Search

As the complexities of urban environments continue to grow, the modelling of transportation systems become increasingly challenging. This paper explores the application of advanced Graph Neural Network (GNN) architectures as surrogate models for strategic transport planning. Building upon a prior work that laid the foundation with graph convolution networks (GCN), our study delves into the comparative analysis of established GCN with the more expressive Graph Attention Network (GAT). Additionally, we propose a novel GAT variant (namely GATv3) to address over-smoothing issues in graph-based models. Our investigation also includes the exploration of a hybrid model combining both GCN and GAT architectures, aiming to investigate the performance of the mixture. The three models are applied to various experiments to understand their limits. We analyse hierarchical regression setups, combining classification and regression tasks, and introduce fine-grained classification with a proposal of a method to convert outputs to precise values. Results reveal the superior performance of the new GAT in classification tasks. To the best of the authors' knowledge, this is the first GAT model in literature to achieve larger depths. Surprisingly, the fine-grained classification task demonstrates the GCN's unexpected dominance with additional training data. This shows that synthetic data generators can increase the training data, without overfitting issues whilst improving model performance. In conclusion, this research advances GNN based surrogate modelling, providing insights for refining GNN architectures. The findings open avenues for investigating the potential of the newly proposed GAT architecture and the modelling setups for other transportation problems.

Using hard (E>10 keV) X-ray observations with NuSTAR, we are able to differentiate between accretion states, and thus compact object types, of neutron stars and black holes in X-ray binaries (XRBs) in M31, our nearest Milky Way-type neighbor. Using ten moderate-depth (20-50 ks) observations of the disk of M31 covering a total of ~0.45 deg$^{2}$, we detect 20 sources at 2$\sigma$ in the 4-25 keV band pass, 14 of which we consider to be XRB candidates. This complements an existing deeper (100-400 ks) survey covering ~0.2 deg$^{2}$ of the bulge and the northeastern disk. We make tentative classifications of 9 of these sources with the use of diagnostic color-intensity and color-color diagrams, which separate sources into various neutron star and black hole regimes, identifying 3 black holes and 6 neutron stars. In addition, we create X-ray luminosity functions for both the full (4-25 keV) and hard (12-25 keV) band, as well as sub-populations of the full band based on compact object type and association with globular clusters. Our best fit globular cluster XLF is shallower than the field XLF, and preliminary BH and NS XLFs suggest a difference in shape based on compact object type. We find that the cumulative disk XLFs in the full and hard band are best fit by power laws with indices of 1.32 and 1.28 respectively. This is consistent with models of the Milky Way XLF from Grimm et al. (2002), Voss & Ajello (2010), and Doroshenko et al. (2014)., Comment: 19 pages, 6 Figures, Published in ApJ

Alternative theories of gravity may be put to the test by making use of gravitational wave observations. Scalar-tensor theories provide a relatively simple framework that allows for compact object solutions with characteristics different from those of their general relativity counterparts. In shift-symmetric theories, black hole hair may arise when a linear coupling with the Gauss-Bonnet (GB) invariant is introduced. The effect of this coupling on the black hole quasinormal modes (QNMs) has been shown to be small, therefore limiting the observational interest in the theory. In general, however, one expects that additional shift-symmetric terms in the Lagrangian may be relevant, and indeed, it has been shown that they can have a nontrivial impact on the mass and scalar charge of the stationary solutions. In this work we explore the effect that various such terms have on the axial and polar QNMs of the hairy black hole solutions., Comment: 14 pages, 6 figures, 2 tables

Neural architecture search (NAS) finds high performing networks for a given task. Yet the results of NAS are fairly prosaic; they did not e.g. create a shift from convolutional structures to transformers. This is not least because the search spaces in NAS often aren't diverse enough to include such transformations a priori. Instead, for NAS to provide greater potential for fundamental design shifts, we need a novel expressive search space design which is built from more fundamental operations. To this end, we introduce einspace, a search space based on a parameterised probabilistic context-free grammar. Our space is versatile, supporting architectures of various sizes and complexities, while also containing diverse network operations which allow it to model convolutions, attention components and more. It contains many existing competitive architectures, and provides flexibility for discovering new ones. Using this search space, we perform experiments to find novel architectures as well as improvements on existing ones on the diverse Unseen NAS datasets. We show that competitive architectures can be obtained by searching from scratch, and we consistently find large improvements when initialising the search with strong baselines. We believe that this work is an important advancement towards a transformative NAS paradigm where search space expressivity and strategic search initialisation play key roles., Comment: Project page at https://linusericsson.github.io/einspace/

