Search

Biomedical research requires large, diverse samples to produce unbiased results. Automated methods for matching variables across datasets can accelerate this process. Research in this area has been limited, primarily focusing on lexical matching and ontology based semantic matching. We aimed to develop new methods, leveraging large language models (LLM) and ensemble learning, to automate variable matching. Methods: We utilized data from two GERAS cohort (European and Japan) studies to develop variable matching methods. We first manually created a dataset by matching 352 EU variables with 1322 candidate JP variables, where matched variable pairs were positive and unmatched pairs were negative instances. Using this dataset, we developed and evaluated two types of natural language processing (NLP) methods, which matched variables based on variable labels and definitions from data dictionaries: (1) LLM-based and (2) fuzzy matching. We then developed an ensemble-learning method, using the Random Forest model, to integrate individual NLP methods. RF was trained and evaluated on 50 trials. Each trial had a random split (4:1) of training and test sets, with the model's hyperparameters optimized through cross-validation on the training set. For each EU variable, 1322 candidate JP variables were ranked based on NLP-derived similarity scores or RF's probability scores, denoting their likelihood to match the EU variable. Ranking performance was measured by top-n hit ratio (HRn) and mean reciprocal rank (MRR). Results:E5 performed best among individual methods, achieving 0.90 HR-30 and 0.70 MRR. RF performed better than E5 on all metrics over 50 trials (P less than 0.001) and achieved an average HR 30 of 0.98 and MRR of 0.73. LLM-derived features contributed most to RF's performance. One major cause of errors in automatic variable matching was ambiguous variable definitions within data dictionaries., Comment: 32 pages, 2 figures

The combined use of data from different sources can be critical in emergencies, where accurate models are needed to make real-time decisions, but high-fidelity representations and detailed information are simply unavailable. This study presents a data assimilation framework based on an ensemble Kalman filter that sequentially exploits and improves an advection-diffusion model in a case study concerning an airborne contaminant dispersion problem over a complex two-dimensional domain. An autonomous aerial drone is used to sequentially observe the actual contaminant concentration in a small fraction of the domain, orders of magnitude smaller than the total domain area. Such observations are synchronized with the data assimilation framework, iteratively adjusting the simulation. The path of the drone is sequentially optimized by balancing exploration and exploitation according to the available knowledge at each decision time. Starting from an erroneous initial model based on approximated assumptions that represent the limited initial knowledge available during emergency scenarios, results show how the proposed framework sequentially improves its belief about the dispersion dynamics, thus providing a reliable contaminant concentration map.

A digital twin framework for rapid predictions of atmospheric contaminant dispersion is developed to support informed decision making in emergency situations. In an offline preparation phase, the geometry of a built environment is discretized with a finite element (FEM) mesh and a reduced-order model (ROM) of the steady-state incompressible Navier-Stokes equations is constructed for various wind conditions. Subsequently, the ROM provides a fast wind field estimate based on the current wind speed during the online phase. To support crisis management, several methodological building blocks are combined. Automatic FEM meshing of built environments and numerical flow solver capabilities enable fast forward-simulations of contaminant dispersion using the advection-diffusion equation as transport model. Further methods are integrated in the framework to address inverse problems such as contaminant source localization based on sparse concentration measurements. Additionally, the contaminant dispersion model is coupled with a continuum-based pedestrian crowd model to derive fast and safe evacuation routes for people seeking protection during contaminant dispersion emergencies. The interplay of these methods is demonstrated in two critical infrastructure protection (CIP) test cases. Based on simulated real world interaction (measurements, communication), this article demonstrates a full Measurement-Inversion-Prediction-Steering (MIPS) cycle including a Bayesian formulation of the inverse problem.

Labor market tightness tremendously increased in Germany between 2012 and 2022. We analyze the effect of tightness on wages by combining social security data with unusually rich information on vacancies and job seekers. Instrumental variable regressions reveal positive elasticities between 0.004 and 0.011, implying that higher tightness explains between 7 and 19 percent of the real wage increase. We report greater elasticities for new hires, high-skilled workers, the Eastern German labor market, and the service sector. In particular, tightness raised wages at the bottom of the wage distribution, contributing to the decline in wage inequality over the last decade.

