fast simulation mode based on the assumption that the instantaneous Doppler spectrum within a range gate varies on a timescale significantly larger than the PRI. The sea surface can then be updated at a frequency much smaller than the PRF, drastically reducing the computational cost. When the surface is updated, Doppler spectra are computed for all range gates. Signals segments are then obtained through 1D inverse Fourier transforms and pondered to ensure a smooth time evolution between surface updates. We validate this fast simulation mode with a radar altimeter simulation case of the Sentinel-3 SRAL instrument, showing that simulated raw data can be focused and retrieved using state-of-the-art algorithms. Finally, we show that, using a modest hardware configuration, our simulator can generate enough data in one day to compute the SWH and SSH spectra of a scene. This demonstrate that we achieved an important state-of-the-art speed-up.

Incomplete Hippocampal Inversion (IHI), sometimes called hippocampal malrotation, is an atypical anatomical pattern of the hippocampus found in about 20% of the general population. IHI can be visually assessed on coronal slices of T1 weighted MR images, using a composite score that combines four anatomical criteria. IHI has been associated with several brain disorders (epilepsy, schizophrenia). However, these studies were based on small samples. Furthermore, the factors (genetic or environmental) that contribute to the genesis of IHI are largely unknown. Large-scale studies are thus needed to further understand IHI and their potential relationships to neurological and psychiatric disorders. However, visual evaluation is long and tedious, justifying the need for an automatic method. In this paper, we propose, for the first time, to automatically rate IHI. We proceed by predicting four anatomical criteria, which are then summed up to form the IHI score, providing the advantage of an interpretable score. We provided an extensive experimental investigation of different machine learning methods and training strategies. We performed automatic rating using a variety of deep learning models (conv5-FC3, ResNet and SECNN) as well as a ridge regression. We studied the generalization of our models using different cohorts and performed multi-cohort learning. We relied on a large population of 2,008 participants from the IMAGEN study, 993 and 403 participants from the QTIM/QTAB studies as well as 985 subjects from the UKBiobank. We showed that deep learning models outperformed a ridge regression. We demonstrated that the performances of the conv5-FC3 network were at least as good as more complex networks while maintaining a low complexity and computation time. We showed that training on a single cohort may lack in variability while training on several cohorts improves generalization., Comment: Accepted for publication at the Journal of Machine Learning for Biomedical Imaging (MELBA) https://melba-journal.org/2024:016

We show the existence and uniqueness of a continuous solution to a path-dependent volatility model introduced by Guyon and Lekeufack (2023) to model the price of an equity index and its spot volatility. The considered model for the trend and activity features can be written as a Stochastic Volterra Equation (SVE) with non-convolutional and non-bounded kernels as well as non-Lipschitz coefficients. We first prove the existence and uniqueness of a solution to the SVE under integrability and regularity assumptions on the two kernels and under a condition on the second kernel weighting the past squared returns which ensures that the activity feature is bounded from below by a positive constant. Then, assuming in addition that the kernel weighting the past returns is of exponential type and that an inequality relating the logarithmic derivatives of the two kernels with respect to their second variables is satisfied, we show the positivity of the volatility process which is obtained as a non-linear function of the SVE's solution. We show numerically that the choice of an exponential kernel for the kernel weighting the past returns has little impact on the quality of model calibration compared to other choices and the inequality involving the logarithmic derivatives is satisfied by the calibrated kernels. These results extend those of Nutz and Valdevenito (2023).

