31325 research outputs found
Sort by
High-dimensional Change Point Detection using Graph-Spanning Ratio
No full textDie Wendepunktdetektion (Change Point Detection, CPD) identifiziert Verteilungs- änderungen in sequenziellen Daten in verschiedenen Bereichen, darunter Biologie, Ingenieurwesen und Finanzen. Moderne CPD steht aufgrund der hohen Dimensionalität vor großen Herausforderungen, da die Datendimension die Anzahl der Beobachtungen übersteigt. Inspiriert von Graphstrukturen schlagen wir einen neuartigen CPD-Algorithmus vor, der darauf ausgelegt ist, Änderungen des Mittelwerts und der Varianz sowohl in Offline- als auch in Online-Daten niedriger bis hoher Dimensionen zu erkennen. Zentral für unseren Ansatz ist ein Ähnlichkeitsmaß, das aus dem graph-spanning Ratio (GSR) abgeleitet wird und statistische Tests ermöglicht, um festzustellen, ob eine Änderung aufgetreten ist. Die Methode ist vielseitig, dass sie auf euklidische oder graphisch strukturierte Daten aus unbekannten Verteilungen mit vorgegebenen Fehlerwahrscheinlichkeiten angewendet werden kann. Der GSR-Rahmen wird theoretisch durch das Studium der Teststatistik, der Testqualität und der Validierung des Testverfahrens unterstützt. Theoretisch zeigen wir, dass eine hohe Erkennungsleistung erreicht werden kann, wenn das Ausmaß der Änderung die Untergrenze der Minimax-Trennungsrate für Gaußsche Daten überschreitet. Durch die Nutzung von Gaußschen Näherungen unter leichten Momentbeschränkungen erweitern wir diese theoretischen Ergebnisse auf nicht-Gaußsche Daten. Durch eine numerische Studie mit künstlichen Datensätzen zeigt unser datengesteuerter Ansatz eine hohe Erkennungsleistung für hochdimensionale Daten, während die Falschalarmrate (Fehler Typ I) auf einem nominalen signifikanten Niveau gehalten wird. Im Vergleich zu bestehenden Methoden erreicht unser Algorithmus eine überlegene Genauigkeit sowohl für Gaußsche als auch für nicht-Gaußsche Daten. Insbesondere funktioniert der Graph-Spanning-Ansatz außergewöhnlich gut bei kleinen Scanfenstern, was ihn besonders effektiv für Online-Umgebungen macht, in denen eine zeitnahe und genaue Erkennung entscheidend ist. Schließlich wenden wir die Methode auf Finanzdaten an und erkennen erfolgreich Wendepunke in S&P 500-Aktien, was ihre Anwendbarkeit in der realen Welt demonstriert.Change point detection (CPD) identifies distributional changes in sequential data in various domains, including biology, engineering, and finance. Modern CPD faces significant challenges due to high dimensionality, as the data dimension exceeds the number of observations. Inspired by graph structures, we propose a novel CPD algorithm designed to detect changes in mean and variance in both offline and online data of low to high dimensions. Central to our approach is a similarity measure derived from the graph-spanning ratio (GSR), which enables statistical tests to determine whether a change has occurred. The method is versatile that it can be applied to Euclidean or graph-structured data from unknown distributions with prescribed error probabilities. The GSR framework is theoretically supported by the study of test statistics, test quality, and validation of the test procedure. Theoretically, we show that high detection power can be achieved when the magnitude of the change exceeds the lower bound of the minimax separation rate for Gaussian data. By leveraging Gaussian approximations under slight moment constraints, we extend these theoretical results to non-Gaussian data. Through a numerical study on artificial datasets, our data-driven approach shows high detection performance for high-dimensional data while keeping the false alarm rate (Type I error) at a nominally significant level. Compared to existing methods, our algorithm achieves superior accuracy for both Gaussian and non-Gaussian data. In particular, the graph-spanning approach performs exceptionally well for small scan windows, making it particularly effective for online environments where timely and accurate detection is crucial. Finally, we apply the method to financial data and successfully detect change points in S&P 500 stocks, demonstrating its applicability in the real world
Scientific literature on carbon dioxide removal revealed as much larger through AI-enhanced systematic mapping
