bonndoc - Der Publikationsserver der Universität Bonn

Bayesian Approaches for Robust and Efficient Sequential Decision-Making

2026-09-04T11:45:53Z

Bayesian Approaches for Robust and Efficient Sequential Decision-Making Weichert, Dorina In many situations, the solution to a problem is not found in one fell swoop, but iteratively, by making a series of decisions to ultimately achieve the desired goal. The formal framework for this process is referred to as sequential decision-making. Here, the next step is determined by taking into account the available information and the possible consequences of the decision.
Methods from the field of sequential decision-making are used in chemistry, robotics or engineering, but their application is still limited due to open challenges. This work deals with two of them: sample efficiency and robustness. Sample efficiency is crucial when the data required for decision-making is costly, which is the case for many real-world problems. Moreover, in these cases, the solution to the sequential decision-making problem is often found in a setting that differs from the final application, so additional robustness is required. We focus on finding adversarially robust solutions, i.e., solutions that are optimal even under the worst application conditions - a typical but conservative robustness assumption.
In this work, we develop sample-efficient and adversarially robust algorithms for three subdomains of sequential decision-making: Bayesian Optimization, Active Learning, and Reinforcement Learning. In all algorithms, we achieve high sample efficiency through Bayesian methods that efficiently combine complex expert knowledge with data and, by exploiting their inherent uncertainty estimates, enable information-based methods that are known for their high sample efficiency. Furthermore, in order to find solutions that are adversarially robust at the time of application, we already include both the possible application scenarios and the robustness requirements during the sampling process.
The results obtained extend the state of the art considerably. In Bayesian Optimization, we search for the optimum of a costly-to-evaluate black-box function that is robust to a worst-case realization of parameters that are uncontrollable at the time of application. Current algorithms focus either on sample efficiency or on robustness to adversarial attacks, but not on these two problems simultaneously. In Active Learning we deal with a specific problem that has arisen in practice: We learn a probabilistic binary classifier and focus on finding the worst possible sublevel set of the class probability function under the possible application conditions. Competing approaches are either sample-efficient Bayesian approaches to level set estimation, which thus lack the robustness component; or they focus on general Active Learning problems with corruptions during the learning process, and thus do not fit the specific setup. In Reinforcement Learning, we find an optimal sequence of actions to reach a certain final state that also works under the worst-case realization of the environment. Therefore, we assume the environment to be represented by a discrete set of oracles that all match the ground truth environment for different application cases. In the case of multiple oracles, the current state-of-the-art methods try to identify the correct model, while optimizing a risk measure over the oracles. Our approach differs as follows: we restrict ourselves to finding a worst-case robust solution, so we do not devote resources to the identification of the correct oracle. Additionally, we further increase the sample efficiency by exploiting knowledge about the variation of the samples gained from the oracles.
For all mentioned subdomains of sequential decision-making we derive adversarially robust and sample-efficient algorithms. Their superior performance over the state of the art is demonstrated empirically on two kinds of problems. First, on variants of traditional benchmark problems from the individual subdomains that consider the robustness requirement. Here, we show that our approaches are both robust (as they find the robust optima), and sample-efficient as they find them faster than competing robust approaches. Secondly, we demonstrate the usefulness of our approaches on engineering problems from the real-life applications that originally motivated our work. In this way, we can show that our approaches are not just theoretical, but really relevant in practice.

Associations of social determinants of health and patient safety in perinatal care

2026-09-04T10:31:41Z

Associations of social determinants of health and patient safety in perinatal care Averdunk, Katharina; Miani, Céline; Strizek, Brigitte; Weigl, Matthias Background: Associations between adverse social conditions and poor health are well documented – also in perinatal care. However, research into the actual ramifications of such disparities for perinatal patient safety remains inconclusive. Therefore, to achieve a comprehensive understanding of the risk and burden of patient harm to disadvantaged people, we aim to systematically review current evidence on social determinants of health (SDoH) and perinatal patient safety.
Objective: This study protocol outlines definitions, methods, and procedures for a systematic literature review with meta-analysis aiming to synthesise the research base on the associations between SDoH and patient safety in perinatal care.
Methods: Adhering to PRISMA guidelines, a literature search will be conducted for a systematic review in MEDLINE (PubMed), Scopus database, CINAHL (EBSCO), and Embase (Elsevier) for quantitative studies reporting associations between SDoH and patient safety measures in perinatal care. Data extraction will include study design, population, SDoH variables, outcome measures, effect sizes, and control variables. If deemed feasible after assessment of heterogeneity, narrative synthesis of findings will be complemented by conducting meta-analyses of pooled effect sizes. Methodological quality of included studies will be assessed using JBI Critical Appraisal Tools, and the certainty of evidence using the GRADE tool. This protocol is registered on PROSPERO (CRD420251090149) and OSF (https://doi.org/10.17605/OSF.IO/UP3JS).
Discussion: The review aligns well with current global efforts to promote safe perinatal care and presents an innovative, comprehensive approach for assessing the associations between SDoH and patient safety. The review will provide the first systematic synthesis of current evidence of SDoH and patient safety in perinatal care. Anticipated limitations include heterogeneity of study designs, measures, and outcomes, with expected predominance of observational studies, which may limit causal inferences. However, this review will provide a valuable foundation for further empirical research and interventions to enhance equitable and safe perinatal care.