Die studies are fundamental to quantifying ancient monetary production, providing insights into the relationship between coinage, politics, and history. The process requires tedious manual work, which limits the size of the corpora that can be studied. Few works have attempted to automate this task, and none have been properly released and evaluated from a computer vision perspective. We propose a fully automatic approach that introduces several innovations compared to previous methods. We rely on fast and robust local descriptors matching that is set automatically. Second, the core of our proposal is a clustering-based approach that uses an intrinsic metric (that does not need the ground truth labels) to determine its critical hyper-parameters. We validate the approach on two corpora of Greek coins, propose an automatic implementation and evaluation of previous baselines, and show that our approach significantly outperforms them., Comment: code: https://cea-list-lasti.github.io/projects/studies/studies.html

We report the discovery and characterization of a small planet, TOI-1408 c, on a 2.2-day orbit located interior to a previously known hot Jupiter, TOI-1408 b ($P=4.42$ d, $M=1.86\pm0.02\,M_\mathrm{Jup}$, $R=2.4\pm0.5\,R_\mathrm{Jup}$) that exhibits grazing transits. The two planets are near 2:1 period commensurability, resulting in significant transit timing variations (TTVs) for both planets and transit duration variations (TDVs) for the inner planet. The TTV amplitude for TOI-1408 c is 15% of the planet's orbital period, marking the largest TTV amplitude relative to the orbital period measured to date. Photodynamical modeling of ground-based radial velocity (RV) observations and transit light curves obtained with the Transiting Exoplanet Survey Satellite (TESS) and ground-based facilities leads to an inner planet radius of $2.22\pm0.06\,R_\oplus$ and mass of $7.6\pm0.2\,M_\oplus$ that locates the planet into the Sub-Neptune regime. The proximity to the 2:1 period commensurability leads to the libration of the resonant argument of the inner planet. The RV measurements support the existence of a third body with an orbital period of several thousand days. This discovery places the system among the rare systems featuring a hot Jupiter accompanied by an inner low-mass planet., Comment: Accepted to ApJL, 17 pages, 6 figures, 4 tables

3D object detection plays a crucial role in various applications such as autonomous vehicles, robotics and augmented reality. However, training 3D detectors requires a costly precise annotation, which is a hindrance to scaling annotation to large datasets. To address this challenge, we propose a weakly supervised 3D annotator that relies solely on 2D bounding box annotations from images, along with size priors. One major problem is that supervising a 3D detection model using only 2D boxes is not reliable due to ambiguities between different 3D poses and their identical 2D projection. We introduce a simple yet effective and generic solution: we build 3D proxy objects with annotations by construction and add them to the training dataset. Our method requires only size priors to adapt to new classes. To better align 2D supervision with 3D detection, our method ensures depth invariance with a novel expression of the 2D losses. Finally, to detect more challenging instances, our annotator follows an offline pseudo-labelling scheme which gradually improves its 3D pseudo-labels. Extensive experiments on the KITTI dataset demonstrate that our method not only performs on-par or above previous works on the Car category, but also achieves performance close to fully supervised methods on more challenging classes. We further demonstrate the effectiveness and robustness of our method by being the first to experiment on the more challenging nuScenes dataset. We additionally propose a setting where weak labels are obtained from a 2D detector pre-trained on MS-COCO instead of human annotations.

We study the boundary behavior of the Kobayashi-Royden metric and the Kobayashi hyperbolicity of domains in Riemannian manifolds. As an application, we prove a Fatou type theorem on the existence, almost everywhere, of non tangential limits for bounded conformal harmonic immersed discs., Comment: 12 pages

This paper introduces a computational framework designed to delineate gender distribution biases in topics covered by French TV and radio news. We transcribe a dataset of 11.7k hours, broadcasted in 2023 on 21 French channels. A Large Language Model (LLM) is used in few-shot conversation mode to obtain a topic classification on those transcriptions. Using the generated LLM annotations, we explore the finetuning of a specialized smaller classification model, to reduce the computational cost. To evaluate the performances of these models, we construct and annotate a dataset of 804 dialogues. This dataset is made available free of charge for research purposes. We show that women are notably underrepresented in subjects such as sports, politics and conflicts. Conversely, on topics such as weather, commercials and health, women have more speaking time than their overall average across all subjects. We also observe representations differences between private and public service channels., Comment: Accepted to Interspeech 2024