Gravitational wave observations can test the validity of General Relativity (GR) in the strong field regime. Certain classes of scalar-tensor theories indeed predict that compact objects can exhibit significant deviations from their GR counterparts. Here we explore the quasinormal modes of axial perturbations in spherically symmetric black holes in three such classes: (i) dilatonic black holes with an additional scalar-Ricci coupling (EdRGB), (ii) spontaneously scalarized black holes (EsRGB) with a quadratic coupling to the Gauss-Bonnet invariant and the Ricci scalar, (iii) spontaneously scalarized black holes with a quadratic and a quartic coupling to the Gauss-Bonnet invariant., Comment: 12 pages, 7 figures

Despite recent advances in video action recognition achieving strong performance on existing benchmarks, these models often lack robustness when faced with natural distribution shifts between training and test data. We propose two novel evaluation methods to assess model resilience to such distribution disparity. One method uses two different datasets collected from different sources and uses one for training and validation, and the other for testing. More precisely, we created dataset splits of HMDB-51 or UCF-101 for training, and Kinetics-400 for testing, using the subset of the classes that are overlapping in both train and test datasets. The other proposed method extracts the feature mean of each class from the target evaluation dataset's training data (i.e. class prototype) and estimates test video prediction as a cosine similarity score between each sample to the class prototypes of each target class. This procedure does not alter model weights using the target dataset and it does not require aligning overlapping classes of two different datasets, thus is a very efficient method to test the model robustness to distribution shifts without prior knowledge of the target distribution. We address the robustness problem by adversarial augmentation training - generating augmented views of videos that are "hard" for the classification model by applying gradient ascent on the augmentation parameters - as well as "curriculum" scheduling the strength of the video augmentations. We experimentally demonstrate the superior performance of the proposed adversarial augmentation approach over baselines across three state-of-the-art action recognition models - TSM, Video Swin Transformer, and Uniformer. The presented work provides critical insight into model robustness to distribution shifts and presents effective techniques to enhance video action recognition performance in a real-world deployment.

The perturbative treatment of realistic quantum field theories, such as quantum electrodynamics, requires the use of mathematical idealizations in the approximation series for scattering amplitudes. Such mathematical idealisations are necessary to derive empirically relevant models from the theory. Mathematical idealizations can be either controlled or uncontrolled, depending on whether current scientific knowledge can explain whether the effects of the idealization are negligible or not. Drawing upon negative formal results in asymptotic analysis (failure of Borel summability) and renormalization group theory (failure of asymptotic safety), we argue that the mathematical idealizations applied in perturbative quantum electrodynamics should be understood as uncontrolled. This, in turn, leads to the problematic conclusion that such theories do not have theoretical models in the standard understanding of this term. The existence of unquestionable empirically successful theories without theoretical models has significant implications both for our understanding of the theory-model relationship in physics and the concept of empirical adequacy.

We present a new five-epoch Chandra X-ray Observatory monitoring survey of the nearby spiral galaxy M33 which probes X-ray variability with time sampling between two weeks and four months. We characterize the X-ray variability of 55 bright point sources outside of the nucleus, many of which are expected to be high-mass X-ray binaries (HMXBs). We detect eight new candidate transients not detected in previous X-ray catalogs of M33 and discuss their possible nature. The final catalog includes 26 known HMXB candidates identified in the literature. We extend the baseline of the X-ray light curves up to 21 years by including archival X-ray observations of these sources. We compare the detection and non-detection epochs of the sources to suites of simulated source duty cycles and infer that most of our detected sources have duty cycles > 30%. We find only four sources whose detection patterns are consistent with having duty cycles below 30%. This large fraction of sources with high duty cycles is unexpected for a population of HMXBs, thus more frequent X-ray monitoring will likely reveal many more low duty cycle HMXBs in M33., Comment: Accepted for publication in ApJ. 31 pages, 14 figures

