Search

During the past decade, Deep Neural Networks (DNNs) proved their value on a large variety of subjects. However despite their high value and public accessibility, the protection of the intellectual property of DNNs is still an issue and an emerging research field. Recent works have successfully extracted fully-connected DNNs using cryptanalytic methods in hard-label settings, proving that it was possible to copy a DNN with high fidelity, i.e., high similitude in the output predictions. However, the current cryptanalytic attacks cannot target complex, i.e., not fully connected, DNNs and are limited to special cases of neurons present in deep networks. In this work, we introduce a new end-to-end attack framework designed for model extraction of embedded DNNs with high fidelity. We describe a new black-box side-channel attack which splits the DNN in several linear parts for which we can perform cryptanalytic extraction and retrieve the weights in hard-label settings. With this method, we are able to adapt cryptanalytic extraction, for the first time, to non-fully connected DNNs, while maintaining a high fidelity. We validate our contributions by targeting several architectures implemented on a microcontroller unit, including a Multi-Layer Perceptron (MLP) of 1.7 million parameters and a shortened MobileNetv1. Our framework successfully extracts all of these DNNs with high fidelity (88.4% for the MobileNetv1 and 93.2% for the MLP). Furthermore, we use the stolen model to generate adversarial examples and achieve close to white-box performance on the victim's model (95.8% and 96.7% transfer rate).

Modern science is formally structured around scholarly publication, where scientific knowledge is canonized through citation. Precisely how citations are given and accrued can provide information about the value of discovery, the history of scientific ideas, the structure of fields, and the space or scope of inquiry. Yet parsing this information has been challenging because citations are not simply present or absent; rather, they differ in purpose, function, and sentiment. In this paper, we investigate how critical and favorable sentiments are distributed across citations, and demonstrate that citation sentiment tracks sociocultural norms across scales of collaboration, discipline, and country. At the smallest scale of individuals, we find that researchers cite scholars they have collaborated with more favorably (and less critically) than scholars they have not collaborated with. Outside collaborative relationships, higher h-index scholars cite lower h-index scholars more critically. At the mesoscale of disciplines, we find that wetlab disciplines tend to be less critical than drylab disciplines, and disciplines that engage in more synthesis through publishing more review articles tend to be less critical. At the largest scale of countries, we find that greater individualism (and lesser acceptance of the unequal distribution of power) is associated with more critical sentiment. Collectively, our results demonstrate how sociocultural factors can explain variations in sentiment in scientific communication. As such, our study contributes to the broader understanding of how human factors influence the practice of science, and underscore the importance of considering the larger sociocultural contexts in which science progresses., Comment: 16 pages, 8 figures in main; 13 pages, 3 figures in supplement

Within the context of the ALICE ITS3 collaboration, a set of MAPS small-scale test structures were developed using the 65 nm TPSCo CMOS imaging process with the upgrade of the ALICE inner tracking system as its primary focus. One such sensor, the Circuit Exploratoire 65 nm (CE-65), and its evolution the CE-65v2, were developed to explore charge collection properties for varying configurations including collection layer process (standard, blanket, modified with gap), pixel pitch (15, 18, \SI{22.5}{\micro\meter}), and pixel geometry (square vs hexagonal/staggered). In this work the characterisation of the CE-65v2 chip, based on $^{55}$Fe lab measurements and test beams at CERN SPS, is presented. Matrix gain uniformity up to the $\mathcal{O}$(5\%) level was demonstrated for all considered chip configurations. The CE-65v2 chip achieves a spatial resolution of under \SI{2}{\micro\meter} during beam tests. Process modifications allowing for faster charge collection and less charge sharing result in decreased spatial resolution, but a considerably wider range of operation, with both the \SI{15}{\micro\meter} and \SI{22.5}{\micro\meter} chips achieving over 99\% efficiency up to a $\sim$180 e$^{-}$ seed threshold. The results serve to validate the 65 nm TPSCo CMOS process, as well as to motivate design choices in future particle detection experiments., Comment: 7 pages, 12 figures; Proceedings for iWoRiD 2024 (Lisbon)

In this article, we investigate the possibility of generating all the configurations of a subshift in a local way. We propose two definitions of local generation, explore their properties and develop techniques to determine whether a subshift satisfies these definitions. We illustrate the results with several examples.

In this article, we apply the techniques developed in our previous article ``Local generation of tilings'', in which we introduced two definitions capturing the intuitive idea that some subshifts admit a procedure that can generate any tiling and working in a local way. We classify all the Wang tilesets with two colors in which each tile has an even number of each color., Comment: arXiv admin note: text overlap with arXiv:2411.08697

Due to its fundamental principles, quantum theory holds the promise to enhance the security of modern cryptography, from message encryption to anonymous communication, digital signatures, online banking, leader election, one-time passwords and delegated computation. While quantum key distribution (QKD) has already enabled secure key exchange over hundreds of kilometers, a myriad of other quantum-cryptographic primitives are being developed to secure future applications against quantum adversaries. This article surveys the theoretical and experimental developments in quantum cryptography beyond QKD over the past decades, along with advances in secure quantum computation. It provides an intuitive classification of the main quantum primitives and their security levels, summarizes their possibilities and limits, and discusses their implementation with current photonic technology., Comment: Submitted to Reviews of Modern Physics. Comments are welcome

Thermal transport in nanostructures plays a critical role in modern technologies. As devices shrink, techniques that can measure thermal properties at nanometer and nanosecond scales are increasingly needed to capture transient, out-of-equilibrium phenomena. We present a novel pump-probe photon-electron method within a scanning transmission electron microscope (STEM) to map temperature dynamics with unprecedented spatial and temporal resolutions. By combining focused laser-induced heating and synchronized time-resolved monochromated electron energy loss spectroscopy (EELS), we track phonon, exciton and plasmon signals in various materials, including silicon nitride, aluminum thin film, and transition metal dichalcogenides. Our results demonstrate the technique's ability to follow temperature changes at the nanometer and nanosecond scales. The experimental data closely matched theoretical heat diffusion models, confirming the method's validity. This approach opens new opportunities to investigate transient thermal phenomena in nanoscale materials, offering valuable insights for applications in thermoelectric devices and nanoelectronics.

