Search

We study the drivers of the Gilchrist and Zakraj\v{s}ek (2012) excess bond premium (EBP) through the lens of the news. The monthly attention the news pays to 180 topics (Bybee et al., 2024) captures up to 80% of the variation in the EBP, and this component of variation forecasts macroeconomic movements. Greater news attention to financial intermediaries and crises tends to drive up the EBP and portend macroeconomic downturns, while greater news attention to politics and science tends to drive down the EBP. Attention-based estimates of EBP largely drive out the forecast power of direct sentiment measures for macroeconomic fluctuations and predict the business cycle going back to the early 1900's. Overall, we attribute predictive variation about the EBP for macroeconomic movements to variation in news attention to financial intermediaries, crises, and politics.

The ultra high frequency emission of gravitational waves by binary systems of black holes has recently been investigated in details in the framework of new experimental ideas around resonant cavities. In this article, we consider the case of elliptic trajectories. At fixed masses and frequency, we conclude that the total amount of energy radiated by the system within the bandwidth of the detector can be significantly higher than for circular orbits. However, due to subtle experimental effects, the signal-to-noise ratio is, overall, a decreasing function of the eccentricity. Limits on the maximum distance at which a merging system of black holes can be detected derived for circular orbits are therefore not improved by considering elliptic trajectories. The article is written as pedagogically as possible so as to be accessible to the non-familiar reader.

Despite its widespread adoption, Adam's advantage over Stochastic Gradient Descent (SGD) lacks a comprehensive theoretical explanation. This paper investigates Adam's sensitivity to rotations of the parameter space. We demonstrate that Adam's performance in training transformers degrades under random rotations of the parameter space, indicating a crucial sensitivity to the choice of basis. This reveals that conventional rotation-invariant assumptions are insufficient to capture Adam's advantages theoretically. To better understand the rotation-dependent properties that benefit Adam, we also identify structured rotations that preserve or even enhance its empirical performance. We then examine the rotation-dependent assumptions in the literature, evaluating their adequacy in explaining Adam's behavior across various rotation types. This work highlights the need for new, rotation-dependent theoretical frameworks to fully understand Adam's empirical success in modern machine learning tasks.

Privacy and regulatory constraints make data generation vital to advancing machine learning without relying on real-world datasets. A leading approach for tabular data generation is the Forest Flow (FF) method, which combines Flow Matching with XGBoost. Despite its good performance, FF is slow and makes errors when treating categorical variables as one-hot continuous features. It is also highly sensitive to small changes in the initial conditions of the ordinary differential equation (ODE). To overcome these limitations, we develop Heterogeneous Sequential Feature Forest Flow (HS3F). Our method generates data sequentially (feature-by-feature), reducing the dependency on noisy initial conditions through the additional information from previously generated features. Furthermore, it generates categorical variables using multinomial sampling (from an XGBoost classifier) instead of flow matching, improving generation speed. We also use a Runge-Kutta 4th order (Rg4) ODE solver for improved performance over the Euler solver used in FF. Our experiments with 25 datasets reveal that HS3F produces higher quality and more diverse synthetic data than FF, especially for categorical variables. It also generates data 21-27 times faster for datasets with $\geq20%$ categorical variables. HS3F further demonstrates enhanced robustness to affine transformation in flow ODE initial conditions compared to FF. This study not only validates the HS3F but also unveils promising new strategies to advance generative models.

The Fr\'echet Video Distance (FVD) is a widely adopted metric for evaluating video generation distribution quality. However, its effectiveness relies on critical assumptions. Our analysis reveals three significant limitations: (1) the non-Gaussianity of the Inflated 3D Convnet (I3D) feature space; (2) the insensitivity of I3D features to temporal distortions; (3) the impractical sample sizes required for reliable estimation. These findings undermine FVD's reliability and show that FVD falls short as a standalone metric for video generation evaluation. After extensive analysis of a wide range of metrics and backbone architectures, we propose JEDi, the JEPA Embedding Distance, based on features derived from a Joint Embedding Predictive Architecture, measured using Maximum Mean Discrepancy with polynomial kernel. Our experiments on multiple open-source datasets show clear evidence that it is a superior alternative to the widely used FVD metric, requiring only 16% of the samples to reach its steady value, while increasing alignment with human evaluation by 34%, on average.