The rapid rise of air pollution events necessitates accurate, real-time monitoring for informed mitigation strategies. Data Assimilation (DA) methods provide promising solutions, but their effectiveness hinges heavily on optimal measurement locations. This paper presents a novel approach for air quality mapping where autonomous drones, guided by a collaborative multi-agent reinforcement learning (MARL) framework, act as airborne detectives. Ditching the limitations of static sensor networks, the drones engage in a synergistic interaction, adapting their flight paths in real time to gather optimal data for Data Assimilation (DA). Our approach employs a tailored reward function with dynamic credit assignment, enabling drones to prioritize informative measurements without requiring unavailable ground truth data, making it practical for real-world deployments. Extensive experiments using a real-world dataset demonstrate that our solution achieves significantly improved pollution estimates, even with limited drone resources or limited prior knowledge of the pollution plume. Beyond air quality, this solution unlocks possibilities for tackling diverse environmental challenges like wildfire detection and management through scalable and autonomous drone cooperation., Comment: 8 pages, 4 figures

This paper describes the submissions of team TalTech-IRIT-LIS to the DISPLACE 2024 challenge. Our team participated in the speaker diarization and language diarization tracks of the challenge. In the speaker diarization track, our best submission was an ensemble of systems based on the pyannote.audio speaker diarization pipeline utilizing powerset training and our recently proposed PixIT method that performs joint diarization and speech separation. We improve upon PixIT by using the separation outputs for speaker embedding extraction. Our ensemble achieved a diarization error rate of 27.1% on the evaluation dataset. In the language diarization track, we fine-tuned a pre-trained Wav2Vec2-BERT language embedding model on in-domain data, and clustered short segments using AHC and VBx, based on similarity scores from LDA/PLDA. This led to a language diarization error rate of 27.6% on the evaluation data. Both results were ranked first in their respective challenge tracks., Comment: accepted at Interspeech 2024

Retrieval-Augmented Generation (RAG) allows overcoming the limited knowledge of LLMs by extending the input with external information. As a consequence, the contextual inputs to the model become much longer which slows down decoding time directly translating to the time a user has to wait for an answer. We address this challenge by presenting COCOM, an effective context compression method, reducing long contexts to only a handful of Context Embeddings speeding up the generation time by a large margin. Our method allows for different compression rates trading off decoding time for answer quality. Compared to earlier methods, COCOM allows for handling multiple contexts more effectively, significantly reducing decoding time for long inputs. Our method demonstrates a speed-up of up to 5.69 $\times$ while achieving higher performance compared to existing efficient context compression methods., Comment: 10 pages

How can technical advice on operating experience feedback be managed efficiently in an organization that has never used knowledge engineering techniques and methods? This article explains how an industrial company in the nuclear and defense sectors adopted such an approach, adapted to its "TA KM" organizational context and falls within the ISO30401 framework, to build a complete system with a "SARBACANES" application to support its business processes and perpetuate its know-how and expertise in a knowledge base. Over and above the classic transfer of knowledge between experts and business specialists, SARBACANES also reveals the ability of this type of engineering to deliver multi-functional operation. Modeling was accelerated by the use of a tool adapted to this type of operation: the Ardans Knowledge Maker platform., Comment: in French language. 35es Journ{\'e}es francophones d'Ing{\'e}nierie des Connaissances (IC 2024) @ Plate-Forme Intelligence Artificielle (PFIA 2024), Association Fran\c{c}aise pour l'Intelligence Artificielle; Laboratoire L3i La Rochelle Universit{\'e}, Jul 2024, La Rochelle, France

We present an experimental study on the perturbed evolution of Korteweg-deVries soliton gases in a weakly dissipative nonlinear electrical transmission line. The system's dynamics reveal that an initially dense, fully randomized, soliton gas evolves into a coherent macroscopic state identified as a soliton condensate through nonlinear spectral analysis. The emergence of the soliton condensate is driven by the spatial rearrangement of the systems's eigenmodes and by the proliferation of new solitonic states due to nonadiabatic effects, a phenomenon not accounted for by the existing hydrodynamic theories.

