TU Dortmund University
Eldorado - Ressourcen aus und für Lehre, Studium und ForschungNot a member yet
19914 research outputs found
Sort by
Mathematische Grundlagen im maschinellen Lernen
No full textThemen des Maschinellen Lernens (ML) werden in der Lehre von Mathematik bereits als Anwendungsbeispiele zur Vertiefung unterschiedlicher mathematischer Inhalte genutzt. Doch welche mathematischen Inhalte können den Umgang mit Verfahren des ML unterstützen? Im Beitrag wird exemplarisch an Verfahren des überwachten ML gezeigt, welche mathematischen Inhalte dem Erstellen eines ML-Modells zugrunde liegen. Methodisch gestützt durch das so genannte Modellkonzept werden mathematische Voraussetzungen sowie inhaltliche Anknüpfungspunkte zum Erlernen von ML analysiert und strukturiert
Bounded variation in multi-stage optimization
No full textThis thesis studies the computational complexity of a multi-stage optimization problem with either penalized or bounded variation.
The input of the problem is a sequence of instances (one for each time stage), and the task is to find a sequence of solutions (one for each time stage) that achieves a tradeoff between the quality of the solutions in each time stage and their similarity.
The amount of change in the sequence of solutions is quantified by the so-called variation which will be measured time-wise, component-wise, or in total.
The variation of the solution sequence is either incorporated into the model by a penalty term or by imposing a hard upper bound in the set of constraints.
This gives rise to two related but not equivalent multi-stage problems.
When only one binary decision can be made in each stage, the tractability of the problems with either of the three types of variation will be settled by presenting respective compact extended LP-reformulations.
A thorough polyhedral study of the problem versions will yield complete, and usually exponentially large, descriptions of the convex hull of feasible solutions in the original variable space and also corresponding efficient separation algorithms.
For a higher-dimensional underlying combinatorial problem, the complexity of the problem versions will turn out to depend on the type of variation and also on whether this is penalized or bounded.
Indeed, even for an underlying problem as trivial as a selection problem, one version will be strongly NP-hard, while others will be tractable.
An oracle-based approach will shift the perspective from a complexity-theoretical one to a rather information-related one in order to investigate the relative complexity of the bounded and penalized problem version with respect to the underlying combinatorial problem
Datenbasierte Entscheidungsbäume mit unplugged Datenkarten als Einstieg in das maschinelle Lernen
No full textDie Integration von maschinellem Lernen in den Schulunterricht gewinnt zunehmend an Bedeutung. Im ProDaBi-Projekt wurde eine Unterrichtseinheit für die Sekundarstufe I entwickelt, in der Lernende mithilfe von analogen Datenkarten datenbasierte Entscheidungsbäume erstellen. Drei Studien mit Schülerinnen und Schülern zeigten, dass Grundlagen von datenbasierten Entscheidungsbäume erfolgreich erlernt und angewendet werden können. Herausforderungen lagen in der optimalen Nutzung von Strategien wie Sortieren oder Gruppieren der Datenkarten und der datenbasierten Argumentation
Bridging scales in digital pathology
No full textHistopathology is a cornerstone of disease diagnosis and treatment, traditionally relying on manually assessing tissue specimens under a microscope. However, the advent of slide scanners to produce digital tissue representations, so-called whole-slide images (WSI), has enabled computational pathology to perform quantitative and automated tissue analysis. Current developments in Artificial Intelligence, particularly Deep Learning, have accelerated the progress in this field. This thesis proposes a comprehensive Deep Learning pipeline for quantitative histopathological image analysis, integrating WSI