Offline Reinforcement Learning (RL) has emerged as a powerful alternative to imitation learning for behavior modeling in various domains, particularly in complex navigation tasks. An existing challenge with Offline RL is the signal-to-noise ratio, i.e. how to mitigate incorrect policy updates due to errors in value estimates. Towards this, multiple works have demonstrated the advantage of hierarchical offline RL methods, which decouples high-level path planning from low-level path following. In this work, we present a novel hierarchical transformer-based approach leveraging a learned quantizer of the space. This quantization enables the training of a simpler zone-conditioned low-level policy and simplifies planning, which is reduced to discrete autoregressive prediction. Among other benefits, zone-level reasoning in planning enables explicit trajectory stitching rather than implicit stitching based on noisy value function estimates. By combining this transformer-based planner with recent advancements in offline RL, our proposed approach achieves state-of-the-art results in complex long-distance navigation environments., Comment: Under review. Code will be released upon acceptance

This work aims at investigating the impact of DNA geometry, compaction and calculation chain on DNA break and chromosome aberration predictions for high charge and energy (HZE) ions, using the Monte Carlo codes Geant4-DNA, RITRACKS and RITCARD. To ensure consistency of ion transport of both codes, we first compared microdosimetry and nanodosimetry spectra for different ions of interest in hadrontherapy and space research. The Rudd model was used for the transport of ions in both models. Developments were made in Geant4 (v11.2) to include periodic boundary conditions (PBC) to account for electron equilibrium in small targets. Excellent agreements were found for both microdosimetric and nanodosimetric spectra for all ion types, with and without PBC. Some discrepancies remain for low-energy deposition events, likely due to differences in electron interaction models. The latest results obtained using the newly available Geant4 example ``dsbandrepair'' will be presented and compared to DNA break predictions obtained with RITCARD.

We demonstrate a quantum walk comb in synthetic frequency space formed by externally modulating a semiconductor optical amplifier operating in the telecommunication wavelength range in a unidirectional ring cavity. The ultrafast gain saturation dynamics of the gain medium and its operation at high current injections is responsible for the stabilization of the comb in a broad frequency modulated state. Our device produces a nearly flat broadband comb with a tunable repetition frequency reaching a bandwidth of 1.8THz at the fundamental repetition rate of 1GHz while remaining fully locked to the RF drive. Comb operation at harmonics of the repetition rate up to 14.1GHz is also demonstrated. This approach paves the way for next-generation optical frequency comb devices with potential applications in precision ranging and high-speed communications.

In the context of particle therapy monitoring, we are developing a gamma-ray detector to determine the ion range in vivo from the measurement of particle time-of-flight. For this application, a beam monitor capable to tag in time the incident ion with a time resolution below 235 ps FWHM (100 ps rms) is required to provide a start signal for the acquisition. We have therefore developed a dedicated detector based on a fast organic scintillator (EJ-204) of 25x25x1 mm3 coupled to four SiPM strips that allow measuring the particle incident position by scintillation light sharing. The prototype was characterised with single protons of energies between 63 and 225 MeV at the MEDICYC and ProteusONE facilities of the Antoine Lacassagne proton therapy centre in Nice. We obtained a time resolution of 120 ps FWHM at 63 MeV, and a spatial resolution of ~2 mm rms for single particles. Two identical detectors also allowed to measure the MEDICYC proton energy with 0.3% accuracy.

We calculate the gravitational waves (GWs) produced by primordial black holes (PBHs) in the presence of the inflaton condensate in the early Universe. Combining the GW production from the evaporation process, the gravitational scattering of the inflaton itself, and the density fluctuations due to the inhomogeneous distribution of PBHs, we propose for the first time a complete coherent analysis of the spectrum, revealing three peaks, one for each source. Three frequency ranges ($\sim$ kHz, GHz, and PHz, respectively) are expected, each giving rise to a similar GW peak amplitude $\Omega_{\rm GW}$. We also compare our predictions with current and future GWs detection experiments., Comment: 21 pages, 10 figures

This paper is devoted to the study of the nonlinear Schr\"odinger-Poisson system with a doping profile. We are interested in the strong instability of standing waves associated with ground state solutions in the $L^2$-supercritical case. The presence of a doping profile causes several difficulties, especially in examining geometric shapes of fibering maps along an $L^2$-invariant scaling curve. Furthermore, the classical approach by Berestycki-Cazenave for the strong instability cannot be applied to our problem due to a remainder term caused by the doping profile. To overcome these difficulties, we establish a new energy inequality associated with the $L^2$-invariant scaling and adopt the strong instability result developed by Fukaya-Ohta(2018). When the doping profile is a characteristic function supported on a bounded smooth domain, some geometric quantities related to the domain, such as the mean curvature, are responsible for the strong instability of standing waves., Comment: arXiv admin note: substantial text overlap with arXiv:2411.02103; text overlap with arXiv:2409.01842

