256869 research outputs found
Sort by
Annemarie Hubacher-Constam: Hauptentwerferin im Schatten ihrer Partner und ihrer Zeit
No full textThis chapter examines the life and work of Annemarie Hubacher-Constam, a Swiss architect and founding partner of “Hans Hubacher Architekt”. Despite her central role as the main designer of numerous notable projects, her contributions have been overshadowed by her male colleagues, reflecting broader gender dynamics in post-war architectural practice. Through archival research and oral histories, the chapter interrogates how female architects’ biographies can be narrated ethically: Should the focus remain on individual contributions or should their work be framed within a network of collaborative efforts? Using the 1959 Siedlung Rietholz near Zurich as a case study – an innovative prefabricated housing project suggestive of Annemarie’s involvement – the discussion highlights the challenges of attributing credit in collective design processes. The chapter ultimately questions how to represent women architects’ integrative roles without perpetuating biases, offering a nuanced perspective on constructing architectural histories
Simulation of catalytic and corrosive processes via Thermopotentiostat Molecular Dynamics
No full textEffectiveness of a Novel Vent-And-Slide Window for Short-Term Single-Sided Ventilation: A Study Based on Measurements and Detailed CFD Simulations
No full textNatural ventilation is regarded as one of the methods for maintain-ing fresh air indoors, in addition to having a significant impact on the overall performance of a building. However, pertinent data about details how processes and impacts of natural ventilation for different window technologies and oper-ation schemes is something still missing in a comprehensive fashion. Toward this end this contribution presents results of a study focusing on a novel vent-and-slide window. Thereby, the ventilation effectiveness of the ventilation posi-tion/state of this window is examined both empirically via measurements (results of tracer gas measurements) and simulation (corresponding computational-fluid-dynamics/CFD simulations) for single-sided ventilation. Monitoring efforts in a test building in Lower Austria formed the foundation for evaluating CFD mod-els in a well-used CFD-simulation-instrument, OpenFOAM. The study assesses the accuracy of simulated air exchange using data collected from short-duration single-sided ventilation in a test room during the summer and winter seasons in 2024/2025. Building on this knowledge, a virtual test room was generated and utilized to analyse air movement, air exchange, and the resulting thermal comfort in detail. Additionally, the study thoroughly examines the placement of air quality sensors, considering automated window ventilation
Between Subversion and Adaptation: Women in Rowing as Spatial Practitioners at the Turn of the Last Century
No full textThis chapter broadens the understanding of architecture as a field of spatial practice to examine how women engaged with and shaped everyday spaces and urban life around 1900, even within male-dominated domains. Focusing on rowing – a typical “gentleman’s sport” that emerged during industrialisation and excluded women and the working class – the chapter explores how women created alternative spaces for participation. Using the Friedrichshagener Damen-Ruder-Club, Germany’s first women’s rowing club in Berlin, as a case study, it examines how early female rowers navigated gendered exclusions to claim physical and architectural space. Through practices such as organising club activities and eventually building a boathouse, these women redefined access to leisure and urban waterscapes. Drawing on club chronicles, photo albums, newspaper clippings, water maps, and rowing guides, the chapter situates rowing as a spatial practice and explores how gender and class identities were historically mutually constituted through these interventions
LLMs and fuzzing in tandem: a new approach to automatically generating weakest preconditions
