• Digital Humanities and Cultural Heritage
• Publications close
• Research data close
• Other research products close
• Bern Open Repository and Informatio... close
• OpenAIRE close

The interpretability of ML models is important, but it is not clear what it amounts to. So far, most philosophers have discussed the lack of interpretability of black-box models such as neural networks, and methods such as explainable AI that aim to make these models more transparent. The goal of this paper is to clarify the nature of interpretability by focussing on the other end of the 'interpretability spectrum'. The reasons why some models, linear models and decision trees, are highly interpretable will be examined, and also how more general models, MARS and GAM, retain some degree of interpretability. I find that while there is heterogeneity in how we gain interpretability, what interpretability is in particular cases can be explicated in a clear manner. Comment: 22 pages, 4 figures

During marine oxygen isotope stage (MIS) 2, the Swiss Plateau temporarily hosted large piedmont lobe glaciers that retreated after their maximum advance back to the fringe of the Alps. The presence of moraines in this region indicates that overall glacier recession was punctuated by repeated phases of ice-marginal stability and re-advances. The timing of these events in the region formerly covered by the eastern lobe of the Rhône (or Valais) glacier has been controversial but remains poorly constrained due to the lack of chronological data. To fill this gap, 10Be cosmic-ray exposure (CRE) dating was applied to erratic boulders inside the assumed MIS 2 maximum extent of this piedmont lobe. Erratic boulders close to the suspected MIS 2 maximum extent gave unrealistically young CRE ages. Erratic boulders at a presumably younger ice-marginal position (Brästenberg position) yielded an average age of ~19 ka, consistent with recalibrated basal radiocarbon ages from lakes and recomputed radiocarbon ages from large Late Pleistocene mammals buried in glacio-fluvial deposits. However, several erratic boulders beyond the Brästenberg position gave internally consistent, but stratigraphically too young ages of ~17 ka. We cannot rule out that glacier recession from the Brästenberg position began no later than ~17 ka. CRE dating of a moraine of a presumably younger ice-marginal position (Solothurn position) gave unrealistically old ages and an incredibly young age (86 ka, 41 ka, and 4 ka). 14C CRE dating should be applied to check whether the boulders associated with the Solothurn and Brästenberg positions were previously exposed to cosmicradiation. Nevertheless, despite outlying ages, the presented chronological data contribute to an overall consistent and increasingly refined chronology of the last deglaciation of the Swiss Plateau when compared with 59 previously published CRE ages.

Deep Learning hat bereits neue Erkenntnisse in den digitalen Geisteswissenschaften ermöglicht. Vormoderne Sprachen und Sprachen des globalen Südens bringen allerdings Herausforderungen mit sich, die aktuell diesen analytischen Zugriff in diesem Bereich noch nicht erlauben. Das Projekt "The Flow" entwickelt Lösungen für solche historische Korpora in den Bereichen Handschriftenerkennung, Entitätsidentifikation, Event-Extraktion, Topic Modeling und Clustering. Die Entwicklung der Webanwendung nopaque zielt darauf ab, diese bestehenden Methoden bzw. Werkzeuge in einem übergreifenden Workflow zu verbinden. Der Workshop stellt den aktuellen Stand von nopaque bzw. den Workflow vor. Ziel ist es, bei der Etablierung eines allgemein anwendbaren Workflows für Deep Learning die Vielfalt der Quellen und geisteswissenschaftlichen Forschung zu berücksichtigen. Wir laden Teilnehmer:innen ein, Ideen und Erfahrungen einzubringen und Implementierungen für nicht standardisierte Layouts, Schriften und Sprachen zu diskutieren. Der Workshop trägt dazu bei, unser Projektziel zu erreichen: maschinelles Lernen in allen Bereichen der Geschichtswissenschaft zugänglicher zu machen.

