335598 research outputs found
Sort by
Erklärbare Künstliche Intelligenz - Steigerung der Nachvollziehbarkeit überwachter maschineller Lernverfahren
No full textKünstliche Intelligenz und maschinelles Lernen gewinnen zunehmend an Bedeutung, sind jedoch aufgrund der Komplexität oft nur eingeschränkt nachvollziehbar. Die erklärbare Künstliche Intelligenz setzt genau hier an. Diese Arbeit leistet einen Beitrag zu erklärbaren überwachten Lernverfahren durch ein Vorgehensmodell mit fünf Erklärungstypen sowie globale und lokale Surrogatmodelle, die sowohl Modelle als auch einzelne Vorhersagen verständlich machen und in Benutzerstudien evaluiert wurden
Trio induced cytoskeleton rearrangement drives endothelial cell enlargement and outward arterial remodeling
No full textDynamic N→B coordination and anion-selective turn-on fluorescence in oxadiazole-functionalized organoboranes
No full textA versatile route for the preparation of chemically and electronically diverse Mes2BPh-based boranes (Mes = mesityl, 2,4,6-trimethylphenyl) is presented that allows the conversion of tetrazolyl rings in a common borane precursor (2H) into boranes bearing variously substituted oxadiazolyl groups. A series of eight boranes (4a–4h) was prepared with functional groups on the 5-position of the oxadiazole ranging from electron donating (4a: 4-Me-phenyl, 4b: 4-MeO-phenyl, etc.) to strongly electron withdrawing (4d: 4-O2N-phenyl, 4e: 3,5-bis(CF3)-phenyl, 4h: CF3) and also including a bifunctionalized example bearing two Mes2B moieties (4g). A full characterization study of the optical, electrochemical and electronic properties, both experimentally and by DFT calculations, was carried out. Our investigation shows that the boranes exhibit dynamic equilibria between closed intramolecularly N→B-coordinated and open non-coordinated conformers, as indicated by variable temperature NMR, 11B NMR and anion binding studies with F− and CN−. The anion binding studies reveal substantial differences in the fluorescence response of the compounds ranging from differing degrees of quenching to fluorescence shifts (4g) and enhanced emission (4c) (4-OMe-phenyl). These results show that this synthetic strategy allows easy creation of a series of compounds with incrementally varied optical properties and Lewis acidities
Sensor Differences of Dual Sphere Superconducting Gravimeters
No full textWe analyse the difference signal of dual sphere superconducting gravimeters (SG) to identify systematic instrumental disturbances, which would otherwise go unnoticed. Compared to classical spring gravimeters, SGs excel by their superior long-term stability. However, SG measurements also suffer from systematic errors. One possibility for characterising and quantifying these errors is the analysis of sensor differences, either of dual sphere instruments or between two colocated instruments. For perfect instruments, the sensor difference should vanish. We study the sensor differences of all dual sphere SGs in the database of the International Geodynamics and Earth Tide Service (IGETS). As expected, they show a relative drift between the sensors and steps related to operator interventions. However, we can also identify periods of unexpected drift rate changes that last for several months. Afterwards, the drift rate returns to its old value. The observed differences are too big to be caused by local gravity gradients. Therefore, we think, they indicate more complex systematic disturbances of SGs. These disturbances are at the level of a few tens of nm/ and could not be clearly identified in the gravity residuals of only one sensor. These findings are corroborated by the analysis of differences between colocated single sphere SGs at the J9 observatory in Strasbourg. Knowledge of the characteristics and size of these disturbances is important if gravity changes of a few tens of nm/ are studied on long time scales, like signals from hydrology or polar motion
Where to harvest solar energy in Iran? A geographic information system (GIS) analysis for supporting the siting of photovoltaic (PV) parks and concentrating solar power (CSP) plants
No full textIran\u27s electricity generation relies heavily on fossil fuels, resulting in frequent power shortages and widespread blackouts in major cities. Given the high levels of solar irradiance across the country, photovoltaic (PV) and concentrating solar power (CSP) technologies could provide a sustainable alternative. Existing studies focus on specific technologies or individual regions. Currently, there is no consistent, comprehensive mapping of the scope for political decision-making in Iran. This study aims to address this issue by providing the first nationwide assessment of solar energy potential in Iran, evaluating both PV and CSP. This GIS-based assessment uses an expanded set of environmental and technical criteria and performs sensitivity analyses to ensure robust results and identify the most effective and sustainable locations for PV and CSP plants. The model incorporates specific constraints, such as protected natural areas, to exclude unsuitable sites, and assesses suitability based on criteria such as solar irradiation levels and proximity to grid infrastructure. These factors are categorised into four suitability classes, ranging from \u27high\u27 to \u27very low\u27 for both PV and CSP installations. By synthesising the constraint and suitability maps, the