No full textThe weakest precondition (WP) of a program describes the largest set of initial states from which all terminating executions of the program satisfy a given postcondition. The generation of WPs is an important task with practical applications in areas ranging from verification to run-time error checking. This paper proposes the combination of Large Language Models (LLMs) and fuzz testing for generating WPs. In pursuit of this goal, we introduce Fuzzing Guidance (FG); FG acts as a means of directing LLMs towards correct WPs using program execution feedback. FG utilises fuzz testing for approximately checking the validity and weakness of candidate WPs, this information is then fed back to the LLM as a means of context refinement. We demonstrate the effectiveness of our approach on a comprehensive benchmark set of deterministic array programs in Java. Our experiments indicate that LLMs are capable of producing viable candidate WPs, and that this ability can be practically enhanced through FG
Mixed Reality Feuerlöschsimulator
No full textFeuerlöschen ist eine wichtige Fähigkeit im Alltag. Ein plötzlich ausbrechender Brand kann enorme Schäden an der Umwelt verursachen und sogar zu Todesfällen führen. Die meisten Menschen haben jedoch noch nie in ihrem Leben einen Feuerlöscher benutzt, was in kritischen Situationen zu falschen Entscheidungen führen kann. Herkömmliche Feuerlöschertrainings zielen darauf ab, dieses Problem zu beheben, erfordern jedoch einen erheblichen organisatorischen Aufwand und finanzielle Kosten. Eine vielversprechende Alternative, die wiederholbare, sichere und kostengünstige Trainingsszenarien ermöglicht, bietet die virtuelle Realität (VR).Bestehende VR-Simulatoren simulieren bereits gut die Struktur herkömmlicher Trainings, aber den meisten fehlt es an Realismus, wenn es um physisches Feedback geht. Aufbauend auf diesen Ansätzen wurde in dieser Diplomarbeit ein Mixed Reality (MR) Feuerlöschersimulator entwickelt, der einen realen, getrackten Feuerlöscher in die virtuelle Umgebung integriert und zusätzlich durch haptisches Feedback erweitert. Zwei zentrale Hardware-Mechanismen wurden dabei umgesetzt: ein Gewichtsbalance-System, um den Masseverlust während des Löschvorgangs zu simulieren, sowie eine Rückstoßsimulation, um die auftretenden Kräfte beim Sprühen nachzustellen.Um den Ablauf realer Feuerlöschertrainings möglichst genau nachzubilden, wurde der Simulator in drei verschiedene Lernszenarien unterteilt: Präsentation, Übung und Test. In Kombination mit dem gleichzeitigen Tracking von Händen und Controllern sowie der Implementierung grundlegender Löschmechaniken erschafft dieses System ein intuitives und immersives Trainingserlebnis.Durch einen Expertentest und eine User Study mit 29 Teilnehmern wurde die Applikation bewertet, wobei die Ergebnisse überwiegend sehr positives Feedback hinsichtlich Bedienbarkeit und Trainingsrelevanz zeigen. Trotz bestehender Einschränkungen, insbesondere in Bezug auf die Erzeugung von Stress und Realismus, hat der entwickelte Prototyp ein hohes Potenzial als ergänzende Alternative zu konventionellen Feuerlöschertrainings.Fire extinguishing is an essential skill in everyday life. The sudden outbreak of a fire can cause huge damage to the environment and even lead to fatalities. However, most people have never used a fire extinguisher in their whole lives, which can lead to wrong decisions in critical situations. Conventional fire extinguisher training aims to address this issue, but it requires significant organizational effort and financial costs. A promising alternative, by enabling repeatable, safe, and cost-efficient training scenarios, is offeredby Virtual Reality (VR).Existing VR simulators already simulate the structure of conventional training well,but most of them lack realism when it comes to physical feedback. Building on these approaches, this thesis developed a mixed reality (MR) fire extinguisher simulator that integrates a real, tracked fire extinguisher into the virtual environment and additionally enhances it with haptic feedback. Two major hardware constructions were implemented: a weight balance system to simulate mass loss during usage and a recoil simulation to represent the forces experienced while spraying.To closely resemble conventional training, three different scenarios were integrated: presentation, exercise, and test. Combined with simultaneous tracking of hands and controllers, and the implementation of basic fire extinguishing functionalities, the system provides an intuitive and immersive training experience.In an expert test and a user study with 29 participants, the application was evaluated and received great feedback regarding usability and training relevance through questionnaire answers. Although the system had limitations in terms of stress creation and realism, it demonstrates strong potential as a complementary addition to conventional fire extinguisher training