The accelerating growth of the Internet of Things (IoT) and its integration with Low-Earth Orbit (LEO) satellites demand efficient, reliable, and scalable communication protocols. Among these, the Long-Range Frequency Hopping Spread Spectrum (LR-FHSS) modulation, tailored for LEO satellite IoT communications, sparks keen interest. This work presents a joint approach to enhancing the scalability of LR-FHSS, addressing the demand for massive connectivity. We deepen into Frequency Hopping Sequence (FHS) mechanisms within LR-FHSS, spotlighting the potential of leveraging Wide-Gap sequences. Concurrently, we introduce two novel demodulator allocation strategies, namely, ``Early-Decode" and ``Early-Drop," to optimize the utilization of LoRa-specific gateway decoding resources. Our research further validates these findings with extensive simulations, offering a comprehensive look into the future potential of LR-FHSS scalability in IoT settings.

Retrieval-Augmented Generation allows to enhance Large Language Models with external knowledge. In response to the recent popularity of generative LLMs, many RAG approaches have been proposed, which involve an intricate number of different configurations such as evaluation datasets, collections, metrics, retrievers, and LLMs. Inconsistent benchmarking poses a major challenge in comparing approaches and understanding the impact of each component in the pipeline. In this work, we study best practices that lay the groundwork for a systematic evaluation of RAG and present BERGEN, an end-to-end library for reproducible research standardizing RAG experiments. In an extensive study focusing on QA, we benchmark different state-of-the-art retrievers, rerankers, and LLMs. Additionally, we analyze existing RAG metrics and datasets. Our open-source library BERGEN is available under \url{https://github.com/naver/bergen}., Comment: 29 pages

Retrieval-augmented generation (RAG) has recently emerged as a promising solution for incorporating up-to-date or domain-specific knowledge into large language models (LLMs) and improving LLM factuality, but is predominantly studied in English-only settings. In this work, we consider RAG in the multilingual setting (mRAG), i.e. with user queries and the datastore in 13 languages, and investigate which components and with which adjustments are needed to build a well-performing mRAG pipeline, that can be used as a strong baseline in future works. Our findings highlight that despite the availability of high-quality off-the-shelf multilingual retrievers and generators, task-specific prompt engineering is needed to enable generation in user languages. Moreover, current evaluation metrics need adjustments for multilingual setting, to account for variations in spelling named entities. The main limitations to be addressed in future works include frequent code-switching in non-Latin alphabet languages, occasional fluency errors, wrong reading of the provided documents, or irrelevant retrieval. We release the code for the resulting mRAG baseline pipeline at https://github.com/naver/bergen.

Face recognition and verification are two computer vision tasks whose performance has progressed with the introduction of deep representations. However, ethical, legal, and technical challenges due to the sensitive character of face data and biases in real training datasets hinder their development. Generative AI addresses privacy by creating fictitious identities, but fairness problems persist. We promote fairness by introducing a demographic attributes balancing mechanism in generated training datasets. We experiment with an existing real dataset, three generated training datasets, and the balanced versions of a diffusion-based dataset. We propose a comprehensive evaluation that considers accuracy and fairness equally and includes a rigorous regression-based statistical analysis of attributes. The analysis shows that balancing reduces demographic unfairness. Also, a performance gap persists despite generation becoming more accurate with time. The proposed balancing method and comprehensive verification evaluation promote fairer and transparent face recognition and verification.

In a general fair division model with transferable utilities we discuss endogenous lower and upper guarantees on individual shares of benefits or costs. Like the more familiar exogenous bounds on individual shares described by an outside option or a stand alone utility, these guarantees depend on my type but not on others' types, only on their number and the range of types. Keeping the range from worst share to best share as narrow as permitted by the physical constraints of the model still leaves a large menu of tight guarantee functions. We describe in detail these design options in several iconic problems where each tight pair of guarantees has a clear normative meaning: the allocation of indivisible goods or costly chores, cost sharing of a public facility and the exploitation of a commons with substitute or complementary inputs. The corresponding benefit or cost functions are all sub- or super-modular, and for this class we characterise the set of minimal upper and maximal lower guarantees in all two agent problems.