No full textCarbon dioxide removal plays an important role in any strategy to limit global warming to well below 2 °C. Keeping abreast with the scientific evidence using rigorous evidence synthesis methods is an important prerequisite for sustainably scaling these methods. Here, we use artificial intelligence to provide a comprehensive systematic map of carbon dioxide removal research. We find a total of 28,976 studies on carbon dioxide removal—3–4 times more than previously suggested. Growth in research is faster than for the field of climate change research as a whole, but very concentrated in specific areas—such as biochar, certain research methods like lab and field experiments, and particular regions like China. Patterns of carbon dioxide removal research contrast with trends in patenting and deployment, highlighting the differing development stages of these technologies. As carbon dioxide removal gains importance for the Paris climate goals, our systematic map can support rigorous evidence synthesis for the IPCC and other assessments.ERC-2020-SyG "GENIE" (grant ID 951542), German Ministry of Research and Education under the CDR-SynTra project (01LS2101F)Peer Reviewe
Introduction: Imagery of blackness and colonial fantasies in Central and Eastern European modernism
No full textThe introduction addresses the broader context of the research topic, presents the state of the art and outlines the objectives and sources of the special issue. Its articles examine literary, artistic and scientific representations of race and the Other, analysing the way Central and Eastern European societies have adopted, adapted or subverted colonial ideas. As a whole, the volume offers interdisciplinary insights into how imagery of blackness and colonial fantasies shaped identities in Central and Eastern Europe, revealing the complexities of modernity, colonialism and racial constructs.HORIZON EUROPE Marie Sklodowska-Curie Actions https://doi.org/10.13039/100018694Peer Reviewe
Monostudiengang - Überfachlicher Wahlpflichtbereich für andere Bachelorstudiengänge und -studienfächer
No full textKorrigendum zu: Die Judäer/Aramäer von Elephantine und ihre Religion
No full textDieser Beitrag ist mit Zustimmung des Rechteinhabers aufgrund einer (DFG-geförderten) Allianz- bzw. Nationallizenz frei zugänglich.Korrigendum zu: Die Judäer/Aramäer von Elephantine und ihre ReligionPeer Reviewe
Detecting newly installed bat boxes: Bats’ prior familiarity with artificial roosts may play a bigger role than improved echo-reflective properties
No full textHabitat loss in Europe severely affects bats, particularly tree-roosting species, due to the decreasing availability of tree cavities. One common conservation strategy is the installation of artificial roost boxes. However, the occupation of newly installed roost boxes can take up to several years, and the underlying mechanisms for successful roost detection in bats are still poorly understood. This study proposes enhancing the detectability of roost boxes to echolocating bats by incorporating hollow hemispheres that provide highly conspicuous echoes. The hemispheres strongly reflect the echolocation calls of passing bats and are thus well detectable over a broad range of angles. We hypothesized that roost boxes equipped with these hemispheres would attract more bats and exhibit greater bat activity than standard, unmodified boxes. To evaluate this, we placed 30 modified boxes and 30 unmodified boxes across three forest areas in Northern Germany, each differing in proximity to known bat hibernation sites and the prior presence of artificial roosts. We monitored bat activity by measuring light beam interruptions at each box and found that the activity of bats at the boxes varied considerably. Our findings indicate that, contrary to our hypothesis, bat activity was more strongly influenced by their prior experience with artificial roosts than by the increased detectability provided by hollow hemispheres. Furthermore, our study revealed that light beam interruptions indicated bat presence at the boxes earlier than visual checks for bats or feces, showcasing the benefits of non-invasive monitoring techniques. Conservation efforts are complex, and these results imply that for effective bat conservation, increasing bats’ familiarity with artificial roosts may be more important than merely enhancing the detectability of these structures.Peer Reviewe
Answer to the Letter to the Editor of M. Hamza, et al. concerning “Association between lumbar paraspinal muscle activities and quality in chronic low back pain: a cross-sectional analysis” by Zhang T, Firouzabadi A, Becker L, et al. (Eur Spine J [2025]: doi: 10.1007/s00586-025-08727-x)
No full textOpen Access funding enabled and organized by Projekt DEAL.Charité - Universitätsmedizin Berlin (3093)Peer Reviewe
Evaluation of gridded precipitation datasets over Iran
No full textThe article processing charge was funded by the Open Access Publication Fund of Humboldt-Universität zu Berlin.Study region:
Six climate zones in Iran: Zone 1 (humid Caspian region), Zone 2 (cold and semi-arid northwestern Iran), Zone 3 (mild and semi-humid western Iran), Zone 4 (mild and semi-arid northeastern Iran), Zone 5 (warm and semi-arid southwestern Iran), and Zone 6 (warm and arid eastern, southern, and southeastern Iran).