Entanglement distillation, the process of converting weakly entangled states into maximally entangled ones using Local Operations and Classical Communication (LOCC), is pivotal for robust entanglement-assisted quantum information processing in error-prone environments. A construction based on stabilizer codes offers an effective method for designing such protocols. By analytically investigating the effective action of stabilizer protocols for codes in arbitrary dimensions, we establish a standard form for the output states of recurrent stabilizer-based distillation. This links the properties of input states, stabilizers, and encodings to the properties of the protocol. Based on those insights, we present a novel two-copy distillation protocol, applicable to all bipartite states, that maximizes the fidelity increase per iteration for Bell-diagonal states in any prime dimension. The power of this framework and the protocol is demonstrated through numerical investigations, which provide evidence of its superior performance in terms of efficiency and distillability of low-fidelity states compared to other well-established recurrence protocols. By elucidating the interplay between states, errors, and protocols, our contribution advances the systematic development of highly effective distillation protocols across arbitrary dimensions, enhancing our understanding of distillability., Comment: 20 pages, 9 figures

The probabilities for gamma-ray and particle emission as a function of the excitation energy of a decaying nucleus are valuable observables for constraining the ingredients of the models that describe the de-excitation of nuclei near the particle emission threshold. These models are essential in nuclear astrophysics and applications. In this work, we have for the first time simultaneously measured the gamma-ray and neutron emission probabilities of 208Pb. The measurement was performed in inverse kinematics at the Experimental Storage Ring (ESR) of the GSI/FAIR facility, where a 208Pb beam interacted through the 208Pb(p,p') reaction with a hydrogen gas jet target. Instead of detecting the gamma-rays and neutrons emitted by 208Pb, we detected the heavy beam-like residues produced after gamma and neutron emission. These heavy residues were fully separated by a dipole magnet of the ESR and were detected with outstanding efficiencies. The comparison of the measured probabilities with model calculations has allowed us to test and select different descriptions of the gamma-ray strength function and the nuclear level density available in the literature., Comment: 23 pages, 14 figures

Coronary angioplasty with stent implantation is the most frequently used interventional treatment for coronary artery disease. However, reocclusion within the stent, referred to as in-stent restenosis, occurs in up to 10% of lesions. It is widely accepted that mechanical loads on the vessel wall strongly affect adaptive and maladaptive mechanisms. Yet, the role of procedural and lesion-specific influence on restenosis risk remains understudied. Computational modeling of the stenting procedure can provide new mechanistic insights, such as local stresses, that play a significant role in tissue growth and remodeling. Previous simulation studies often featured simplified artery and stent geometries and cannot be applied to real-world examples. Realistic simulations were computationally expensive since they featured fully resolved stenting device models. The aim of this work is to develop and present a mixed-dimensional formulation to simulate the patient-specific stenting procedure with a reduced-dimensional beam model for the stent and 3D models for the artery. In addition to presenting the numerical approach, we apply it to realistic cases to study the intervention's mechanical effect on the artery and correlate the findings with potential high-risk locations for in-stent restenosis. We found that high artery wall stresses develop during the coronary intervention in severely stenosed areas and at the stent boundaries. Herewith, we lay the groundwork for further studies towards preventing in-stent restenosis after coronary angioplasty.