Methylglyoxal (MGO) is an endogenous, highly reactive dicarbonyl metabolite generated under hyperglycaemic conditions. MGO plays a role in developing pathophysiological conditions, including diabetic cardiomyopathy. However, the mechanisms involved and the molecular targets of MGO in the heart have not been elucidated. In this work, we studied the exposure-related effects of MGO on cardiac function in an isolated perfused rat heart ex vivo model. The effect of MGO on calcium homeostasis in cardiomyocytes was studied in vitro by the fluorescence indicator of intracellular calcium Fluo-4. We demonstrated that MGO induced cardiac dysfunction, both in contractility and diastolic function. In rat heart, the effects of MGO treatment were significantly limited by aminoguanidine, a scavenger of MGO, ruthenium red, a general cation channel blocker, and verapamil, an L-type voltage-dependent calcium channel blocker, demonstrating that this dysfunction involved alteration of calcium regulation. MGO induced a significant concentration-dependent increase of intracellular calcium in neonatal rat cardiomyocytes, which was limited by aminoguanidine and verapamil. These results suggest that the functionality of various calcium channels is altered by MGO, particularly the L-type calcium channel, thus explaining its cardiac toxicity. Therefore, MGO could participate in the development of diabetic cardiomyopathy through its impact on calcium homeostasis in cardiac cells.

Efficiently quantifying predictive uncertainty in medical images remains a challenge. While Bayesian neural networks (BNN) offer predictive uncertainty, they require substantial computational resources to train. Although Bayesian approximations such as ensembles have shown promise, they still suffer from high training and inference costs. Existing approaches mainly address the costs of BNN inference post-training, with little focus on improving training efficiency and reducing parameter complexity. This study introduces a training procedure for a sparse (partial) Bayesian network. Our method selectively assigns a subset of parameters as Bayesian by assessing their deterministic saliency through gradient sensitivity analysis. The resulting network combines deterministic and Bayesian parameters, exploiting the advantages of both representations to achieve high task-specific performance and minimize predictive uncertainty. Demonstrated on multi-label ChestMNIST for classification and ISIC, LIDC-IDRI for segmentation, our approach achieves competitive performance and predictive uncertainty estimation by reducing Bayesian parameters by over 95\%, significantly reducing computational expenses compared to fully Bayesian and ensemble methods.

PLATO (PLAnetary Transits and Oscillations of stars) is ESA's M3 mission designed to detect and characterise extrasolar planets and perform asteroseismic monitoring of a large number of stars. PLATO will detect small planets (down to <2 R_(Earth)) around bright stars (<11 mag), including terrestrial planets in the habitable zone of solar-like stars. With the complement of radial velocity observations from the ground, planets will be characterised for their radius, mass, and age with high accuracy (5 %, 10 %, 10 % for an Earth-Sun combination respectively). PLATO will provide us with a large-scale catalogue of well-characterised small planets up to intermediate orbital periods, relevant for a meaningful comparison to planet formation theories and to better understand planet evolution. It will make possible comparative exoplanetology to place our Solar System planets in a broader context. In parallel, PLATO will study (host) stars using asteroseismology, allowing us to determine the stellar properties with high accuracy, substantially enhancing our knowledge of stellar structure and evolution. The payload instrument consists of 26 cameras with 12cm aperture each. For at least four years, the mission will perform high-precision photometric measurements. Here we review the science objectives, present PLATO's target samples and fields, provide an overview of expected core science performance as well as a description of the instrument and the mission profile at the beginning of the serial production of the flight cameras. PLATO is scheduled for a launch date end 2026. This overview therefore provides a summary of the mission to the community in preparation of the upcoming operational phases.