model identifies feasible sites and assesses their relative desirability. A sensitivity analysis, focusing on the weighting of the suitability criteria, confirms the robustness of the results. The results highlight Iran\u27s considerable capacity for solar power generation and suggest that the country could exceed its current electricity production by a multiple through the development of solar power plants. The model applies 14 exclusion criteria, revealing that 70% of Iran’s land is unsuitable for PV and 83% for CSP. The results show that 14.5% of Iran’s land is suitable for PV and 7.5% for CSP (medium and high suitable), with central and eastern regions offering the highest potential. Additionally, the study highlights the promising prospects of GIS modeling in renewable energy siting, emphasizing improved data integration, global scalability, environmental impact assessment, and policy harmonization
Impact of Connectivity-Preserving Loss Functions on the Segmentation of Thin Tubular Structures: Application to Coronary Arteries From CT Angiography Data
No full textCoronary artery disease remains a leading cause of mortality worldwide. Coronary Computed
Tomography Angiography (CCTA) provides a non-invasive basis for diagnosis; however, an
accurate and connectivity-preserving segmentation of the coronary artery tree is essential for
robust automatic and quantitative analyses. Convolutional Neural Networks (CNNs)-based
architectures, in particular U-Net and no-new-U-Net (nnU-Net), have shown outstanding
performance across a wide range of medical image segmentation benchmarks, yet they may
frequently produce fragmented vessel trees when segmenting thin, tubular structures such as
coronary arteries. Recent studies indicate that connectivity-aware loss functions can mitigate
these discontinuities by explicitly penalizing missing centerline segments, but their efficacy
for coronary artery tree segmentation remains to be demonstrated.
This thesis quantifies the benefits and challenges of integrating connectivity-preserving loss
functions into an nnU-Net-based pipeline for one-step coronary artery tree segmentation from
CCTA images. Performance is assessed using complementary metrics covering vessel mask
accuracy, vessel accuracy, centerline completeness, and runtime, capturing volumetric overlap,
connectivity-related effects, and computational cost. To contextualize the quantitative
results, we conduct qualitative case studies with targeted visualizations.
The comparison of connectivity-preserving losses reveals that the Skeleton Recall (SR)
loss provides the most consistent improvements in connectivity metrics while incurring
substantially lower training time than the other connectivity-preserving loss formulations. In
the final statistical analysis, augmenting the generic loss with a SR term improves coronary
artery tree connectivity in a statistically significant and practically relevant manner, without
substantially degrading volumetric overlap and with negligible computational overhead.
These findings identify SR as an effective, computationally efficient, and straightforward-to-implement loss function, making it a practically viable choice for accurate and connectivitypreserving coronary artery tree segmentation
Escape from a coupled bi-quartic potential well
No full textIn this paper, we investigate the dynamics of particles within a bi-quartic potential well, characterized by the coupled potential function . Our focus is on the safe basins of escape and level-crossing under arbitrary initial conditions, i.e., the spatial region of initial conditions from where an initiated motion of the particle remains bounded. The coupling term allows energy exchange between the modes. If the total energy is sufficient, a particle starting from a given set of initial conditions within the potential well can reach the escape boundary over time, which would not occur without coupling. We find that escape trajectories often pass near one of the four saddles of the potential. Numerical simulations reveal that the safe basins of escape have fractal boundaries due to the energy-exchange mechanism. To address safety-critical applications where these chaotic regimes must be avoided, we introduce a factor of safety that defines a safety region. Crossing the safety region\u27s boundary shifts the problem from escape to level-crossing. Assuming harmonic-like solutions of the differential equations with slowly varying amplitudes and phases, we transform our system into an appropriate form for averaging. By eliminating time as a variable and realizing that only the phase difference is significant, we derive two first integrals of the particle motion in analytic form, which allows us to analytically determine the safe region boundary and calculate its size based on the coupling parameter
Temperature station matching for elevation-standardised ecological meta-analysis
No full textDie Standardisierung von Temperaturdaten über heterogene Untersuchungsstandorte hinweg ist für ökologische Metaanalysen unerlässlich, doch durch Höhengradienten bedingte Temperaturunterschiede erschweren oft den direkten Vergleich von Klimadaten, insbesondere bei grober Rasterauflösung von Thermaldaten. Ökologische Studien erfassen häufig nur die Standorthöhe – insbesondere historische Datensätze –, was die Analyse thermischer Einflüsse auf die räumliche Verbreitung von Organismen einschränkt.