Despite the evidence that supermassive black holes (SMBHs) co-evolve with their host galaxy, and that most of the growth of these SMBHs occurs via merger-free processes, the underlying mechanisms which drive this secular co-evolution are poorly understood. We investigate the role that both strong and weak large-scale galactic bars play in mediating this relationship. Using 72,940 disc galaxies in a volume-limited sample from Galaxy Zoo DESI, we analyse the active galactic nucleus (AGN) fraction in strongly barred, weakly barred, and unbarred galaxies up to z = 0.1 over a range of stellar masses and colours. After controlling for stellar mass and colour, we find that the optically selected AGN fraction is 31.6 +/- 0.9 per cent in strongly barred galaxies, 23.3 +/- 0.8 per cent in weakly barred galaxies, and 14.2 +/- 0.6 per cent in unbarred disc galaxies. These are highly statistically robust results, strengthening the tantalising results in earlier works. Strongly barred galaxies have a higher fraction of AGNs than weakly barred galaxies, which in turn have a higher fraction than unbarred galaxies. Thus, while bars are not required in order to grow a SMBH in a disc galaxy, large-scale galactic bars appear to facilitate AGN fuelling, and the presence of a strong bar makes a disc galaxy more than twice as likely to host an AGN than an unbarred galaxy at all galaxy stellar masses and colours., Comment: 11 pages, 8 figures, accepted for publication in MNRAS

: Non-resonant background (NRB) plays a significant role in coherent anti-Stokes Raman scattering (CARS) spectroscopic applications. All the recent works primarily focused on removing the NRB using different deep learning methods, and only one study explored the effect of NRB. Hence, in this work, we systematically investigated the impact of NRB variation on Raman signal retrieval. The NRB is simulated as a linear function with different strengths relative to the resonant Raman signal, and the variance also changed for each NRB strength. The resonant part of nonlinear susceptibility is extracted from real experimental Raman data; hence, the simulated CARS data better approximate the experimental CARS spectra. Then, the corresponding Raman signal is retrieved by four different methods: maximum entropy method (MEM), Kramers-Kronig (KK), convolutional neural network (CNN), and long short-term memory (LSTM) network. Pearson correlation measurements and principal component analysis combined with linear discriminant analysis (PCA-LDA) modelling revealed that MEM and KK methods have an edge over LSTM and CNN for higher NRB strengths. It is also demonstrated that normalizing the input data favors LSTM and CNN predictions. In contrast, background removal from the predictions significantly influenced Pearson correlation but not the classification accuracies for MEM and KK. This comprehensive study is done for the first time to the best of our knowledge and has the potential to impact the CARS spectroscopy and microscopy applications in different areas., Comment: 15 pages, 7 figures

Bound entanglement is a special form of quantum entanglement that cannot be used for distillation, i.e., the local transformation of copies of arbitrarily entangled states into a smaller number of approximately maximally entangled states. Implying an inherent irreversibility of quantum resources, this phenomenon highlights the gaps in our current theory of entanglement. This review provides a comprehensive exploration of the key findings on bipartite bound entanglement. We focus on systems of finite dimensions, an area of high relevance for many quantum information processing tasks. We elucidate the properties of bound entanglement and its interconnections with various facets of quantum information theory and quantum information processing. The article illuminates areas where our understanding of bound entangled states, particularly their detection and characterization, is yet to be fully developed. By highlighting the need for further research into this phenomenon and underscoring relevant open questions, this article invites researchers to unravel its relevance for our understanding of entanglement in Nature and how this resource can most effectively be used for applications in quantum technology., Comment: Review Paper, 62 pages, 9 figures, comments from colleagues added (version 2)

Sufficiently clean geometrically frustrated (GF) magnets are the largest class of candidate materials that may host quantum spin liquids (QSLs). Some of them have been shown to exhibit spin-glass freezing, potentially precluding QSLs, at the "hidden energy scale", which is significantly lower than the microscopic energy scale of spin interactions. Here, we investigate the origin of the hidden energy scale and its relationship to the $f$-ratio, the figure of merit for the degree of frustration in GF magnetic materials. The available experimental and numerical data provide evidence that GF magnets display, universally, two distinct temperature scales in the specific heat, the lowest of which is of the order of the hidden energy scale $T^*$. We argue that this scale is determined by non-magnetic excitations, similar to spin exchanges in chains of spins. The collective entropy of such excitations matches the entropy of the ground states of the Ising model on the same lattice, which provides a way to verify the proposed scenario in experiment. We demonstrate that in the presence of quenched disorder, a broad class of materials exhibits spin-glass freezing at temperatures of order $T^*$, in accordance with experimental observations. As $T^*$ is a property of the clean GF medium, it leads to a constraint on the $f$-ratio., Comment: 4.5+4.5 pages, 2+2 figures, 2 tables