The issue of consistency is crucial in quantum gravity. It has recently been intensively addressed for effective symmetry-reduced models. In this article, we exhaustively study the anomaly freedom of effective loop quantum cosmology with generalized holonomy corrections, considering loop correction of the constraints at the perturbative order. We pedagogically explain why, although the holonomy correction -- including the details of the chosen scheme -- applied on the background part of the constraints is crucial, it becomes irrelevant when implemented on perturbative expansions, in the sense that all consequences are "absorbed" in the counter-terms used for the regularization. The possibility of closing the algebra of constraints without counter-terms is also studied. It is argued that, although enforcing a first-class algebra is a strong requirement, this can be achieved in several different ways, often overlooked, which generates ambiguities on the restriction of the form of the generalized holonomy correction. Those ambiguities are examined in details, leading to the conclusion that the consistency of the effective theory for cosmological perturbations, especially when considering scalar modes, cannot be achieved without counter-terms. We also take the opportunity of this work to clarify, as much as possible, all the required steps so that future works have a clear material at disposal. In particular, a highly detailed calculation of all the brackets is provided, emphasizing the (usually implicit) assumptions, hypotheses and manipulations required to ensure the closure of the algebra. Prospects for future works are underlined., Comment: 45 pages, no figures. Any questions, comments, and criticism are welcome

This is an index of the contributions by the Giant Radio Array for Neutrino Detection (GRAND) Collaboration to the 10th International Workshop on Acoustic and Radio EeV Neutrino Detection Activities (ARENA 2024, University of Chicago, June 11-14, 2024). The contributions include an overview of GRAND in its present and future incarnations, methods of radio-detection that are being developed for them, and ongoing joint work between the GRAND and BEACON experiments., Comment: Note: To access the list of contributions, please follow the "HTML" link that can be found on the arXiv page

Neutrinos at ultrahigh energies can originate both from interactions of cosmic rays at their acceleration sites and through cosmic-ray interactions as they propagate through the universe. These neutrinos are expected to have a low flux which drives the need for instruments with large effective areas. Radio observations of the inclined air showers induced by tau neutrino interactions in rock can achieve this, because radio waves can propagate essentially unattenuated through the hundreds of kilometers of atmosphere. Proposed arrays for radio detection of tau neutrinos focus on either arrays of inexpensive receivers distributed over a large area, the GRAND concept, or compact phased arrays on elevated mountains, the BEACON concept, to build up a large detector area with a low trigger threshold. We present a concept that combines the advantages of these two approaches with a trigger driven by phased arrays at a moderate altitude (1 km) and sparse, high-gain outrigger receivers for reconstruction and background rejection. We show that this design has enhanced sensitivity at 100 PeV over the two prior designs with fewer required antennas and discuss the need for optimized antenna designs., Comment: ARENA2024 Conference Proceeding PoS(ARENA2024)058

One of the major challenges for the radio detection of extensive air showers, as encountered by the Giant Radio Array for Neutrino Detection (GRAND), is the requirement of an autonomous radio self-trigger. This work presents the current development of self-triggering techniques at the detection-unit level -- the so-called first-level trigger (FLT) -- in the context of the NUTRIG project. A second-level trigger (SLT) at the array level is described in a separate contribution. Two FLT methods are described, based on a template-fitting algorithm and a convolutional neural network (CNN). In this work, we compare the preliminary offline performance of both FLT methods in terms of signal selection efficiency and background rejection efficiency. We find that for both methods, ${\gtrsim}40\%$ of the background can be rejected if a signal selection efficiency of 90\% is required at the $5\sigma$ level., Comment: Presented at the 10th International Workshop on Acoustic and Radio EeV Neutrino Activities (ARENA2024). 8 pages, 4 figures