We analyze VeLO (versatile learned optimizer), the largest scale attempt to train a general purpose "foundational" optimizer to date. VeLO was trained on thousands of machine learning tasks using over 4000 TPU months with the goal of producing an optimizer capable of generalizing to new problems while being hyperparameter free, and outperforming industry standards such as Adam. We independently evaluate VeLO on the MLCommons optimizer benchmark suite. We find that, contrary to initial claims: (1) VeLO has a critical hyperparameter that needs problem-specific tuning, (2) VeLO does not necessarily outperform competitors in quality of solution found, and (3) VeLO is not faster than competing optimizers at reducing the training loss. These observations call into question VeLO's generality and the value of the investment in training it.

We present some of the salient aspects of the scientific motivation for high resolution soft X-ray spectroscopy of early-type stars with the Line Emission Mapper X-ray Probe. The major strength of {\it LEM} for hot star physics is its large effective area, aided by the inherent energy resolution of its microcalorimeter that readily achieves resolving powers of 1000 and obviates the need for relatively inefficient dispersive optical elements. This increased sensitivity enables much fainter and more distant high mass stars to be observed than are accessible with present-day facilities, greatly increasing the pool of potential targets. For brighter sources, the sensitivity opens up time domain studies, wherein sufficient signal can be garnered in short order and exposure times, probing source variations on ks timescales. We argue that these capabilities of {\it LEM} will yield breakthroughs in all types of hot star systems, from understanding single OB and WR star winds and how they vary with metallicity, to probing the shocks of colliding wind systems and the magnetically channeled winds of magnetic OB stars. {\it LEM} will also study the energetics of WR star bubbles and feedback from their powerful pre-SN stellar winds., Comment: A Line Emission Mapper White Paper

The ion injectors of the CERN accelerator chain, in particular the Super Proton Synchrotron (SPS) and the Low Energy Ion Ring (LEIR), operate in a strong Space Charge (SC) and Intra-Beam Scattering (IBS) regime, which can degrade beam quality. Optimizing the ion beam performance requires thus to study the interplay of these two effects in tracking simulations by incorporating both SC and IBS effects interleaved with lattice non-linearities. In this respect, the kinetic theory approach of treating IBS effects has been deployed. A new, modified approach has been introduced using the formalism of the Bjorken and Mtingwa model and the complete integrals of the second kind for faster numerical evaluation. This IBS kick is implemented in PyORBIT and extensive benchmarking cases against analytical models are shown. Results of combined space charge and intra-beam scattering simulations for the SPS and LEIR are presented and compared with observations from beam measurements., Comment: 13 pages, 13 figures, to be submitted to Physical Review Accelerators and Beams

Computed Tomography (CT) is commonly used to image acute ischemic stroke (AIS) patients, but its interpretation by radiologists is time-consuming and subject to inter-observer variability. Deep learning (DL) techniques can provide automated CT brain scan assessment, but usually require annotated images. Aiming to develop a DL method for AIS using labelled but not annotated CT brain scans from patients with AIS, we designed a convolutional neural network-based DL algorithm using routinely-collected CT brain scans from the Third International Stroke Trial (IST-3), which were not acquired using strict research protocols. The DL model aimed to detect AIS lesions and classify the side of the brain affected. We explored the impact of AIS lesion features, background brain appearances, and timing on DL performance. From 5772 unique CT scans of 2347 AIS patients (median age 82), 54% had visible AIS lesions according to expert labelling. Our best-performing DL method achieved 72% accuracy for lesion presence and side. Lesions that were larger (80% accuracy) or multiple (87% accuracy for two lesions, 100% for three or more), were better detected. Follow-up scans had 76% accuracy, while baseline scans 67% accuracy. Chronic brain conditions reduced accuracy, particularly non-stroke lesions and old stroke lesions (32% and 31% error rates respectively). DL methods can be designed for AIS lesion detection on CT using the vast quantities of routinely-collected CT brain scan data. Ultimately, this should lead to more robust and widely-applicable methods., Comment: 12 pages, 5 figures