In open-world scenarios, where both novel classes and domains may exist, an ideal segmentation model should detect anomaly classes for safety and generalize to new domains. However, existing methods often struggle to distinguish between domain-level and semantic-level distribution shifts, leading to poor out-of-distribution (OOD) detection or domain generalization performance. In this work, we aim to equip the model to generalize effectively to covariate-shift regions while precisely identifying semantic-shift regions. To achieve this, we design a novel generative augmentation method to produce coherent images that incorporate both anomaly (or novel) objects and various covariate shifts at both image and object levels. Furthermore, we introduce a training strategy that recalibrates uncertainty specifically for semantic shifts and enhances the feature extractor to align features associated with domain shifts. We validate the effectiveness of our method across benchmarks featuring both semantic and domain shifts. Our method achieves state-of-the-art performance across all benchmarks for both OOD detection and domain generalization. Code is available at https://github.com/gaozhitong/MultiShiftSeg., Comment: Published in NeurIPS 2024

Accurate stabilization of facial motion is essential for applications in photoreal avatar construction for 3D games, virtual reality, movies, and training data collection. For the latter, stabilization must work automatically for the general population with people of varying morphology. Distinguishing rigid skull motion from facial expressions is critical since misalignment between skull motion and facial expressions can lead to animation models that are hard to control and can not fit natural motion. Existing methods struggle to work with sparse sets of very different expressions, such as when combining multiple units from the Facial Action Coding System (FACS). Certain approaches are not robust enough, some depend on motion data to find stable points, while others make one-for-all invalid physiological assumptions. In this paper, we leverage recent advances in neural signed distance fields and differentiable isosurface meshing to compute skull stabilization rigid transforms directly on unstructured triangle meshes or point clouds, significantly enhancing accuracy and robustness. We introduce the concept of a stable hull as the surface of the boolean intersection of stabilized scans, analogous to the visual hull in shape-from-silhouette and the photo hull from space carving. This hull resembles a skull overlaid with minimal soft tissue thickness, upper teeth are automatically included. Our skull carving algorithm simultaneously optimizes the stable hull shape and rigid transforms to get accurate stabilization of complex expressions for large diverse sets of people, outperforming existing methods., Comment: 4 pages, 3 figures

This paper is devoted to the study of the nonlinear Schr\"odinger-Poisson system with a doping profile. We are interested in the existence of ground state solutions by considering the minimization problem on a Nehari-Pohozaev set. The presence of a doping profile causes several difficulties, especially in the proof of the uniqueness of a maximum point of a fibering map. A key ingredient is to establish the energy inequality. We also establish the relation between ground state solutions and $L^2$-constraint minimizers. When the doping profile is a characteristic function supported on a bounded smooth domain, some geometric quantities related to the domain, such as the mean curvature,are responsible for the existence of ground state solutions.

We will cover the basics of several complex variables in 4 lectures: Basic properties of holomorphic functions in several variables, the notion of pseudoconvexity, CR functions and CR geometry, and the $\bar\partial$-problem. The main underlying idea is to connect various characterizations of domains of holomorphy, that is, the natural domains of definition for holomorphic functions. In the process we will connect the function theory on the domain to geometric properties of the boundary, and discuss the relationship between the boundary values and the functions themselves and extension of holomorphic functions from subspaces. These notes are based on a series of lectures held during the S-Matrix Marathon workshop at the Institute for Advanced Study on 11-22 March 2024., Comment: A chapter to be published by Springer Lecture Notes in Physics

3D Gaussian Splatting (3DGS) has demonstrated remarkable effectiveness for novel view synthesis (NVS). However, the 3DGS model tends to overfit when trained with sparse posed views, limiting its generalization capacity for broader pose variations. In this paper, we alleviate the overfitting problem by introducing a self-ensembling Gaussian Splatting (SE-GS) approach. We present two Gaussian Splatting models named the $\mathbf{\Sigma}$-model and the $\mathbf{\Delta}$-model. The $\mathbf{\Sigma}$-model serves as the primary model that generates novel-view images during inference. At the training stage, the $\mathbf{\Sigma}$-model is guided away from specific local optima by an uncertainty-aware perturbing strategy. We dynamically perturb the $\mathbf{\Delta}$-model based on the uncertainties of novel-view renderings across different training steps, resulting in diverse temporal models sampled from the Gaussian parameter space without additional training costs. The geometry of the $\mathbf{\Sigma}$-model is regularized by penalizing discrepancies between the $\mathbf{\Sigma}$-model and the temporal samples. Therefore, our SE-GS conducts an effective and efficient regularization across a large number of Gaussian Splatting models, resulting in a robust ensemble, the $\mathbf{\Sigma}$-model. Experimental results on the LLFF, Mip-NeRF360, DTU, and MVImgNet datasets show that our approach improves NVS quality with few-shot training views, outperforming existing state-of-the-art methods. The code is released at https://github.com/sailor-z/SE-GS.