The Giant Radio Array for Neutrino Detection (GRAND) is a distributed, sparse ground antenna array designed to detect the radio emission from highly inclined extensive air showers induced by ultra-high-energy particles in the atmosphere. We use CoREAS air-shower simulations to adapt an existing signal model of the radio emission of inclined showers to the 50-200 MHz frequency band GRAND is sensitive to. The model uses a parameterisation of the charge excess emission to isolate the geomagnetic component. By fitting a one-dimensional lateral distribution function to the geomagnetic energy fluence of a shower, we reconstruct its radiation energy. This work details the signal model and the intrinsic event reconstruction of our method, as well as the adaptations to the new frequency band. This work is part of the NUTRIG project., Comment: 8 pages, 6 figures, Conference proceeding for 10th International Workshop on Acoustic and Radio EeV Neutrino Detection Activities

Large-scale cosmic-ray detectors like the Giant Radio Array for Neutrino Detection (GRAND) are pushing the boundaries of our ability to identify air shower events. Existing trigger schemes rely solely on the timing of signals detected by individual antennas, which brings many challenges in distinguishing true air shower signals from background. This work explores novel event-level radio trigger methods specifically designed for GRAND, but also applicable to other systems, such as the Radio Detector (RD) of the Pierre Auger Observatory. In addition to an upgraded plane wave front reconstruction technique, we introduce orthogonal and complementary approaches that analyze the radio-emission footprint, the spatial distribution of signal strength across triggered antennas, to refine event selection. We test our methods on mock data sets constructed with simulated showers and real background noise measured with the GRAND prototype, to assess the performance potential in terms of sensitivity and background rejection in GRAND. Our preliminary results are a first step to identifying the most discriminating radio signal features at event-level, and optimizing the techniques for future implementation on experimental data., Comment: 8 pages, 5 figures, 10th International Workshop on Acoustic and Radio EeV Neutrino Detection Activities - ARENA2024

The Giant Radio Array for Neutrino Detection (GRAND) aims to detect highly inclined extensive air showers (EAS) with down-going and up-going trajectories. Several working groups in the GRAND collaboration are developing methods to reconstruct the incoming direction, core position, primary energy, and composition of the showers. The reconstruction pipeline -- currently under development in the France/IAP working group -- relies on a model of spherical wavefront emission for arrival times, which is possible because the radio signals are generated far away from the antenna stations. The amplitude distribution of the signals at the antenna level is described by an Angular Distribution Function that considers various asymmetries in the data, including geomagnetic effects. In this contribution, we present preliminary results from testing our EAS reconstruction procedure using realistic mock observations., Comment: V2: Fixed typo in axis labels of Fig.5 and collaboration author list updated

The operation of upcoming ultra-high-energy cosmic-ray, gamma-ray, and neutrino radio-detection experiments, like the Giant Radio Array for Neutrino Detection (GRAND), poses significant computational challenges involving the production of numerous simulations of particle showers and their detection, and a high data throughput. GRANDlib is an open-source software tool designed to meet these challenges. Its primary goal is to perform end-to-end simulations of the detector operation, from the interaction of ultra-high-energy particles, through -- by interfacing with external air-shower simulations -- the ensuing particle shower development and its radio emission, to its detection by antenna arrays and its processing by data-acquisition systems. Additionally, GRANDlib manages the visualization, storage, and retrieval of experimental and simulated data. We present an overview of GRANDlib to serve as the basis of future GRAND analyses., Comment: 11 pages, 9 figures, plus appendices