The carbon isotopic signature inherited from soil and epikarst processes may be modified by degassing and prior calcite precipitation (PCP) before its imprint on speleothem calcite. Despite laboratory demonstration of PCP effects on carbon isotopes and increasingly sophisticated models of the governing processes, to date, there has been limited effort to deconvolve the dual PCP and soil-epikarst components in measured speleothem isotopic time series. In this contribution, we explore the feasibility, advantages, and disadvantages of using trace element ratios and δ⁴⁴Ca to remove the overprinting effect of PCP on measured δ¹³C to infer the temporal variations in the initial δ¹³C of drip water prior to degassing and PCP. In nine examined stalagmites, the most widely utilized PCP indicators Mg/Ca and δ⁴⁴Ca covary as expected. However, Srg/gCa does not show consistent relationships with δ⁴⁴Ca so PCP is not the dominant control on Srg/gCa. From δ⁴⁴Ca and Mg/Ca, our calculation of PCP as f_Ca, the fraction of initial Ca remaining in solution at the time the stalagmite layer is deposited, yields multiple viable solutions depending on the assumed δ⁴⁴Ca fractionation factor and inferred variation in DMg. Uncertainty in the effective fractionation of δ¹³C during degassing and precipitation contributes to uncertainty in the absolute value of estimated initial δ¹³C. Nonetheless, the trends in initial δ¹³C are less sensitive to these uncertainties. In coeval stalagmites from the same cave spanning the 94 to 82 ka interval, trends in calculated initial δ¹³C are more similar than those in measured δ¹³C and reveal a common positive-Anomaly initial δ¹³C during a stadial cooling event. During deglaciations, calculated initial δ¹³C implies a trend of greater respiration rates and higher soil CO₂, although the higher interglacial drip water saturation favors more extensive PCP. Initial δ¹³C can be estimated for active and fossil speleothems from a range of settings, wherever there is confidence that Mg/Ca and/or δ⁴⁴Ca provides a quantitative indication of past changes in PCP. Further study of Mg partitioning in speleothems will improve the robustness of Mg/Ca as a PCP proxy. Climate of the Past, 19 (12) ISSN:1814-9332 ISSN:1814-9324

Because of the mass gap, lattice QCD simulations exhibit stochastic locality: distant regions of the lattice fluctuate independently. There is a long history of exploiting this to increase statistics by obtaining multiple spatially-separated samples from each gauge field; in the extreme case, we arrive at the master-field approach in which a single gauge field is used. Here we develop techniques for studying hadronic observables using position-space correlators, which are more localized, and compare with the standard time-momentum representation. We also adapt methods for estimating the variance of an observable from autocorrelated Monte Carlo samples to the case of correlated spatially-separated samples. Journal of high energy physics 11, 167 (2023). doi:10.1007/JHEP11(2023)167 Published by SISSA, [Trieste]

Anthropogenic activities have exerted strong influence on ecosystems worldwide, particularly since 1950 CE. The local impact of past human activities often started much earlier and deserves detailed study. Here, we present an environmental record from a 278 cm long sedimentary core from Lake Höglwörth (Bavaria, Germany). Sedimentological and geochemical parameters indicate that the organic-rich bottom sediments of the record consist of peat that formed prior to 870-160+140 CE, when lake sediments started to accumulate. After 870-160+140 CE, distinct shifts in lithology, elemental composition, and the biological record are visible and are interpreted to result from the construction of a monastery on the lake peninsula in 1125 CE and/or the damming of the lake. From 1120±120 to 1240-120+110 CE, the lake environment was relatively stable. This period was followed by enhanced deforestation that led to a more open landscape and soil erosion, visible in increased allochthonous input from 1240-120+110 to 1380-110+90 CE. This was accompanied by high aquatic productivity and bottom or interstitial water anoxia from 1310-120+100 to 1470-100+90 CE, possibly triggered by increased nutrient availability. Enhanced allochthonous input and a substantial shift in the aquatic community can be assigned to the construction of a flour mill and related rerouting of a small creek in 1701 CE. High aquatic productivity and bottom or interstitial water anoxia after 1960±10 CE correspond to recent eutrophication resulting from accelerated local anthropogenic activities. The sedimentary record from Lake Höglwörth exemplarily demonstrates that anthropogenic activities have had substantial environmental impacts on aquatic environments during the past millennium.