Study focus:
This study evaluates the performance of seven gridded precipitation datasets (CRU TS, TMPA 3B43, GPCC, IMERG, MERRA-2, ERA5, and PDIR) against rain gauge data. The accuracy of these datasets in representing precipitation patterns is assessed from 2000 to 2017, with a focus on their performance across diverse climate zones in Iran.
New insights for the region:
In Zone 1, ERA5 performs best in winter and autumn, while GPCC excels in spring and summer. In Zones 2 and 3, GPCC is the top performer in most seasons, while CRU TS performs best in autumn and winter for Zone 2. In Zone 4, GPCC outperforms all datasets year-round. In Zone 5, TMPA 3B43 performs best in winter, with GPCC slightly outperforming others in other seasons. Overall, ERA5 is most accurate in Zone 1, while GPCC performs best in other zones, with PDIR consistently underperforming across all zones. GPCC captures interannual precipitation variability well, making it the best overall performer. We recommend ERA5 for real-time monitoring in Zone 1, while GPCC is better suited for long-term studies in areas with dense rain gauge networks.Peer Reviewe
Follicle architecture and innervation of functionally distinct rat vibrissae
No full textThe vibrissa follicle is a complex mechanotransducer with intricate accessory structures such as vibrissa, ring sinus and ringwulst as well as rich innervation by diverse afferent types. Establishing how afferent types and accessory structures operate together to derive specific kinds of sensory information has been challenging, because we often lack precise information on afferent types, accessory structures and vibrissa function. Here we address this challenge by synchrotron X-ray imaging of vibrissa follicles of rat vibrissae with distinct function. Specifically, we characterize accessory structures and trace myelinated axons of the all-purpose-sensing C2-, an object-sensing micro-, the wind-sensing supraorbital- and the ground-sensing trident-vibrissa. We find that while vibrissa length and follicle size differ widely across these vibrissae, the ringwulst and the associated club-like afferents are of near constant diameter and height and appear to form a non-scalable sensory module. The two longer vibrissae (supraorbital and C2 vibrissa) have noticeably more club like afferents, suggesting a special role of the ringwulst in transducing presumably smaller deflection angles encountered by long sensory hairs. The trident vibrissa receives overall few afferents, which are strongly polarized to the posterior vibrissa-shaft, a putative specialization to sensing forward-egomotion. We conclude that high-resolution structural analysis allows relating follicle architecture and function.Deutsche Forschungsgemeinschaft (German Research Foundation)https://doi.org/10.13039/501100001659Deutsches Elektronen-Synchrotron (DESY)https://doi.org/10.13039/501100001647Max-Planck-Gesellschaft (Max Planck Society)https://doi.org/10.13039/501100004189Sonderforschungsbereich SFB1456/A3Peer Reviewe
Were there STAMP morphs in earlier stages of Niger-Congo?
No full textSTAMP morphs–portmanteau subject-TAM-polarity morphemes–are found in numerous attested subgroups of Niger-Congo and have been proposed for at least one intermediate proto-language (Proto-Cross-River within Benue-Congo). This paper suggests that other older states in the history of Niger-Congo might also have made use of STAMP morphs, as the environment for creating such on the left edge of the verbal complex has endured for centuries in different Niger-Congo groups. Thus, different languages and subgroups have likely availed themselves of this structural option at different points in the history of Niger-Congo and this should be considered when attempting reconstruction in this domain