Conversational AI (CAI) systems which encompass voice- and text-based assistants are on the rise and have been largely integrated into people's everyday lives. Despite their widespread adoption, users voice concerns regarding privacy, security and trust in these systems. However, the composition of these perceptions, their impact on technology adoption and usage and the relationship between privacy, security and trust perceptions in the CAI context remain open research challenges. This study contributes to the field by conducting a Systematic Literature Review and offers insights into the current state of research on privacy, security and trust perceptions in the context of CAI systems. The review covers application fields and user groups and sheds light on empirical methods and tools used for assessment. Moreover, it provides insights into the reliability and validity of privacy, security and trust scales, as well as extensively investigating the subconstructs of each item as well as additional concepts which are concurrently collected. We point out that the perceptions of trust, privacy and security overlap based on the subconstructs we identified. While the majority of studies investigate one of these concepts, only a few studies were found exploring privacy, security and trust perceptions jointly. Our research aims to inform on directions to develop and use reliable scales for users' privacy, security and trust perceptions and contribute to the development of trustworthy CAI systems., Comment: 39 pages, 4 figures

Background: The aim of this study was to investigate the role of clinical, dosimetric and pretherapeutic magnetic resonance imaging (MRI) features for lesion-specific outcome prediction of stereotactic radiotherapy (SRT) in patients with brain metastases from malignant melanoma (MBM). Methods: In this multicenter, retrospective analysis, we reviewed 517 MBM from 130 patients treated with SRT (single fraction or hypofractionated). For each gross tumor volume (GTV) 1576 radiomic features (RF) were calculated (788 each for the GTV and for a 3 mm margin around the GTV). Clinical parameters, radiation dose and RF from pretherapeutic contrast-enhanced T1-weighted MRI from different institutions were evaluated with a feature processing and elimination pipeline in a nested cross-validation scheme. Results: Seventy-two (72) of 517 lesions (13.9%) showed a local failure (LF) after SRT. The processing pipeline showed clinical, dosimetric and radiomic features providing information for LF prediction. The most prominent ones were the correlation of the gray level co-occurrence matrix of the margin (hazard ratio (HR): 0.37, confidence interval (CI): 0.23-0.58) and systemic therapy before SRT (HR: 0.55, CI: 0.42-0.70). The majority of RF associated with LF was calculated in the margin around the GTV. Conclusions: Pretherapeutic MRI based RF connected with lesion-specific outcome after SRT could be identified, despite multicentric data and minor differences in imaging protocols. Image data analysis of the surrounding metastatic environment may provide therapy-relevant information with the potential to further individualize radiotherapy strategies.

The current standard model of cosmology successfully describes a variety of measurements, but the nature of its main ingredients, dark matter and dark energy, remains unknown. Euclid is a medium-class mission in the Cosmic Vision 2015-2025 programme of the European Space Agency (ESA) that will provide high-resolution optical imaging, as well as near-infrared imaging and spectroscopy, over about 14,000 deg^2 of extragalactic sky. In addition to accurate weak lensing and clustering measurements that probe structure formation over half of the age of the Universe, its primary probes for cosmology, these exquisite data will enable a wide range of science. This paper provides a high-level overview of the mission, summarising the survey characteristics, the various data-processing steps, and data products. We also highlight the main science objectives and expected performance., Comment: Accepted for publication in the A&A special issue`Euclid on Sky'

Urban populations continue to grow, highlighting the critical need to safeguard civilians against potential disruptions, such as dangerous gas contaminant dispersion. The digital twin (DT) framework offers promise in analyzing and predicting such events. This study presents a computational framework for modelling airborne contaminant dispersion in built environments. Leveraging automatic generation of computational domains and solution processes, the proposed framework solves the underlying physical model equations with the finite element method (FEM) for numerical solutions. Model order reduction (MOR) methods are investigated to enhance computational efficiency without compromising accuracy. The study outlines the automatic model generation process, the details of the employed model, and the future perspectives for the realization of a DT. Throughout this research, the aim is to develop a reliable predictive model combining physics and data in a hybrid DT to provide informed real-time support within evacuation scenarios.