Generating novel molecules is challenging, with most representations leading to generative models producing many invalid molecules. Spanning Tree-based Graph Generation (STGG) is a promising approach to ensure the generation of valid molecules, outperforming state-of-the-art SMILES and graph diffusion models for unconditional generation. In the real world, we want to be able to generate molecules conditional on one or multiple desired properties rather than unconditionally. Thus, in this work, we extend STGG to multi-property-conditional generation. Our approach, STGG+, incorporates a modern Transformer architecture, random masking of properties during training (enabling conditioning on any subset of properties and classifier-free guidance), an auxiliary property-prediction loss (allowing the model to self-criticize molecules and select the best ones), and other improvements. We show that STGG+ achieves state-of-the-art performance on in-distribution and out-of-distribution conditional generation, and reward maximization., Comment: Code: https://github.com/SamsungSAILMontreal/AnyMolGenCritic

Organizations across various industries are still exploring the potential of Generative Artificial Intelligence (GenAI) to enhance knowledge work. While innovation is often viewed as a product of individual creativity, it more commonly unfolds through a highly structured, collaborative process where creativity intertwines with knowledge work. However, the extent and effectiveness of GenAI in supporting this process remain open questions. Our study investigates this issue using a collaborative practice research approach focused on three GenAI-enabled innovation projects conducted over a year within three different organizations. We explored how, why, and when GenAI could be integrated into design sprints, a highly structured, collaborative, and human-centered innovation method. Our research identified challenges and opportunities in synchronizing AI capabilities with human intelligence and creativity. To translate these insights into practical strategies, we propose four recommendations for organizations eager to leverage GenAI to both streamline and bring more value to their innovation processes: (1) establish a collaborative intelligence value loop with GenAI; (2) build trust in GenAI, (3) develop robust data collection and curation workflows, and (4) cultivate a craftsmanship mindset., Comment: 46 pages, 4 figures

With recent advances in video prediction, controllable video generation has been attracting more attention. Generating high fidelity videos according to simple and flexible conditioning is of particular interest. To this end, we propose a controllable video generation model using pixel level renderings of 2D or 3D bounding boxes as conditioning. In addition, we also create a bounding box predictor that, given the initial and ending frames' bounding boxes, can predict up to 15 bounding boxes per frame for all the frames in a 25-frame clip. We perform experiments across 3 well-known AV video datasets: KITTI, Virtual-KITTI 2 and BDD100k.

A good initialization of deep learning models is essential since it can help them converge better and faster. However, pretraining large models is unaffordable for many researchers, which makes a desired prediction for initial parameters more necessary nowadays. Graph HyperNetworks (GHNs), one approach to predicting model parameters, have recently shown strong performance in initializing large vision models. Unfortunately, predicting parameters of very wide networks relies on copying small chunks of parameters multiple times and requires an extremely large number of parameters to support full prediction, which greatly hinders its adoption in practice. To address this limitation, we propose LoGAH (Low-rank GrAph Hypernetworks), a GHN with a low-rank parameter decoder that expands to significantly wider networks without requiring as excessive increase of parameters as in previous attempts. LoGAH allows us to predict the parameters of 774-million large neural networks in a memory-efficient manner. We show that vision and language models (i.e., ViT and GPT-2) initialized with LoGAH achieve better performance than those initialized randomly or using existing hypernetworks. Furthermore, we show promising transfer learning results w.r.t. training LoGAH on small datasets and using the predicted parameters to initialize for larger tasks. We provide the codes in https://github.com/Blackzxy/LoGAH ., Comment: 16 pages