Verbesserung des analytischen Verfahrens zur Diagnostik hereditärer thrombozytärer Erkrankungen

2026-09-04T08:20:23Z

Verbesserung des analytischen Verfahrens zur Diagnostik hereditärer thrombozytärer Erkrankungen Loos, Kevin Hereditäre Thrombozytopathien zählen zu den seltenen Erkrankungen, deren Diagnostik aufgrund ihrer unterschiedlichen phänotypischen Ausprägungen eine Kombination verschiedener labordiagnostischer Verfahren erfordert. Die Immunfluoreszenzmikroskopie (IFM) von Thrombozyten stellt hierbei laut aktueller AWMF-Leitlinie eine anerkannte Referenzmethode dar, welche jedoch nur an wenigen nationalen Zentren angeboten wird. Ziel dieser Arbeit war die Entwicklung und Etablierung eines standardisierten Immunfluoreszenzfärbeprotokolls in Kombination mit einer automatisierten Mikroskopie und computergestützten Bildanalyse. Dadurch sollen einige Vorteile der Durchflusszytometrie, insbesondere die quantitative Diagnostik, mit denen der IFM, insbesondere der qualitativen bildmorphologischen Analyse in einem labordiagnostischen Verfahren kombiniert werden. Hierdurch soll eine reliabel, objektiv und valide beurteilbare Diagnostik ausgewählter hereditärer Thrombozytopathien erfolgen. Im Rahmen dieser Dissertation wurde so ein neuer labordiagnostischer Assay entwickelt, welcher die Diagnostik hereditärer Thrombozytopathien um ein standardisiertes und automatisiertes Verfahren ergänzt und damit das bisherige Analysespektrum am Institut für Experimentelle Hämatologie und Transfusionsmedizin des Universitätsklinikums Bonn erweitert.; Inherited platelet disorders are rare diseases whose diagnosis requires a combination of different laboratory diagnostic methods due to their diverse phenotypic manifestations. According to the current AWMF guideline, platelet immunofluorescence microscopy (IFM) is an established reference method; however, it is currently offered at only a few specialized centers in Germany. The aim of this study was to develop and establish a standardized immunofluorescence staining protocol in combination with automated microscopy and computer-assisted image analysis. This approach is intended to combine some of the advantages of flow cytometry, particularly quantitative diagnostics, with those of IFM, especially qualitative morphological image analysis, within a single laboratory diagnostic procedure. This should enable reliable, objective, and valid diagnostic assessment of selected inherited platelet disorders. Within the framework of this doctoral thesis, a novel laboratory diagnostic assay was developed that complements the diagnosis of inherited platelet disorders with a standardized and automated method, thereby expanding the existing diagnostic capabilities at the Institute of Experimental Haematology and Transfusion Medicine at the University Hospital Bonn.

The role of corticothalamic circuits in avoidance learning

2026-09-04T07:05:25Z

The role of corticothalamic circuits in avoidance learning Haetzel, Laura Maria Avoidance learning entails linking a threat-predictive stimulus (CS) to aversive outcomes (US) and developing a behavioural strategy (shuttling) to forestall harm. This multifactorial process necessitates neural circuits capable of prioritising relevant sensory cues and promoting optimal actions. Auditory cortex (AC) and thalamic reticular nucleus (TRN) send excitatory and inhibitory afferents to auditory thalamus (medial geniculate body; MGB), respectively, and MGB neurons play an active role in associative fear learning. However, it is unclear how AC-MGB and TRN-MGB projections channel information to guide stimulus-outcome associations and behavioural strategy during avoidance learning. To better understand the role of AC-MGB and TRN-MGB projections, I recorded their activity during a 5-day avoidance learning paradigm using genetically restricted calcium indicators and in vivo fibre photometry. I found that AC-MGB projections primarily provide stable sensory representations by encoding aversive cues and auditory cue familiarity. In contrast, inhibitory TRN-MGB projections represented escape decisions and tracked learning-related variables that develop across trials, such as CS value and prediction error. Further downstream, MGB sends projections to medium spiny neurons (MSNs) in tail of striatum (TS), a region that receives dense auditory input and has been implicated in threat prediction. Using optogenetics-assisted circuit mapping in ex vivo slices, I found that thalamostriatal projections from MGB modulated action potential firing in TS MSNs. This regulation of striatal output proceeded in part through a newly identified long-range inhibitory projection from MGB that triggered slow, GABA_B-receptor mediated currents in TS MSNs. Though the in vivo relevance of MGB-TS projections remains to be studied, thalamic tuning of striatal output is likely implicated in avoidance learning. Altogether, cortico-thalamo-striatal circuits surrounding MGB exhibit distinct signalling modes and dynamic encoding properties that guide avoidance learning.