In this study, we investigate the potential of fast-to-evaluate surrogate modeling techniques for developing a hybrid digital twin of a steel-reinforced concrete beam, serving as a representative example of a civil engineering structure. As surrogates, two distinct models are developed utilizing physics-informed neural networks, which integrate experimental data with given governing laws of physics. The experimental data (sensor data) is obtained from a previously conducted four-point bending test. The first surrogate model predicts strains at fixed locations along the center line of the beam for various time instances. This time-dependent surrogate model is inspired by the motion of a harmonic oscillator. For this study, we further compare the physics-based approach with a purely data-driven method, revealing the significance of physical laws for the extrapolation capabilities of models in scenarios with limited access to experimental data. Furthermore, we identify the natural frequency of the system by utilizing the physics-based model as an inverse solver. For the second surrogate model, we then focus on a fixed instance in time and combine the sensor data with the equations of linear elasticity to predict the strain distribution within the beam. This example reveals the importance of balancing different loss components through the selection of suitable loss weights.

Multi-modal foundation models such as CLIP have showcased impressive zero-shot capabilities. However, their applicability in resource-constrained environments is limited due to their large number of parameters and high inference time. While existing approaches have scaled down the entire CLIP architecture, we focus on training smaller variants of the image encoder, which suffices for efficient zero-shot classification. The use of synthetic data has shown promise in distilling representations from larger teachers, resulting in strong few-shot and linear probe performance. However, we find that this approach surprisingly fails in true zero-shot settings when using contrastive losses. We identify the exploitation of spurious features as being responsible for poor generalization between synthetic and real data. However, by using the image feature-based L2 distillation loss, we mitigate these problems and train students that achieve zero-shot performance which on four domain-specific datasets is on-par with a ViT-B/32 teacher model trained on DataCompXL, while featuring up to 92% fewer parameters.

Monoclinic $\beta$-Ga$_2$O$_3$ films grown on $c$-plane sapphire have been shown to exhibit six $(\bar{2}01)$-plane oriented domains, which are equally-spaced-by-rotation around the surface normal and equally-sized-by-volume that render the film optical response effectively uniaxial. We derive and discuss an optical model suitable for ellipsometry data analysis of such films. We model mid- and far-infrared ellipsometry data from undoped and electrically insulating films with an effective uniaxial dielectric tensor based on projections of all phonon modes within the rotation domains parallel and perpendicular to the sample normal, i.e., to the reciprocal lattice vector $\mathbf{g}_{\bar{2}01}$. Two effective response functions are described by model, and found sufficient to calculate ellipsometry data that best-match measured ellipsometry data from a representative film. We propose to render either effective dielectric functions, or inverse effective dielectric functions, each separately for electric field directions parallel and perpendicular to $\mathbf{g}_{\bar{2}01}$, by sums of Lorentz oscillators, which permit to determine either sets of transverse optical phonon mode parameters, or sets of longitudinal optical phonon mode parameters, respectively. Transverse optical modes common to both dielectric functions can be traced back to single crystal modes with $B_{\mathrm{u}}$ character, while modes with $A_{\mathrm{u}}$ character only appear within the dielectric function for polarization perpendicular to the sample surface. The thereby obtained parameter sets reveal all phonon modes anticipated from averaging over the six-fold rotation domains of single crystal $\beta$-Ga$_2$O$_3$, but with slightly shifted transverse optical, and completely different longitudinal optical phonon modes., Comment: 14 pgaes, 8 figures