Despite the large amount of brain CT data generated in clinical practice, the availability of CT datasets for deep learning (DL) research is currently limited. Furthermore, the data can be insufficiently or improperly prepared for machine learning and thus lead to spurious and irreproducible analyses. This lack of access to comprehensive and diverse datasets poses a significant challenge for the development of DL algorithms. In this work, we propose a complete semi-automatic pipeline to address the challenges of preparing a clinical brain CT dataset for DL analysis and describe the process of standardising this heterogeneous dataset. Challenges include handling image sets with different orientations (axial, sagittal, coronal), different image types (to view soft tissues or bones) and dimensions, and removing redundant background. The final pipeline was able to process 5,868/10,659 (45%) CT image datasets. Reasons for rejection include non-axial data (n=1,920), bone reformats (n=687), separated skull base/vault images (n=1,226), and registration failures (n=465). Further format adjustments, including image cropping, resizing and scaling are also needed for DL processing. Of the axial scans that were not localisers, bone reformats or split brains, 5,868/6,333 (93%) were accepted, while the remaining 465 failed the registration process. Appropriate preparation of medical imaging datasets for DL is a costly and time-intensive process., Comment: 9 pages, 5 figures

Recent developments in the field of gravitational physics, including the emergence of gravitational wave astronomy, black hole images, and more accurate telescopes, have allowed us to probe the strong-field character of gravity in a novel and revolutionary manner. This accessibility related to strong gravity brings into the foreground discussions about potential modifications to General Relativity (GR) that are particularly relevant in high curvature regimes. The most straightforward way to generalise GR is to consider an additional degree of freedom, in the form of a scalar field. In this thesis, we study generalised scalar tensor theories that predict interesting strong-gravity phenomenology. First, we review scalar no-hair theorems and the conditions under which they can be evaded. Next, we study solutions of black holes with scalar hair and the way in which higher derivative terms alter their properties. We then move our discussion to the spontaneously scalarized solutions, which only deviate from GR in the strong-field regime. We propose a model consistent with compact object scalarization, that allows for a GR attractor at late times, without fine-tuning (EsRGB model). Then, we proceed to study properties of black holes and neutron stars in this theory, revealing the interesting phenomenology of the solutions. We also study the radial stability of black holes in EsRGB and perform a preliminary analysis of the hyperbolicity of the problem. Finally, we take a look at the shadows of black holes and wormholes in theories with scalar fields, in light of recent observations of black hole shadows., Comment: 132 pages, Ph.D. thesis

We address the problem of learning new classes for semantic segmentation models from few examples, which is challenging because of the following two reasons. Firstly, it is difficult to learn from limited novel data to capture the underlying class distribution. Secondly, it is challenging to retain knowledge for existing classes and to avoid catastrophic forgetting. For learning from limited data, we propose a pseudo-labeling strategy to augment the few-shot training annotations in order to learn novel classes more effectively. Given only one or a few images labeled with the novel classes and a much larger set of unlabeled images, we transfer the knowledge from labeled images to unlabeled images with a coarse-to-fine pseudo-labeling approach in two steps. Specifically, we first match each labeled image to its nearest neighbors in the unlabeled image set at the scene level, in order to obtain images with a similar scene layout. This is followed by obtaining pseudo-labels within this neighborhood by applying classifiers learned on the few-shot annotations. In addition, we use knowledge distillation on both labeled and unlabeled data to retain knowledge on existing classes. We integrate the above steps into a single convolutional neural network with a unified learning objective. Extensive experiments on the Cityscapes and KITTI datasets validate the efficacy of the proposed approach in the self-driving domain. Code is available from https://github.com/ChasonJiang/FSCILSS., Comment: Paper accepted by 4th International Conference on Information Science, Parallel and Distributed Systems (ISPDS 2023)

Binary systems in which a neutron star or black hole accretes material from a high-mass star are known as high-mass X-ray binaries (HMXBs). This chapter provides a brief introduction to the physics of wind accretion and an observational view of HMXBs, including their classification, X-ray spectra, X-ray variability, orbital and compact object properties, as well as studies of Galactic and Magellanic HMXB populations. Two classes of X-ray sources whose possible connections to HMXBs have been debated, ultraluminous X-ray sources and gamma-ray binaries, are also discussed. Approximately 300 HMXBs residing either in the Milky Way or the Magellanic Clouds have been discovered. The majority of these HMXBs host wind-accreting neutron stars. Their X-ray properties depend both on the interaction of the accreting material with the neutron star's strong magnetic field and the properties of the donor star's wind. Most HMXBs are classified as either supergiant XBs or Be XBs based on the spectral type of the donor star; these classes exhibit different patterns of X-ray variability and occupy different phase space in diagrams of neutron star spin versus orbital period. While studies of HMXBs in the Milky Way and Magellanic Clouds find that their luminosity functions have similar shapes, an overabundance of Be XBs in the Small Magellanic Cloud points to important variations of the HMXB population with metallicity and age., Comment: Invited chapter for Handbook of X-ray and Gamma-ray Astrophysics, 44 pages of text, 11 pages of references, 18 figures, 1 table