preprocessing, algorithm development for tissue and cell-level segmentation, and clinical application in an end-to-end workflow. The approach not only improves the quantitative evaluation of WSI but also extracts diagnostic and prognostic markers while automatically characterizing tissue dynamics through morphological tissue features. Segmenting entire tissue sections into classes like tumorous or non-tumorous requires the consideration of global tissue patterns as well as local cell morphologies. Following this, we introduce the Memory Attention Framework that can be incorporated into any encoder-decoder segmentation architecture. This framework enables the adaptive incorporation of tissue context during fine-grained local segmentation. The method was evaluated on two public datasets (breast, liver) and an internal kidney cancer dataset, demonstrating superiority over non-context and multiscale segmentation approaches. Notably, the approach reduced the number of false-positive tumor regions. Building on this, we applied the framework to a pancreatic cancer cohort consisting of 400 internal and 182 external patients to quantify the tumor microenvironment and correlate it with patient outcomes. In doing so, we were able to stratify patients into two risk groups based on tissue composition and spatial tumor-stroma distribution, which showed significant (p < 0.05) differences in their survival probabilities. Next to tissue analysis, segmentation on the cellular level is crucial to uncover the cellular composition of tissue samples. While convolutional neural networks have been extensively used for this task, we evaluate the capabilities of Transformer-based networks and incorporate so-called foundation models to improve accuracy compared to existing solutions. The proposed CellViT and CellViT++ models have proven to achieve State-of-the-Art results on several benchmark datasets, covering a broad spectrum of tissue types and cell classes, bringing cell segmentation solutions closer to clinical practice. The models require minimal data for fine-tuning and exhibit remarkable zero-shot cell segmentation quality. This capability allows for a considerably faster adaptation to new research hypotheses without the need for extensive development time. In summary, this work presents Deep Learning techniques for quantifying tissue at both the macro and micro levels, enhancing diagnostic workflows, and identifying prognostic markers
Zusammenhänge von Unterrichtsqualität in der fachschulischen Ausbildung und der Entwicklung des mathematikdidaktischen Wissens angehender Erzieher*innen
No full textUm angehende Erzieher*innen darauf vorzubereiten das mathematische Lernen von Kindern der Kindertagesstätte passgenau begleiten zu können, sollte bereits in der fachschulischen Ausbildung spezifische MPCK aufgebaut werden. Offen ist hier, inwiefern der Wissenszuwachs angehender Erzieher*innen mit einer soliden theoretischen Fundierung des Unterrichts zusammenhängt. Im Beitrag werden die Ergebnisse einer quasi-experimentellen Studie vorgestellt und diskutiert, die die Entwicklung von MPCK bei N = 215 angehenden Erzieher*innen abhängig von der theoretischen Fundierung des Unterrichts untersucht
An Experimental Study: Improved Electrohydraulic Forming Efficiency through Wire-Free Discharge
No full textThis experimental research proposes improved electrohydraulic forming (EHF) methods to
overcome the limitations of traditional high-speed forming and deep-drawing processes.
EHF uses a working fluid to generate forming force, which lasts longer than in
electromagnetic or explosive forming, allowing precise shape control. However,
conventional EHF requires fastening wires and draining/refilling the working fluid for each
experiment, resulting in long preparation times and inconsistent forming due to variation in
wire position. To address these issues, two new methods ‘wire-cross EHF’ and ‘wire-free
EHF’ were introduced and tested using a free-bulging die and 0.5 t STS430 specimens.