Band-to-band tunnelling field-effect transistors (TFETs) have long been considered as promising candidates for future low-power logic applications. However, fabricated TFETs rarely reach sub-60 mV/dec sub-threshold swings (SS) at room temperature. Previous theoretical studies identified Auger processes as possible mechanisms for the observed degradation of SS. Through first-principles quantum transport simulations incorporating carrier-carrier interactions within the Non-equilibrium Green's Function formalism and self-consistent GW approximation, we confirm here that Auger processes are indeed at least partly responsible for the poor performance of TFETs. Using a carbon nanotube TFET as testbed, we show that carrier-carrier scattering alone significantly increases the OFF-state current of these devices, thus worsening their sub-threshold behavior., Comment: 6 pages, 5 figures

This paper describes speech enhancement for realtime automatic speech recognition (ASR) in real environments. A standard approach to this task is to use neural beamforming that can work efficiently in an online manner. It estimates the masks of clean dry speech from a noisy echoic mixture spectrogram with a deep neural network (DNN) and then computes a enhancement filter used for beamforming. The performance of such a supervised approach, however, is drastically degraded under mismatched conditions. This calls for run-time adaptation of the DNN. Although the ground-truth speech spectrogram required for adaptation is not available at run time, blind dereverberation and separation methods such as weighted prediction error (WPE) and fast multichannel nonnegative matrix factorization (FastMNMF) can be used for generating pseudo groundtruth data from a mixture. Based on this idea, a prior work proposed a dual-process system based on a cascade of WPE and minimum variance distortionless response (MVDR) beamforming asynchronously fine-tuned by block-online FastMNMF. To integrate the dereverberation capability into neural beamforming and make it fine-tunable at run time, we propose to use weighted power minimization distortionless response (WPD) beamforming, a unified version of WPE and minimum power distortionless response (MPDR), whose joint dereverberation and denoising filter is estimated using a DNN. We evaluated the impact of run-time adaptation under various conditions with different numbers of speakers, reverberation times, and signal-to-noise ratios (SNRs)., Comment: Accepted to APSIPA2024

Cells use signalling pathways as windows into the environment to gather information, transduce it into their interior, and use it to drive behaviours. MAPK (ERK) is a highly conserved signalling pathway in eukaryotes, directing multiple fundamental cellular behaviours such as proliferation, migration, and differentiation, making it of few central hubs in the signalling circuitry of cells. Despite this versatility of behaviors, population-level measurements have reported low information content (\textless 1 bit) relayed through the ERK pathway, rendering the population barely able to distinguish the presence or absence of stimuli. Here, we contrast the information transmitted by a single cell and a population of cells. Using a combination of optogenetic experiments, data analysis based on information theory framework, and numerical simulations we quantify the amount of information transduced from the receptor to ERK, from responses to singular, brief and sparse input pulses. We show that single cells are indeed able to resolve between graded stimuli, yielding over 2 bit of information, however showing a large population heterogeneity., Comment: 13 pages, 5 figures, 10 pages of supplemental material

In the evolving landscape of deep learning, there is a pressing need for more comprehensive datasets capable of training models across multiple modalities. Concurrently, in digital humanities, there is a growing demand to leverage technology for diverse media adaptation and creation, yet limited by sparse datasets due to copyright and stylistic constraints. Addressing this gap, our paper presents a novel dataset comprising Franco-Belgian comics from the 1950s annotated for tasks including depth estimation, semantic segmentation, saliency detection, and character identification. It consists of two distinct and consistent styles and incorporates object concepts and labels taken from natural images. By including such diverse information across styles, this dataset not only holds promise for computational creativity but also offers avenues for the digitization of art and storytelling innovation. This dataset is a crucial component of the AI4VA Workshop Challenges~\url{https://sites.google.com/view/ai4vaeccv2024}, where we specifically explore depth and saliency. Dataset details at \url{https://github.com/IVRL/AI4VA}., Comment: ECCV 2024 Workshop Proceedings

The Lane-Emden inequality controls $\iint_{\mathbb{R}^{2d}}\rho(x)\rho(y)|x-y|^{-\lambda}\,dx\,dy$ in terms of the $L^1$ and $L^p$ norms of $\rho$. We provide a remainder estimate for this inequality in terms of a suitable distance of $\rho$ to the manifold of optimizers.

Preclinical diffusion MRI (dMRI) has proven value in methods development and validation, characterizing the biological basis of diffusion phenomena, and comparative anatomy. While dMRI enables in vivo non-invasive characterization of tissue, ex vivo dMRI is increasingly being used to probe tissue microstructure and brain connectivity. Ex vivo dMRI has several experimental advantages that facilitate high spatial resolution and high signal-to-noise ratio (SNR) images, cutting-edge diffusion contrasts, and direct comparison with histological data as a methodological validation. However, there are a number of considerations that must be made when performing ex vivo experiments. The steps from tissue preparation, image acquisition and processing, and interpretation of results are complex, with many decisions that not only differ dramatically from in vivo imaging of small animals, but ultimately affect what questions can be answered using the data. This work concludes a 3-part series of recommendations and considerations for preclinical dMRI. Herein, we describe best practices for dMRI of ex vivo tissue, with a focus on image pre-processing, data processing and model fitting, and tractography. In each section, we attempt to provide guidelines and recommendations, but also highlight areas for which no guidelines exist (and why), and where future work should lie. We end by providing guidelines on code sharing and data sharing, and point towards open-source software and databases specific to small animal and ex vivo imaging., Comment: Part 3 of 3 in "Considerations and recommendations for preclinical diffusion MRI"

Demand response is a concept that has been around since the very first electric power systems. However, we have seen an explosion of research on demand response and demand-side technologies in the past 30 years, coinciding with the shift towards liberalized/deregulated electricity markets and efforts to decarbonize the power sector. Now we are also seeing a shift towards more distributed/decentralized electric systems; we have entered the era of "distributed energy resources," which require new grid management, operational, and control strategies. Given this paradigm shift, we argue that the concept of demand response needs to be revisited, and more carefully/consistently defined to enable us to better utilize this massive resource for economic, technical, environmental, and societal aims. In this paper, we survey existing demand response definitions, highlight their shortcomings, propose a new definition, and describe how this new definition enables us to more effectively harness the value of demand response in modern power systems. We conclude with a demand response research agenda informed by a discussion of demand response barriers and enablers., Comment: 12 pages