Future ultra-low emittance rings for electron/positron colliders require extremely high beam brightness and can thus be limited by collective effects. In this paper, the interplay of effects such as synchrotron radiation, intra-beam scattering (IBS) and space charge in the vicinity of excited betatron resonances is assessed. In this respect, two algorithms were developed to simulate IBS and synchrotron radiation effects and integrated in the PyORBIT tracking code, to be combined with its widely used space charge module. The impact of these effects on the achievable beam parameters of the Compact Linear Collider (CLIC) Damping Rings was studied, showing that synchrotron radiation damping mitigates the adverse effects of IBS and space charge induced resonance crossing. The studies include also a full dynamic simulation of the CLIC damping ring cycle starting from the injection beam parameters. It is demonstrated that a careful working point choice is necessary, in order to accommodate the transition from a non-linear lattice induced detuning to a space-charge dominated one and thereby avoid excessive losses and emittance growth generated in the vicinity of strong resonances., Comment: To be published in Physical Review Accelerators and Beams

Shared Mobility Services (SMS), e.g., Demand-Responsive Transit (DRT) or ride-sharing, can improve mobility in low-density areas, often poorly served by conventional Public Transport (PT). Such improvement is mostly quantified via basic performance indicators, like wait or travel time. However, accessibility indicators, measuring the ease of reaching surrounding opportunities (e.g., jobs, schools, shops, ...), would be a more comprehensive indicator. To date, no method exists to quantify the accessibility of SMS based on empirical measurements. Indeed, accessibility is generally computed on graph representations of PT networks, but SMS are dynamic and do not follow a predefined network. We propose a spatial-temporal statistical method that takes as input observed trips of a SMS acting as a feeder for PT and summarized such trips in a graph. On such a graph, we compute classic accessibility indicators. We apply our method to a MATSim simulation study concerning DRT in Paris-Saclay.

We present multi-wavelength characterization of 65 high mass X-ray binary (HMXB) candidates in M33. We use the Chandra ACIS survey of M33 (ChASeM33) catalog to select hard X-ray point sources that are spatially coincident with UV-bright point source optical counterparts in the Panchromatic Hubble Andromeda Treasury: Triangulum Extended Region (PHATTER) catalog, which covers the inner disk of M33 at near infrared, optical, and near ultraviolet wavelengths. We perform spectral energy distribution (SED) fitting on multi-band photometry for each point source optical counterpart to measure its physical properties including mass, temperature, luminosity, and radius. We find that the majority of the HMXB companion star candidates are likely B-type main sequence stars, suggesting that the HMXB population of M33 is dominated by Be-XRBs, as is seen in other Local Group galaxies. We use spatially-resolved recent star formation history (SFH) maps of M33 to measure the age distribution of the HMXB candidate sample and the HMXB production rate for M33. We find a bimodal distribution for the HMXB production rate over the last 80 Myr, with a peak at $\sim$10 Myr and $\sim$40 Myr, which match theoretical formation timescales for the most massive HMXBs and Be X-ray binaries (Be-XRBs), respectively. We measure an HMXB production rate of 107$-$136 HMXBs/(M$_{\odot}$ yr$^{-1}$) over the last 50 Myr and 150$-$199 HMXBs/(M$_{\odot}$ yr$^{-1}$) over the last 80 Myr. For sources with compact object classifications from overlapping NuSTAR observations, we find a preference for giant/supergiant companion stars in BH-HMXBs and main sequence companion stars in neutron star HMXBs (NS-HMXBs)., Comment: Accepted for publication in ApJ, 37 pages, 6 figures, 8 tables