We present the first systematic investigation of supervised scaling laws outside of an ImageNet-like context - on images of galaxies. We use 840k galaxy images and over 100M annotations by Galaxy Zoo volunteers, comparable in scale to Imagenet-1K. We find that adding annotated galaxy images provides a power law improvement in performance across all architectures and all tasks, while adding trainable parameters is effective only for some (typically more subjectively challenging) tasks. We then compare the downstream performance of finetuned models pretrained on either ImageNet-12k alone vs. additionally pretrained on our galaxy images. We achieve an average relative error rate reduction of 31% across 5 downstream tasks of scientific interest. Our finetuned models are more label-efficient and, unlike their ImageNet-12k-pretrained equivalents, often achieve linear transfer performance equal to that of end-to-end finetuning. We find relatively modest additional downstream benefits from scaling model size, implying that scaling alone is not sufficient to address our domain gap, and suggest that practitioners with qualitatively different images might benefit more from in-domain adaption followed by targeted downstream labelling., Comment: 10+6 pages, 12 figures. Appendix C2 based on arxiv:2206.11927. Code, demos, documentation at https://github.com/mwalmsley/zoobot

The anisotropic thermal conductivity and the phonon mean free path (mfp) in monoclinic $\beta$-Ga$_2$O$_3$ single crystals and homoepitaxial films of several $\mu$m were determined using the 3$\omega$-method in the temperature range from 10K-300 K. The measured effective thermal conductivity of both, single crystal and homoepitaxial films are in the order of 20 W/(mK) at room temperature, below 30 K it increases with a maximum of 1000 to 2000 W/(mK) and decreases with T$^3$ below 25 K. Analysis of the phonon mfp shows a dominance of phonon-phonon-Umklapp scattering above 80 K, below which the influence of point-defect scattering is observed. Below 30 K the phonon mfp increases until it is limited by the total $\beta$-Ga$_2$O$_3$ sample size. A crossover from resistive to ballistic phonon transport is observed below 20 K and boundary effects of the total sample size become dominant. This reveals that the homoepitaxial film-substrate interface is highly phonon-transparent. The resistive and ballistic phonon transport regimes in $\beta$-Ga$_2$O$_3$ are discussed corresponding to the models of Callaway and Majumdar, respectively., Comment: 16 pages, 5 figures, plus supplementary material

This paper presents a scalable physics-based block preconditioner for mixed-dimensional models in beam-solid interaction and their application in engineering. In particular, it studies the linear systems arising from a regularized mortar-type approach for embedding geometrically exact beams into solid continua. Due to the lack of block diagonal dominance of the arising 2 x 2 block system, an approximate block factorization preconditioner is used. It exploits the sparsity structure of the beam sub-block to construct a sparse approximate inverse, which is then not only used to explicitly form an approximation of the Schur complement, but also acts as a smoother within the prediction step of the arising SIMPLE-type preconditioner. The correction step utilizes an algebraic multigrid method. Although, for now, the beam sub-block is tackled by a one-level method only, the multi-level nature of the computationally demanding correction step delivers a scalable preconditioner in practice. In numerical test cases, the influence of different algorithmic parameters on the quality of the sparse approximate inverse is studied and the weak scaling behavior of the proposed preconditioner on up to 1000 MPI ranks is demonstrated, before the proposed preconditioner is finally applied for the analysis of steel-reinforced concrete structures in civil engineering.

Predicting the long-term success of endovascular interventions in the clinical management of cerebral aneurysms requires detailed insight into the patient-specific physiological conditions. In this work, we not only propose numerical representations of endovascular medical devices such as coils, flow diverters or Woven EndoBridge but also outline numerical models for the prediction of blood flow patterns in the aneurysm cavity right after a surgical intervention. Detailed knowledge about the post-surgical state then lays the basis to assess the chances of a stable occlusion of the aneurysm required for a long-term treatment success. To this end, we propose mathematical and mechanical models of endovascular medical devices made out of thin metal wires. These can then be used for fully resolved flow simulations of the post-surgical blood flow, which in this work will be performed by means of a Lattice Boltzmann method applied to the incompressible Navier-Stokes equations and patient-specific geometries. To probe the suitability of homogenized models, we also investigate poro-elastic models to represent such medical devices. In particular, we examine the validity of this modeling approach for flow diverter placement across the opening of the aneurysm cavity. For both approaches, physiologically meaningful boundary conditions are provided from reduced-order models of the vascular system. The present study demonstrates our capabilities to predict the post-surgical state and lays a solid foundation to tackle the prediction of thrombus formation and, thus, the aneurysm occlusion in a next step.