Faithful Attention Attribution in Vision Transformers for Chest X-Ray Interpretation
No full textVision Transformers (ViTs) achieve strong performance in natural and medical imaging, yet their decision processes remain opaque. This is especially problematic in high-stakes settings like chest X-ray interpretation. TransMM is among the strongest attribution methods for ViTs, combining attention with class-specific gradients to highlight influential image patches. We ask whether injecting semantic structure from Sparse Autoencoders (SAEs) can further improve the faithfulness of such attributions.We introduce Feature-Gradient Attribution, which extends TransMM’s principle from attention space to feature space. SAEs are trained on residual streams to decompose activations into sparse, interpretable features, providing per-patch feature activations. We project gradients onto the SAE feature basis and compute feature-gradient scores that capture both which learned features are present and how they influence the target logit. These scores yield per-patch gates that modulate TransMM’s attention maps before relevance propagation, forming a lightweight, semantically informed correction.Across three datasets (chest X-rays, endoscopy, natural images), two architectures (finetuned ViT-B/16 and contrastively pre-trained CLIP ViT-B/32), and three complementary faithfulness metrics, our method improves attribution faithfulness consistently. Improvements are statistically significant (p<0.001) on all three metrics for one dataset and on two of three metrics for the remaining datasets. We observe gains of 10.5-34.8% on SaCo and 9.7-43.0% on Faithfulness Correlation, with Pixel Flipping improving by 1.8-10.8%. Notably, we never observe degradation relative to TransMM on any metric–dataset combination
Extending Graph Neural Networks with Global Features
No full textWir untersuchen die Verwendung globaler Graphmerkmale, um die Ausdruckskraft und die Vorhersageleistung von graph neural networks (GNNs) zu verbessern. Während Message Passing Neural Networks (MPNNs) ein gängiger Ansatz für das Lernen von Graphrepräsentationen sind, sind sie in ihrer Ausdruckskraft beschränkt. Diese Einschränkung beeinträchtigt die Universalität von MPNNs und folglich die Arten von Grapheneigenschaften, die sie darstellen können.Um diese Herausforderungen anzugehen, schlagen wir vor, expressive globale Graphmerkmale in eine Basis-GNN-Architektur zu integrieren. Im Gegensatz zu herkömmlichen Ansätzen, die die Vorhersageleistung entweder durch Änderungen im Message Passing oder durch Modifikation der Graphen (z. B. Hinzufügen zusätzlicher Knotenmerkmale) verbessern, konzentriert sich unsere Methode auf globale Graphmerkmale, die die gesamte Graphstruktur beschreiben. Diese globalen Merkmale, die in Bereichen wie der Chemoinformatik gut etabliert, jedoch in der GNN Literatur oft übersehen sind, können verwendet werden, um wichtige Aspekte des Graphen zu erfassen, die von MPNNs nicht ausgedrückt werden können, wie etwa den Graphendurchmesser oder den längsten Zyklus.Wir ergänzen die von GNNs gelernten Embeddings mit ausgewählten globalen Graphmerkmalen und deren Einfluss auf die Ausdruckskraft und Modellleistung. Wir zeigen, dass 19 der 23 untersuchten globalen Merkmale, wie der Wiener- oder der Hosoya-Index, nachweislich die Ausdruckskraft von MPNNs erhöhen. Unsere empirischen Studien belegen, dass die Integration globaler Graphmerkmale die Vorhersagekraft von GNNs bei molekularen Benchmark-Datensätzen verbessern kann, was auf eine vielversprechende Richtung zur Steigerung der Effektivität von GNN-basierten Modellen hinweist.We investigate the use of global graph features to enhance the expressive power and predictive performance of graph neural networks (GNNs). While message passing neural networks (MPNNs) are a common approach for learning graph representations, they are inherently limited in their expressive power, which restricts their ability to distinguish between certain graph structures. This limitation impacts the universality of MPNNs, constraining the range of functions they can approximate and, consequently, the types of graph properties they can detect. To address these challenges, we propose a method for incorporating expressive global