Forming distribution was analysed by comparing forming force, forming height, and
preparation time. The wire-free EHF showed lower initial forming force but achieved
comparable forming height through consecutive experiments. It also demonstrated high
consistency and reduced total experiment time by 4.72 times compared to the conventional
EHF, which suffered from increased wire length during twisting, leading to variations in
forming height and forming distribution. Overall, the wire-free EHF proved most effective
in achieving consistent forming results with significantly reduced preparation time, making
it a promising approach for efficient EHF process
Verstehenstypen von Lehrkräften zum Konzept Skalarprodukt und ihre Vorstellungen von Verstehensprozessen
No full textZur Realisierung verständigen Unterrichts in der analytischen Geometrie benötigen Lehrkräfte gegenstandsbezogene Expertise, welche nicht nur durch fach- und fachdidaktisches Wissen, sondern auch durch „Vorstellungen darüber, wie Verstehensprozesse ablaufen“ (Drollinger-Vetter, 2011) bestimmt wird. Dieser Beitrag untersucht, welches eigene Verständnis Lehrkräfte bei der Bearbeitung einer Lernendenaufgabe zum Konzept Skalarprodukt aktivieren und typisiert dieses. Daraus lassen sich unterschiedliche Vorstellungen zum Unterrichten des Skalarproduktes ableiten und Fortbildungsbedarfe spezifizieren
Das ‚Polypersonale‘ Alter Ego – Ein komplexitätsreduzierender Mechanismus im Fremdverstehen von Mathematiklehrkräften
No full textIn diesem Beitrag soll eine qualitativ-empirische Untersuchung des Fremdverstehens von Lehrkräften im Mathematikunterricht vorgestellt werden, in welchem diese auf das Bewusstseinserleben ihrer Schülerinnen und Schüler gerichtet sind. Im Rahmen dieser Untersuchung konnten mehrere Merkmale rekonstruiert werden, die ein solches Fremdverstehen aufweisen kann. Eines dieser Merkmale – das ‚polypersonale‘ alter ego – soll beschrieben und diskutiert werden
Development of optochemical strategies to study proteome-wide ubiquitination dynamics in a linkage-specific manner and target protein-specific ubiquitination events
No full textProteins are the primary drivers of cellular processes and normal cellular functioning depends on their precise balance of homeostasis and degradation. Protein degradation is largely brought about by ubiquitination, a post-translational modification that adds multiple molecules of the protein Ubiquitin (Ub) to substrates and targets them for proteasome-mediated degradation. Ubiquitination is a highly conserved, complex phenomenon involving multiple enzymatic reactions that covalently adds Ub to substrates via an isopeptide linkage between the ɛ-amino group of a substrate lysine (Lys, K) and the carboxyl group of the terminal glycine (Gly, G) of Ub. Ub itself has seven Lys on which more Ub molecules can be added giving rise to polyUb chains. Initially discovered as an important mediator of protein degradation and thus turn over, subsequent studies have revealed that ubiquitination also plays important roles in cellular signal transduction. Linkage-specificity of Ub chains is derived from the specific Lys on which chains are built and their specific structure, giving rise to a complex code of Ub chains Consequently, this linkage-specificity decides the biological outcome of the ubiquitination event. Due to the complexity and inter-twined dynamics of ubiquitination, the study of its linkage specific kinetics becomes largely significant. However, due to its highly transient behaviour, studying protein ubiquitination in vivo has proved to be challenging and is often met with background noise from endogenous Ub and the various possibilities of Ub chain structure formation. In this study, a light-activated Ub has been developed by incorporating a photocaged lysine at specific sites, through amber codon suppression, for the monitoring of proteome-wide linkage-specific polyubiquitination and to gain insights into the kinetics of the same. This study reveals rapid, minute-scale ubiquitination kinetics for Lys11 (K11), Lys48 (K48) and Lys63 (K63) linkages. Also, the roles of individual components of the ubiquitin-proteasome system have been studied in K48-initiated chain synthesis by small molecule inhibition. This approach expands the repertoire of current cellular ubiquitination perturbation strategies with the ability to control linkage-specific ubiquitination at high temporal resolution and is a promising tool for studying ubiquitinome dynamics. Furthermore, in this study, advances were made towards looking into the kinetics of a monoubiquitination event in Ten-Eleven translocation (TET) proteins
Profilanalysen von (angehenden) Mathematiklehrkräften zu inklusionsbezogener professioneller Unterrichtswahrnehmung und zu professionellem Wissen
No full textIm Projekt TEDS-IME wurde die Messung von Lehrkräftekompetenzen um die inklusionsbezogene Perspektive erweitert. Basierend auf 456 Lehramtsstudierenden im Master, Referendar*innen und Lehrkräften wurden explorativ mittels latenter Profilanalysen Kompetenzprofile bezüglich eines inklusiven Algebraunterrichts identifiziert und anhand (berufs)biografischer Angaben näher beleuchtet. Es zeigen sich drei Profile (niedrig, mittel, hoch) hinsichtlich der professionellen Unterrichtswahrnehmung und des Wissens mit einem signifikanten Einfluss der Abiturnote, des Alters und des gymnasialen Lehramts