To obtain high-resolution images of subsurface structures from seismic data, seismic imaging techniques such as Full Waveform Inversion (FWI) serve as crucial tools. However, FWI involves solving a nonlinear and often non-unique inverse problem, presenting challenges such as local minima trapping and inadequate handling of inherent uncertainties. In addressing these challenges, we propose leveraging deep generative models as the prior distribution of geophysical parameters for stochastic Bayesian inversion. This approach integrates the adjoint state gradient for efficient back-propagation from the numerical solution of partial differential equations. Additionally, we introduce explicit and implicit variational Bayesian inference methods. The explicit method computes variational distribution density using a normalizing flow-based neural network, enabling computation of the Bayesian posterior of parameters. Conversely, the implicit method employs an inference network attached to a pretrained generative model to estimate density, incorporating an entropy estimator. Furthermore, we also experimented with the Stein Variational Gradient Descent (SVGD) method as another variational inference technique, using particles. We compare these variational Bayesian inference methods with conventional Markov chain Monte Carlo (McMC) sampling. Each method is able to quantify uncertainties and to generate seismic data-conditioned realizations of subsurface geophysical parameters. This framework provides insights into subsurface structures while accounting for inherent uncertainties.

New JWST/NIRCam wide-field slitless spectroscopy provides redshifts for four z>8 galaxies located behind the lensing cluster MACS J0416.1-2403. Two of them, "Y1" and "JD", have previously reported spectroscopic redshifts based on ALMA measurements of [OIII] 88 $\mu$m and/or [CII] 157.7 $\mu$m lines. Y1 is a merging system of three components, and the existing redshift z=8.31 is confirmed. However, JD is at z=8.34 instead of the previously claimed z=9.28. JD's close companion, "JD-N", which was a previously discovered z>8 candidate, is now identified at the same redshift as JD. JD and JD-N form an interacting pair. A new candidate at z>8, "f090d_018", is also confirmed and is at z=8.49. These four objects are likely part of an overdensity that signposts a large structure extending ~165 kpc in projected distance and ~48.7 Mpc in radial distance. They are magnified by less than one magnitude and have intrinsic $M_{UV}$ ranging from -19.57 to -20.83 mag. Their spectral energy distributions show that the galaxies are all very young with ages ~ 4-18 Myr and stellar masses about $10^{7-8}$ ${\rm M_\odot}$. These infant galaxies have very different star formation rates ranging from a few to over a hundred $\rm{M_\odot}$ yr$^{-1}$, but only two of them (JD and f090d_018) have blue rest-frame UV slopes $\beta<-2.0$ indicative of a high Lyman-continuum photon escape fraction that could contribute significantly to the cosmic hydrogen-reionizing background. Interestingly, these two galaxies are the least massive and least active ones among the four. The other two systems have much flatter UV slopes largely because of their high dust extinction ($A_{\rm V}$=0.9-1.0 mag). Their much lower indicated escape fractions show that even very young, actively star-forming galaxies can have negligible contribution to reionization when they quickly form dust throughout their bodies., Comment: 18 pages, 6 figures, after addressing the referee report

Deep learning models for plant species identification rely on large annotated datasets. The PlantNet system enables global data collection by allowing users to upload and annotate plant observations, leading to noisy labels due to diverse user skills. Achieving consensus is crucial for training, but the vast scale of collected data makes traditional label aggregation strategies challenging. Existing methods either retain all observations, resulting in noisy training data or selectively keep those with sufficient votes, discarding valuable information. Additionally, as many species are rarely observed, user expertise can not be evaluated as an inter-user agreement: otherwise, botanical experts would have a lower weight in the AI training step than the average user. Our proposed label aggregation strategy aims to cooperatively train plant identification AI models. This strategy estimates user expertise as a trust score per user based on their ability to identify plant species from crowdsourced data. The trust score is recursively estimated from correctly identified species given the current estimated labels. This interpretable score exploits botanical experts' knowledge and the heterogeneity of users. Subsequently, our strategy removes unreliable observations but retains those with limited trusted annotations, unlike other approaches. We evaluate PlantNet's strategy on a released large subset of the PlantNet database focused on European flora, comprising over 6M observations and 800K users. We demonstrate that estimating users' skills based on the diversity of their expertise enhances labeling performance. Our findings emphasize the synergy of human annotation and data filtering in improving AI performance for a refined dataset. We explore incorporating AI-based votes alongside human input. This can further enhance human-AI interactions to detect unreliable observations.