graph features into any baseline GNN architecture. Unlike traditional approaches that enhance predictive performance by either changing the message passing or modifying the graphs (e.g. attaching additional node features), our method focuses on global graph properties that describe the entire graph structure. These global features, well-established in fields like chemoinformatics but often overlooked by the GNN community, can be used to capture important aspects of the graph not expressible by MPNNs, such as the graph diameter or the longest cycle.We propose to extend the generated embeddings learned by GNNs with selected global graph features, analyzing their impact on expressiveness and model performance. We show that 19 out of the 23 global features we have investigated, like the Wiener or the Hosoya index, provably increase the expressivity of MPNNs. Our empirical studies demonstrate that incorporating global graph features can improve the performance of GNNs on molecular benchmark datasets, suggesting a promising direction for increasing the effectiveness of GNN-based models in graph-level tasks
Material Imperative
No full textDer landwirtschaftliche Hof meiner Familie in Mroków, Polen dient als Ausgangspunkt dieser Arbeit. Ein Ensemble, das sich über Jahrzehnte hinweg bricolagehaft weiterentwickelt hat und geprägt ist von improvisierten Ergänzungen und pragmatischer Wiederverwendung. Spuren sozialer und ökonomischer Umbrüche lassen ihn nicht als geschlossenes Werk erscheinen, sondern als lebendiges Materialarchiv. Heute stellt sich die Aufgabe, diesen Hof nicht durch einen radikalen Neubeginn, sondern durch ein Weiterschreiben möglichen Zukunftsszenarien zu unterziehen. Ausgehend von der These des Neuen Materialismus, Materie als aktiven Mitspieler zu begreifen und mitsamt der Folgen seiner Prozesse wahrzunehmen, entwickelt die Arbeit fünf Resonanzen, die den Umbauprozess beeinflussen: Iteration, Vernetzung, Ressourcensammlung, Mitsorge und Impulse der Ästhetik. Diese sind keine losgelösten Strategien, sondern verschränken sich im Prozess: Materialien werden gesammelt, in neue Beziehungen gesetzt und ihre Pflege schließlich in einer Ästhetik sichtbar, die Spuren nicht überdeckt, sondern fortschreibt. Das alte Bauernhaus als Katalysator setzt einen iterativen Prozess in Gang, der den Umbau des Hofes in vier Akten versteht. Damit zeigt sich die Architektur für den Hof nicht als Masterplan, sondern als prozessuales Weiterschreiben, das Raum lässt für unvorhersehbare Entwicklungen. Der materielle Imperativ bewirkt ein Umdenken im Umgang mit den Ressourcen unserer Welt und strebt ein Entwerfen mit ihnen an. Der Hof wird so zum Labor einer Umbaukultur, der Materie als wirksame Kraft erkennt und mit ihr neue Formen von Resonanz zwischen Mensch, Stoff und Kontext eröffnet.The agricultural farm of my family in Mroków serves as the starting point for this work. It‘s an ensemble that has evolved over decades in a bricolage-like manner, characterized by improvised additions and pragmatic reuse. Traces of social and economic upheaval render it not a complete work, but rather a living archive of materials. Today, the task is to subject this courtyard not through a radical new beginning, but rather through the continuation of possible future scenarios.Based on the thesis of New Materialism, which understands matter as an active participant and perceives it along with its processes and consequences, the work develops five resonances that influence the conversion process: iteration, networking, resource collection, shared care, and impulses of aesthetic. These are not isolated strategies, but intertwine in the process: Materials are collected, placed in new relationships, and their care ultimately becomes visible in an aesthetic that doesn‘t cover up traces but perpetuates them. The old farmhouse acts as a catalyst to initiate an iterative process that understands the conversion of the courtyard in four acts. The architecture for the courtyard thus presents itself not as a master plan, but as a processual continuation that leaves room for unforeseeable developments.The material imperative prompts a rethinking of how we use our world‘s resources and strives to design with them. The courtyard becomes a laboratory for a culture of transformation that recognizes matter as an effective force and uses it to open up new forms of resonance between people, material, and place