Neutron-induced reaction cross sections of short-lived nuclei are imperative to understand the origin of heavy elements in stellar nucleosynthesis and for societal applications, but their measurement is extremely complicated due to the radioactivity of the targets involved. One way of overcoming this issue is to combine surrogate reactions with the unique possibilities offered by heavy-ion storage rings. In this work, we describe the first surrogate-reaction experiment in inverse kinematics, which we successfully conducted at the Experimental Storage Ring (ESR) of the GSI/FAIR facility, using the 208Pb(p,p') reaction as a surrogate for neutron capture on 207Pb. Thanks to the outstanding detection efficiencies possible at the ESR, we were able to measure for the first time the neutron-emission probability as a function of the excitation energy of 208Pb. We have used this probability to select different descriptions of the gamma-ray strength function and nuclear level density, and provide reliable results for the neutron-induced radiative capture cross section of 207Pb at energies for which no experimental data exist., Comment: 10 pages and 5 figures

Giant Star-forming Clumps (GSFCs) are areas of intensive star-formation that are commonly observed in high-redshift (z>1) galaxies but their formation and role in galaxy evolution remain unclear. High-resolution observations of low-redshift clumpy galaxy analogues are rare and restricted to a limited set of galaxies but the increasing availability of wide-field galaxy survey data makes the detection of large clumpy galaxy samples increasingly feasible. Deep Learning, and in particular CNNs, have been successfully applied to image classification tasks in astrophysical data analysis. However, one application of DL that remains relatively unexplored is that of automatically identifying and localising specific objects or features in astrophysical imaging data. In this paper we demonstrate the feasibility of using Deep learning-based object detection models to localise GSFCs in astrophysical imaging data. We apply the Faster R-CNN object detection framework (FRCNN) to identify GSFCs in low redshift (z<0.3) galaxies. Unlike other studies, we train different FRCNN models not on simulated images with known labels but on real observational data that was collected by the Sloan Digital Sky Survey Legacy Survey and labelled by volunteers from the citizen science project `Galaxy Zoo: Clump Scout'. The FRCNN model relies on a CNN component as a `backbone' feature extractor. We show that CNNs, that have been pre-trained for image classification using astrophysical images, outperform those that have been pre-trained on terrestrial images. In particular, we compare a domain-specific CNN -`Zoobot' - with a generic classification backbone and find that Zoobot achieves higher detection performance and also requires smaller training data sets to do so. Our final model is capable of producing GSFC detections with a completeness and purity of >=0.8 while only being trained on ~5,000 galaxy images., Comment: Accepted for publication in RASTI, 22 pages

Privacy policies play a vital role in safeguarding user privacy as legal jurisdictions worldwide emphasize the need for transparent data processing. While the suitability of privacy policies to enhance transparency has been critically discussed, employing conversational AI systems presents unique challenges in informing users effectively. In this position paper, we propose a dynamic workflow for transforming privacy policies into privacy question-and-answer (Q&A) pairs to make privacy policies easily accessible through conversational AI. Thereby, we facilitate interdisciplinary collaboration among legal experts and conversation designers, while also considering the utilization of large language models' generative capabilities and addressing associated challenges. Our proposed workflow underscores continuous improvement and monitoring throughout the construction of privacy Q&As, advocating for comprehensive review and refinement through an experts-in-the-loop approach., Comment: Position paper presented at CONVERSATIONS 2023 - the 7th International Workshop on Chatbot Research and Design, hosted by the University of Oslo, Norway, November 22-23, 2023