Es wurde ein Protokoll mit zwei Ansätzen entwickelt, um regionale Korrekturfaktoren (ΔH) aus Höhen-Temperatur-Regressionen (Lapse-Rate-Methode: Südwestdeutschland/Italienische Alpen, n = 33 Stationen) und regionenübergreifenden Stationspaaren (TAV-Matching-Methode, n = 27) mit eng aufeinander abgestimmten langfristigen Durchschnittstemperaturen (ΔTAV ≤ 1,2 °C) abzuleiten. Angewandt auf 109 Ixodes ricinus-Untersuchungsstandorte in neun europäischen Regionen wurden Korrekturfaktoren nur für Regionen mit konsistenten Höhenunterschieden (ΔH > 100 m) im Vergleich zu südwestdeutschen Referenzstationen berechnet.
Die regionalen Korrekturfaktoren (ΔH) beider Methoden umfassten +1300 m (Finnland, TAV-Matching), +370 m (Niederlande/Nordostdeutschland, TAV-Matching) und −220 m (italienische Alpen, Lapse-Rate-Methode) über fünf Regionen hinweg. Insgesamt zeigten 27 regionenübergreifende TAV-abgeglichene Paare eine hohe Übereinstimmungsgenauigkeit (Median ΔTAV = 0,05 °C, 89 % ≤ 0,2 °C). Diese Faktoren standardisierten die Standorthöhen auf einen gemeinsamen südwestdeutschen thermischen Referenzrahmen und ermöglichten so die Vergleichbarkeit zwischen den Standorten.
Das Protokoll mit zwei Methoden erfordert keine Automatisierung und ist auf alle Taxa mit dokumentierten Standorthöhen anwendbar. Der vollständige methodische Arbeitsablauf – einschließlich Stationsdaten, Lapse-Rate-Regressionen, Abgleichentscheidungen und Korrekturberechnungen – ist öffentlich zugänglich bei Zenodo [DOI 10.5281/zenodo.18835116] und bietet Ökologen eine pragmatische, vollständig reproduzierbare Vorlage für die höhenstandardisierte Temperaturschätzung in Metaanalysen
Teleoperation in Rail and Tram Systems: A Scoping Review on Work Design, Roles, Tasks, and Human–Machine Interfaces
No full textAs rail transport becomes increasingly automated, ensuring effective human fallback solutions is critical to safety and performance. Teleoperation enables remote intervention, but research on human roles, work places, and performance is fragmented. This scoping review maps current concepts and technologies for teleoperation in rail and tram systems, emphasizing the central roles of human operators and human–machine interface design. This results in an in-depth synthesis of teleoperation studies and identification of key task clusters. Following PRISMA guidelines, a systematic literature search was conducted in five databases. Of 505 unique publications, 71 met the inclusion criteria after title and abstract screening. After full-text analysis, 59 studies were included and categorized into teleoperation (n = 21) and control room contexts (n = 38). Results reveal a wide variety of publications from various perspectives on teleoperation in the rail domain. Included studies ranged from prototype workplaces and interface concepts to task and activity analyses. However, validated performance metrics, comparative studies, and consistent theoretical grounding remain scarce. The findings highlight the need to better integrate human factors, ergonomics, and system design in future research on railway teleoperation