NGC 2506 is a rich, intermediate-age (2.0 Gyr), metal-poor ([Fe/H] $\sim$ -0.2) open cluster. This work presents the results of 12,157 spectroscopic radial-velocity measurements of 2,442 stars in the NGC 2506 field over 41 years, made as part of the WIYN Open Cluster Study. Radial-velocity measurements are complete for the population of proper-motion member stars brighter than a Gaia G magnitude of 15.5, in which 320 proper-motion and radial-velocity cluster members were identified. Within the observation limit of G $<$ 16.5, 469 proper-motion and radial-velocity members were identified. This work reports on the characteristics of NGC 2506, including projected spatial distribution, radial-velocity dispersion, and virial mass. This work also presents orbital solutions for 49 binary members with periods between 1 and 7,580 days. NGC 2506 has an incompleteness-corrected binary frequency for binaries with periods less than $10^4$ days of $35 \pm 5$%. This work also discusses in detail the 14 blue stragglers stars of NGC 2506$\unicode{x2014}$finding at least 64 $\pm$ 21% to be in binaries, 5 yellow straggler stars, and several other stars of note., Comment: 41 pages, 13 figures, open access publication and machine-readable tables available at https://doi.org/10.3847/1538-3881/ad6b1a

Despite significant advancements in neural text-to-audio generation, challenges persist in controllability and evaluation. This paper addresses these issues through the Sound Scene Synthesis challenge held as part of the Detection and Classification of Acoustic Scenes and Events 2024. We present an evaluation protocol combining objective metric, namely Fr\'echet Audio Distance, with perceptual assessments, utilizing a structured prompt format to enable diverse captions and effective evaluation. Our analysis reveals varying performance across sound categories and model architectures, with larger models generally excelling but innovative lightweight approaches also showing promise. The strong correlation between objective metrics and human ratings validates our evaluation approach. We discuss outcomes in terms of audio quality, controllability, and architectural considerations for text-to-audio synthesizers, providing direction for future research., Comment: accepted to NeurIPS 2024 Workshop: Audio Imagination

Modelling partial differential equations (PDEs) is of crucial importance in science and engineering, and it includes tasks ranging from forecasting to inverse problems, such as data assimilation. However, most previous numerical and machine learning approaches that target forecasting cannot be applied out-of-the-box for data assimilation. Recently, diffusion models have emerged as a powerful tool for conditional generation, being able to flexibly incorporate observations without retraining. In this work, we perform a comparative study of score-based diffusion models for forecasting and assimilation of sparse observations. In particular, we focus on diffusion models that are either trained in a conditional manner, or conditioned after unconditional training. We address the shortcomings of existing models by proposing 1) an autoregressive sampling approach that significantly improves performance in forecasting, 2) a new training strategy for conditional score-based models that achieves stable performance over a range of history lengths, and 3) a hybrid model which employs flexible pre-training conditioning on initial conditions and flexible post-training conditioning to handle data assimilation. We empirically show that these modifications are crucial for successfully tackling the combination of forecasting and data assimilation, a task commonly encountered in real-world scenarios., Comment: Accepted at NeurIPS 2024

The rise of large language models (LLMs) for visually rich document understanding (VRDU) has kindled a need for prompt-response, document-based datasets. As annotating new datasets from scratch is labor-intensive, the existing literature has generated prompt-response datasets from available resources using simple templates. For the case of key information extraction (KIE), one of the most common VRDU tasks, past work has typically employed the template "What is the value for the {key}?". However, given the variety of questions encountered in the wild, simple and uniform templates are insufficient for creating robust models in research and industrial contexts. In this work, we present K2Q, a diverse collection of five datasets converted from KIE to a prompt-response format using a plethora of bespoke templates. The questions in K2Q can span multiple entities and be extractive or boolean. We empirically compare the performance of seven baseline generative models on K2Q with zero-shot prompting. We further compare three of these models when training on K2Q versus training on simpler templates to motivate the need of our work. We find that creating diverse and intricate KIE questions enhances the performance and robustness of VRDU models. We hope this work encourages future studies on data quality for generative model training., Comment: Accepted to EMNLP Findings 2024

Purpose: In sleep medicine, assessing the evolution of a subject's sleep often involves the costly manual scoring of electroencephalographic (EEG) signals. In recent years, a number of Deep Learning approaches have been proposed to automate this process, mainly by extracting features from said signals. However, despite some promising developments in related problems, such as Brain-Computer Interfaces, analyses of the covariances between brain regions remain underutilized in sleep stage scoring.Methods: Expanding upon our previous work, we investigate the capabilities of SPDTransNet, a Transformer-derived network designed to classify sleep stages from EEG data through timeseries of covariance matrices. Furthermore, we present a novel way of integrating learned signal-wise features into said matrices without sacrificing their Symmetric Definite Positive (SPD) nature.Results: Through comparison with other State-of-the-Art models within a methodology optimized for class-wise performance, we achieve a level of performance at or beyond various State-of-the-Art models, both in single-dataset and - particularly - multi-dataset experiments.Conclusion: In this article, we prove the capabilities of our SPDTransNet model, particularly its adaptability to multi-dataset tasks, within the context of EEG sleep stage scoring - though it could easily be adapted to any classification task involving timeseries of covariance matrices.