In May 2023, the LIGO Livingston observatory detected the likely black hole-neutron star (BHNS) merger GW230529_181500. That event is expected to be the merger of a 2.5-4.5 $M_{\odot}$ primary with a secondary compact object of mass between 1.2-2.0 $M_{\odot}$. This makes it the first BHNS merger with a significant potential for the production of electromagnetic (EM) counterparts, and provides further evidence for compact objects existing within the suspected lower mass gap. To produce post-merger EM transients, the component of the black hole spin aligned with the orbital angular momentum must be sufficiently high, allowing the neutron star to be tidally disrupted. The disrupting BHNS binary may then eject a few percent of a solar mass of matter, leading to an observable kilonova driven by radioactive decays in ejecta, and/or a compact-binary GRB (cbGRB) resulting from the formation of an accretion disk and relativistic jet. Determining which mergers lead to disruption of the neutron star is necessary to predict the prevalence of EM signals from BHNS mergers, yet most BHNS simulations so far have been performed far from the minimum spin required for tidal disruption. Here, we use the Spectral Einstein Code (SpEC) to explore the behavior of BHNS mergers in a mass range consistent with GW230529_181500 close to that critical spin, and compare our results against the mass remnant model currently used by the LVK collaboration to predict the probability of tidal disruption. Our numerical results reveal the emergence of non-zero accretion disks even below the predicted NS disruption limit, of low mass but capable of powering cbGRBs. Our results also demonstrate that the remnant mass model underpredicts the disk mass for the DD2 EOS, while they are within expected modeling errors for SFHo. In all of our simulations, any kilonova signal would be dim and dominated by post-merger disk outflows., Comment: 9 pages, 5 figures

We review the kinematic effects on a gravitational wave due to either a peculiar motion of the astrophysical source emitting it or a local motion of the observer. Working in the context of general relativity, we show at fully non-linear order in velocity, that the amplitude of the wave is amplified by the Doppler factor in the case in which the source moves with respect to a reference frame, while it is invariant if the observer moves (with respect to a reference observer). However, the observed specific intensity transforms in the same way under a boost of the source or of the observer. We also show at fully non-linear order that under a boost (of either source or observer), the polarization tensor is rotated in the same way the wave direction is rotated by aberration, such that the only net effect of a boost on polarization is to change the phase of the helicity components. We apply these results to a wave emitted by a binary system of compact objects in the cosmological context., Comment: 39 pages, 5 figures

In this brief note, we pursue the systematic investigation of possible gravitational wave sources in the gigahertz band. We focus on hyperbolic encounters of light black holes and evaluate precisely the expected signal when accounting for the detailed characteristics of haloscope experiments. Considering the GraHal setup as a benchmark, we insist on the correct signal-to-noise ratio expression taking into account the appropriate timescales. The associated maximum distance - of the order of a few hundreds of astronomical units - at which an event can be detected is calculated for optimal, sub-optimal, and general trajectories.

Accurate detection and segmentation of diffuse large B-cell lymphoma (DLBCL) from PET images has important implications for estimation of total metabolic tumor volume, radiomics analysis, surgical intervention and radiotherapy. Manual segmentation of tumors in whole-body PET images is time-consuming, labor-intensive and operator-dependent. In this work, we develop and validate a fast and efficient three-step cascaded deep learning model for automated detection and segmentation of DLBCL tumors from PET images. As compared to a single end-to-end network for segmentation of tumors in whole-body PET images, our three-step model is more effective (improves 3D Dice score from 58.9% to 78.1%) since each of its specialized modules, namely the slice classifier, the tumor detector and the tumor segmentor, can be trained independently to a high degree of skill to carry out a specific task, rather than a single network with suboptimal performance on overall segmentation., Comment: 8 pages, 3 figures, 3 tables

This study investigates the acceptability of different artificial intelligence (AI) applications in education from a multi-stakeholder perspective, including students, teachers, and parents. Acknowledging the transformative potential of AI in education, it addresses concerns related to data privacy, AI agency, transparency, explainability and the ethical deployment of AI. Through a vignette methodology, participants were presented with four scenarios where AI's agency, transparency, explainability, and privacy were manipulated. After each scenario, participants completed a survey that captured their perceptions of AI's global utility, individual usefulness, justice, confidence, risk, and intention to use each scenario's AI if available. The data collection comprising a final sample of 1198 multi-stakeholder participants was distributed through a partner institution and social media campaigns and focused on individual responses to four AI use cases. A mediation analysis of the data indicated that acceptance and trust in AI varies significantly across stakeholder groups. We found that the key mediators between high and low levels of AI's agency, transparency, and explainability, as well as the intention to use the different educational AI, included perceived global utility, justice, and confidence. The study highlights that the acceptance of AI in education is a nuanced and multifaceted issue that requires careful consideration of specific AI applications and their characteristics, in addition to the diverse stakeholders' perceptions., Comment: 28 pages, 2 appendices, 3 figures, 5 tables, original research