In this paper we describe our search for galaxy protocluster candidates at $4.5< z < 10$ and explore the environmental and physical properties of their member galaxies identified through JWST wide-field surveys within the CEERS, JADES, and PEARLS NEP-TDF fields. Combining with HST data, we identify 2948 robust $z>4.5$ candidates within an area of 185.4 arcmin$^2$. We determine nearest neighbour statistics and galaxy environments. We find that high-$z$ galaxies in overdense environments exhibit higher star formation activity compared to those in underdense regions. Galaxies in dense environments have a slightly increased SFR at a given mass compared with galaxies in the lower density environments. At the high mass end we also find a gradual flattening of the $M_{\star}$-SFR slope. We find that galaxies in high-density regions often have redder UV slopes than those in low-density regions, suggesting more dust extinction, weaker Lyman-alpha emission and / or a higher damped Lyman-alpha absorption. We also find that the mass-size relation remains consistent and statistically similar across all environments. Furthermore, we quantitatively assess the probability of a galaxy belonging to a protocluster candidate. In total, we identified 26 overdensities at $z=5-7$ and estimate their dark matter halo masses. We find that all protocluster candidates could evolve into clusters with $M_{\rm halo} > 10^{14}M_{\odot}$ at $z = 0$, thereby supporting the theoretical and simulation predictions of cluster formation. Notably, this marks an early search for protocluster candidates in JWST wide field based on photometric data, providing valuable candidates to study cosmic structure formation at the early stages., Comment: 23 pages, 14 figures and 7 table, submitted to MNRAS

Self-supervised features are the cornerstone of modern machine learning systems. They are typically pre-trained on data collections whose construction and curation typically require extensive human effort. This manual process has some limitations similar to those encountered in supervised learning, e.g., the crowd-sourced selection of data is costly and time-consuming, preventing scaling the dataset size. In this work, we consider the problem of automatic curation of high-quality datasets for self-supervised pre-training. We posit that such datasets should be large, diverse and balanced, and propose a clustering-based approach for building ones satisfying all these criteria. Our method involves successive and hierarchical applications of $k$-means on a large and diverse data repository to obtain clusters that distribute uniformly among data concepts, followed by a hierarchical, balanced sampling step from these clusters. Extensive experiments on three different data domains including web-based images, satellite images and text show that features trained on our automatically curated datasets outperform those trained on uncurated data while being on par or better than ones trained on manually curated data. Code is available at https://github.com/facebookresearch/ssl-data-curation.

We describe laboratory experiments in a 2D wave tank that aim at building up and monitor 2D shallow water soliton gas. The water surface elevation is obtained over a large ($\sim 100\,\text{m}^2$) domain, with centimetre-resolution, by stereoscopic vision using two cameras. Floating particles are seeded to get surface texture and determine the wave field by image correlation. With this set-up, soliton propagation and multiple interactions can be measured with a previously unreachable level of detail. The propagation of an oblique soliton is analysed, the amplitude decay and local incidence are compared to analytical predictions. We further present two cases of 2D soliton gas, emerging from multiple line solitons with random incidence ($|\theta|<30^\circ$) and from irregular random waves forced with a {\sc jonswap} spectrum ($|\theta|<45^\circ$). To our knowledge, those are the first observations of random 2D soliton gas for gravity waves. In both cases Mach reflections and Mach expansions result in solitons that mainly propagate in directions perpendicular to the wave-makers., Comment: Accepted for publication in Experiments in Fluids