Artificial Neural Networks (ANNs), including fully-connected networks and transformers, are highly flexible and powerful function approximators, widely applied in fields like computer vision and natural language processing. However, their inability to inherently respect causal structures can limit their robustness, making them vulnerable to covariate shift and difficult to interpret/explain. This poses significant challenges for their reliability in real-world applications. In this paper, we introduce Causal Fully-Connected Neural Networks (CFCNs) and Causal Transformers (CaTs), two general model families designed to operate under predefined causal constraints, as specified by a Directed Acyclic Graph (DAG). These models retain the powerful function approximation abilities of traditional neural networks while adhering to the underlying structural constraints, improving robustness, reliability, and interpretability at inference time. This approach opens new avenues for deploying neural networks in more demanding, real-world scenarios where robustness and explainability is critical.

Tissues of living cells are a prime example of active fluids. There is experimental evidence that tissues generate extensile active stress even though their constituting cells are contractile. Fluctuating forces that could result from cell-substrate interactions have been proposed to be able to induce a transition from contractile to extensile active stress. Through analytic calculations and numerical computations, we show that in isotropic active fluids, nonlinearities and a coupling between fluctuating forces and fluid density are necessary for such a transition to occur. Here, both transitions from extensile to contractile and vice versa are possible., Comment: 10 pages, 2 figures

Reconstruction functions are pivotal in sample compression theory, a framework for deriving tight generalization bounds. From a small sample of the training set (the compression set) and an optional stream of information (the message), they recover a predictor previously learned from the whole training set. While usually fixed, we propose to learn reconstruction functions. To facilitate the optimization and increase the expressiveness of the message, we derive a new sample compression generalization bound for real-valued messages. From this theoretical analysis, we then present a new hypernetwork architecture that outputs predictors with tight generalization guarantees when trained using an original meta-learning framework. The results of promising preliminary experiments are then reported., Comment: Accepted at the NeurIPS 2024 workshop on Compression in Machine Learning

The problem of classifying graphs is ubiquitous in machine learning. While it is standard to apply graph neural networks or graph kernel methods, Gaussian processes can be employed by transforming spatial features from the graph domain into spectral features in the Euclidean domain, and using them as input points. However, this approach only takes into account features on vertices, whereas some graph datasets also support features on edges. In this work, we present a Gaussian process-based classification algorithm that can leverage one or both vertex and edges features. Furthermore, we take advantage of the Hodge decomposition to better capture the intricate richness of vertex and edge features, which can be beneficial on diverse tasks., Comment: NeurIPS 2024 Workshop on Bayesian Decision-Making and Uncertainty (Oral Presentation)

We demonstrate efficient in-plane optical fiber collection of single photon emission from quantum dots embedded in photonic crystal cavities. This was achieved via adiabatic coupling between a tapered optical fiber and a tapered on-chip photonic waveguide coupled to the photonic crystal cavity. The collection efficiency of a dot in a photonic crystal cavity was measured to be 5 times greater via the tapered optical fiber compared to collection by a microscope objective lens above the cavity. The single photon source was also characterized by second order photon correlations measurements giving g(2)(0)=0.17 under non-resonant excitation. Numerical calculations demonstrate that the collection efficiency could be further increased by improving the dot-cavity coupling and by increasing the overlap length of the tapered fiber with the on-chip waveguide. An adiabatic coupling of near unity is predicted for an overlap length of 5 microns., Comment: 22 pages, 12 figures

Optimal stopping is a fundamental problem in optimization that has found applications in risk management, finance, economics, and recently in the fields of computer science. We extend the standard framework to a multi-agent setting, named multi-agent optimal stopping (MAOS), where a group of agents cooperatively solves finite-space, discrete-time optimal stopping problems. Solving the finite-agent case is computationally prohibitive when the number of agents is very large, so this work studies the mean field optimal stopping (MFOS) problem, obtained as the number of agents approaches infinity. We prove that MFOS provides a good approximate solution to MAOS. We also prove a dynamic programming principle (DPP), based on the theory of mean field control. We then propose two deep learning methods: one simulates full trajectories to learn optimal decisions, whereas the other leverages DPP with backward induction; both methods train neural networks for the optimal stopping decisions. We demonstrate the effectiveness of these approaches through numerical experiments on 6 different problems in spatial dimension up to 300. To the best of our knowledge, this is the first work to study MFOS in finite space and discrete time, and to propose efficient and scalable computational methods for this type of problem.

In the algorithm Intersort, Chevalley et al. (2024) proposed a score-based method to discover the causal order of variables in a Directed Acyclic Graph (DAG) model, leveraging interventional data to outperform existing methods. However, as a score-based method over the permutahedron, Intersort is computationally expensive and non-differentiable, limiting its ability to be utilised in problems involving large-scale datasets, such as those in genomics and climate models, or to be integrated into end-to-end gradient-based learning frameworks. We address this limitation by reformulating Intersort using differentiable sorting and ranking techniques. Our approach enables scalable and differentiable optimization of causal orderings, allowing the continuous score function to be incorporated as a regularizer in downstream tasks. Empirical results demonstrate that causal discovery algorithms benefit significantly from regularizing on the causal order, underscoring the effectiveness of our method. Our work opens the door to efficiently incorporating regularization for causal order into the training of differentiable models and thereby addresses a long-standing limitation of purely associational supervised learning.