We examine the main properties of gravitational waves (GWs) emitted by transient hyperbolic encounters of black holes. We begin by building the set of basic variables most relevant to setting our problem. After exposing the ranges of masses and eccentricities accessible at a given GW frequency, we analyze the dependence of the gravitational strain on those parameters and determine the trajectories resulting in the most sizeable strains. Some non-trivial behaviors are unveiled, showing that highly eccentric events can be more easily detectable than parabolic ones. In particular, we underline the correct way to extend formulas from hyperbolic to parabolic orbits. Our reasonings are as general as possible, and we make a point of explaining our considerations pedagogically. The majority of the work is based on Newtonian dynamics and aims at being a benchmark to which more accurate calculations can be compared., Comment: 11 pages, 4 figures

Transcranial ultrasound imaging is currently limited by attenuation and aberration induced by the skull. First used in contrast-enhanced ultrasound (CEUS), highly echoic microbubbles allowed for the development of novel imaging modalities such as ultrasound localization microscopy (ULM). Herein, we develop an inverse problem approach to aberration correction (IPAC) that leverages the sparsity of microbubble signals. We propose to use the \textit{a priori} knowledge of the medium based upon microbubble localization and wave propagation to build a forward model to link the measured signals directly to the aberration function. A standard least-squares inversion is then used to retrieve the aberration function. We first validated IPAC on simulated data of a vascular network using plane wave as well as divergent wave emissions. We then evaluated the reproducibility of IPAC \textit{in vivo} in 5 mouse brains. We showed that aberration correction improved the contrast of CEUS images by 4.6 dB. For ULM images, IPAC yielded sharper vessels, reduced vessel duplications, and improved the resolution from 21.1 $\mu$m to 18.3 $\mu$m. Aberration correction also improved hemodynamic quantification for velocity magnitude and flow direction.

This study addresses critical gaps in automated lymphoma segmentation from PET/CT images, focusing on issues often overlooked in existing literature. While deep learning has been applied for lymphoma lesion segmentation, few studies incorporate out-of-distribution testing, raising concerns about model generalizability across diverse imaging conditions and patient populations. We highlight the need to compare model performance with expert human annotators, including intra- and inter-observer variability, to understand task difficulty better. Most approaches focus on overall segmentation accuracy but overlook lesion-specific measures important for precise lesion detection and disease quantification. To address these gaps, we propose a clinically relevant framework for evaluating deep segmentation networks. Using this lesion measure-specific evaluation, we assess the performance of four deep networks (ResUNet, SegResNet, DynUNet, and SwinUNETR) across 611 cases from multi-institutional datasets, covering various lymphoma subtypes and lesion characteristics. Beyond standard metrics like the Dice similarity coefficient, we evaluate clinical lesion measures and their prediction errors. We also introduce detection criteria for lesion localization and propose a new detection Criterion 3 based on metabolic characteristics. We show that networks perform better on large, intense lesions with higher metabolic activity. Finally, we compare network performance to physicians via intra- and inter-observer variability analyses, demonstrating that network errors closely resemble those made by experts, i.e., the small and faint lesions remain challenging for both humans and networks. This study aims to improve automated lesion segmentation's clinical relevance, supporting better treatment decisions for lymphoma patients. The code is available at: https://github.com/microsoft/lymphoma-segmentation-dnn., Comment: 32 pages, 15 figures, 5 tables