With the rise of SSL and ASR technologies, the Wav2Vec2 ASR-based model has been fine-tuned for automated speech disorder quality assessment tasks, yielding impressive results and setting a new baseline for Head and Neck Cancer speech contexts. This demonstrates that the ASR dimension from Wav2Vec2 closely aligns with assessment dimensions. Despite its effectiveness, this system remains a black box with no clear interpretation of the connection between the model ASR dimension and clinical assessments. This paper presents the first analysis of this baseline model for speech quality assessment, focusing on intelligibility and severity tasks. We conduct a layer-wise analysis to identify key layers and compare different SSL and ASR Wav2Vec2 models based on pre-trained data. Additionally, post-hoc XAI methods, including Canonical Correlation Analysis (CCA) and visualization techniques, are used to track model evolution and visualize embeddings for enhanced interpretability., Comment: Accepted at the Spoken Language Technology (SLT) Conference 2024

We present ZeroComp, an effective zero-shot 3D object compositing approach that does not require paired composite-scene images during training. Our method leverages ControlNet to condition from intrinsic images and combines it with a Stable Diffusion model to utilize its scene priors, together operating as an effective rendering engine. During training, ZeroComp uses intrinsic images based on geometry, albedo, and masked shading, all without the need for paired images of scenes with and without composite objects. Once trained, it seamlessly integrates virtual 3D objects into scenes, adjusting shading to create realistic composites. We developed a high-quality evaluation dataset and demonstrate that ZeroComp outperforms methods using explicit lighting estimations and generative techniques in quantitative and human perception benchmarks. Additionally, ZeroComp extends to real and outdoor image compositing, even when trained solely on synthetic indoor data, showcasing its effectiveness in image compositing.

This paper addresses the problem of long-context linear system identification, where the state $x_t$ of a dynamical system at time $t$ depends linearly on previous states $x_s$ over a fixed context window of length $p$. We establish a sample complexity bound that matches the i.i.d. parametric rate up to logarithmic factors for a broad class of systems, extending previous works that considered only first-order dependencies. Our findings reveal a learning-without-mixing phenomenon, indicating that learning long-context linear autoregressive models is not hindered by slow mixing properties potentially associated with extended context windows. Additionally, we extend these results to (i) shared low-rank representations, where rank-regularized estimators improve rates with respect to dimensionality, and (ii) misspecified context lengths in strictly stable systems, where shorter contexts offer statistical advantages., Comment: 30 pages, 4 figures

Domain Generalization (DG) aims to train models that perform well not only on the training (source) domains but also on novel, unseen target data distributions. A key challenge in DG is preventing overfitting to source domains, which can be mitigated by finding flatter minima in the loss landscape. In this work, we propose Quantization-aware Training for Domain Generalization (QT-DoG) and demonstrate that weight quantization effectively leads to flatter minima in the loss landscape, thereby enhancing domain generalization. Unlike traditional quantization methods focused on model compression, QT-DoG exploits quantization as an implicit regularizer by inducing noise in model weights, guiding the optimization process toward flatter minima that are less sensitive to perturbations and overfitting. We provide both theoretical insights and empirical evidence demonstrating that quantization inherently encourages flatter minima, leading to better generalization across domains. Moreover, with the benefit of reducing the model size through quantization, we demonstrate that an ensemble of multiple quantized models further yields superior accuracy than the state-of-the-art DG approaches with no computational or memory overheads. Our extensive experiments demonstrate that QT-DoG generalizes across various datasets, architectures, and quantization algorithms, and can be combined with other DG methods, establishing its versatility and robustness., Comment: Code will be released soon

We present experimental evidence of multiple blood flow configurations in a relatively simple microfluidic network under constant inlet conditions. We provide evidence of multistability and unsteady dynamics and find good agreement with a theoretical {one-dimensional advection} model for blood flow in microvascular networks{ that relies on the widely used laws for rheology and phase separation}. We discuss the ramifications for microfluidic experiments and measurements using blood and implications for in vivo microcirculation. Our findings suggest that further modeling in microvascular networks should discard the usual assumption of unique, steady-state flow solutions, with crucial consequences regarding gas, nutrient, and waste transport.

Interacting binary star systems play a critical role in many areas of astrophysics. One interesting example of a binary merger product are Thorne-\.Zytkow Objects (T\.ZOs), stars that look like red supergiants but contain neutron stars at their cores. T\.ZOs were theorized nearly five decades ago, and significant work has gone into understanding the physics of their formation, evolution, and stability. Several searches for T\.ZO candidates have also been carried out. Whether or not T\.ZOs could even exist or if they would be stable after formation has also been investigated. Understanding the existence and possible prevalence of T\.ZOs would have important effects on our understanding of binary evolution, stellar mergers, and inform binary population synthesis models. In this chapter, we review the formation channels, evolution and structure, final fates, and observable signatures of T\.ZOs, as well as candidates in the literature, from the inception of T\.ZO theory to recent progress in the field., Comment: This is a pre-print of a chapter for the Encyclopedia of Astrophysics (edited by I. Mandel, section editor J. Andrews) to be published by Elsevier as a Reference Module. 17 pages, 10 figures

Stacking the public Planck CMB temperature maps (NILC, SMICA, SEVEM, Commander) on galaxy clusters from Planck catalogues reveals substantial residual contamination from thermal Sunyaev-Zeldovich (tSZ) emission. Unexpectedly, stacking "tSZ-free" CMB maps, like the Planck SMICA-noSZ or Constrained ILC (CILC) maps, still shows noticeable residual contamination from galaxy clusters. We demonstrate that this persisting residual stems from neglected relativistic SZ (rSZ) corrections in the CMB map estimation. Employing the component separation method outlined in Remazeilles & Chluba (2020) on Planck data, we map the rSZ first-order moment field $y(T_{\rm e}-\bar{T}_{\rm e})$ over the sky for different pivot temperatures $\bar{T}_{\rm e}$ ranging from $2$ to $10$ keV. Stacking these $y(T_{\rm e}-\bar{T}_{\rm e})$-maps on Planck clusters exhibits either an intensity decrement or increment at the centre, contingent upon whether $\bar{T}_{\rm e}$ is above or below the ensemble-averaged cluster temperature $T_{\rm e}$. For the pivot value $\bar{T}_{\rm e}=5$ keV, a vanishing intensity is observed in the stacked Planck $y(T_{\rm e}-\bar{T}_{\rm e})$-map, enabling us to infer the average gas temperature of $T_{\rm e}\simeq 5$ keV for Planck clusters. Building upon this finding, we revisit the Planck tSZ-free CMB map by deprojecting the complete rSZ emission using CILC, assuming an rSZ spectrum with $T_{\rm e} = 5$ keV. Our new, rSZ-free Planck CMB map, when stacked on clusters, shows a clear cancellation of residual SZ contamination in contrast to prior (non-relativistic) tSZ-free Planck CMB maps. Our map-based approach provides compelling evidence for an average temperature of the Planck galaxy clusters of $T_{\rm e} = 4.9 \pm 2.6$ keV using the rSZ effect., Comment: 10 pages, 9 figures, 3 tables. Minor rewording and reduced abstract for MNRAS

The rising interest in leveraging higher-order interactions present in complex systems has led to a surge in more expressive models exploiting high-order structures in the data, especially in topological deep learning (TDL), which designs neural networks on high-order domains such as simplicial complexes. However, progress in this field is hindered by the scarcity of datasets for benchmarking these architectures. To address this gap, we introduce MANTRA, the first large-scale, diverse, and intrinsically high order dataset for benchmarking high-order models, comprising over 43,000 and 249,000 triangulations of surfaces and three-dimensional manifolds, respectively. With MANTRA, we assess several graph- and simplicial complex-based models on three topological classification tasks. We demonstrate that while simplicial complex-based neural networks generally outperform their graph-based counterparts in capturing simple topological invariants, they also struggle, suggesting a rethink of TDL. Thus, MANTRA serves as a benchmark for assessing and advancing topological methods, leading the way for more effective high-order models., Comment: 26 pages, 2 figures, 22 tables

We introduce and study a random non-compact space called the bigeodesic Brownian plane, and prove that it is the tangent plane in distribution of the Brownian sphere at a point of its simple geodesic from the root (for the local Gromov-Hausdorff-Prokhorov-Uniform topology). We also show that it is the local limit of the Brownian plane rerooted further and further on its unique infinite geodesic. Furthermore, we discuss various properties of this space, such as its topology, the behavior of its geodesic rays, and its invariance in distribution under several natural transformations., Comment: 55 pages, 6 figures

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

By directly probing the initial conditions of our universe, cosmological surveys offer us a unique observational handle on quantum field theory in curved spacetime with dynamical gravity and might even allow us to glean information about a full theory of quantum gravity. Here we report on recent progress to study the natural observables in the problem, namely cosmological correlators. After setting the stage, we review results from three different approaches. First, we present the in-out formalism as an interesting alternative to the well-known in-in formalism and stress some of its advantages, such as the derivation of recursion relations, correlators cutting rules and a proposal for a de Sitter scattering matrix. Second, we tackle the important open problem of constructing effective theories in curved spacetime, which generally requires an open quantum system approach. Third, we provide an executive summary of general properties of the field-theoretic wavefunction that follow from symmetries, unitarity, causality and locality. We describe how these properties can be leveraged to bootstrap all tree-level results and we discuss loop contributions. These notes are based on a series of lectures held during the S-Matrix Marathon workshop at the Institute for Advanced Study on 11-22 March 2024., Comment: A chapter to be published by Springer Lecture Notes in Physics

Blue straggler stars in stellar clusters are a subset of stars that are bluer and appear younger than other cluster members, seemingly straggling behind in their evolution. They offer a unique opportunity to understand the stellar dynamics and populations within their hosts. In the collisional formation scenario, a persistent challenge is the excessive angular momentum in the collision product. The consequent significant mass loss during transition to a stable state leads to a star with too low a mass to be a blue straggler, unless it spins down efficiently. While many proposed spin-down mechanisms involve boosted angular momentum loss via magnetic braking within the collision product, the existence or strength of these magnetic fields has not been confirmed. Here, we report three-dimensional magnetohydrodynamical simulations of collisions between two low-mass main-sequence stars and investigate magnetic field amplification. Magnetic field energy is amplified by a factor of $10^{8}-10^{10}$, resulting in the magnetic field strength of $10^{7}-10^{8}$G at the core of the collision product, independent of collision parameters. The surface magnetic field strengths have increased up to $10-10^{4}$ G. In addition, a distinctly flattened, rotating gas structure appears around the collision products in off-axis collisions, which may be a hint of possible disk formation. Such significant magnetic amplification and potential disk formation suggest the possibility of efficient spin-down of collision products via magnetic braking and magnetic disk locking, which can result in their appearance as blue stragglers., Comment: 27 pages, 10 figures, 1 table. Comments welcome! Videos: https://www.youtube.com/watch?v=lkpTEJTI1Ew&list=PLxLK3qI02cQef6wX5wi-QMSFyxBGK4tGL