Interpolation between camera positions is a standard problem in computer graphics and visualization, and can be considered the foundation of camera path planning. As the basis for a new interpolation method, we introduce a new Riemannian metric in camera space, which measures the 3D image flow under a small movement of the camera. Building on this, we define a linear interpolation between two cameras as shortest geodesic in camera space, for which we provide a closed-form solution after a mild simplification of the metric. Furthermore, we propose a geodesic Catmull-Rom interpolant for keyframe camera animation. We compare our approach with several standard camera interpolation methods and obtain consistently better camera paths especially for cameras with extremely varying scales.
 

The development of new biomaterials and in particular implant materials is critical to resolve the socioeconomic challenges posed by demographic changes. Biodegradable implants made of magnesium-based alloys are of particular interest, as they degrade in the body while the tissue that they support is healing. However, the development and testing of these alloys needs to undergo numerous steps before they can be tested clinically, as the degradation process is highly complex and the implant interacts with its biochemical environment. Machine learning and computational modelling can support these testing steps by accelerating image analysis and enabling predictive modelling of certain phenomena. I will give an overview of the work we have recently performed and directions of future research in which machine learning methods will play a major role to improve (bio)materials development.

Exploration is a crucial skill for in-context reinforcement learning in unknown environments. However, it remains unclear if large language models can effectively explore a partially hidden state space. This work isolates exploration as the sole objective, tasking an agent with gathering information that enhances future returns. Within this framework, we argue that measuring agent returns is not sufficient for a fair evaluation. Hence, we decompose missing rewards into their exploration and exploitation components based on the optimal achievable return. Experiments with various models reveal that most struggle to explore the state space, and weak exploration is insufficient. Nevertheless, we found a positive correlation between exploration performance and reasoning capabilities. Our decomposition can provide insights into differences in behaviors driven by prompt engineering, offering a valuable tool for refining performance in exploratory tasks.  

Dr. Fahimeh Farhadifard: "Life Science applications for Deep Learning and AI"

This talk will give an overview of the practical application of state-of-the-art deep learning solutions in the field of life science and microscopy. It will showcase the transition of academic and industrial research into products and provide insights into the challenges of applying deep learning solutions to industry use cases and big data.

Paul Bönisch: "Production Rendering for Life Science and Microscopy"

This talk will give an overview of the unique challenges of developing a production renderer for life science and microscopy. It will discuss the balance of state-of-the-art rendering, hardware limitations and customer demands of visual fidelity, performance, and big data.

Large Language Models have shown capability to utilize In-Context Learning (ICL): They can simply adapt to a new task by using examples provided in the prompt, therefore allowing practitioners to avoid training a new model from scratch or fine-tuning a pre-trained model. So - what happens, if we just include a whole dataset in the prompt? Maybe even tabular datasets, which are traditionally challenging to handle using Machine Learning methods? In this presentation, we will have a look at the recent work by Hollman et al. that presented TabPFN, a foundation model for tabular datasets. We will look at the beginnings of Prior Fitted Networks (PFN), and the proposed changes to the transformer architecture that enable TabPFN to excel on tabular data. Finally, there will be a short demonstration how TabPFN can be applied in practice. Time will tell if TabPFN is just a hype or the beginning of a paradigm shift.

In a time of data abundance, automatic methods increasingly support manual mechanistic modeling. This support can range from data-driven calibration of some parameters up to uncovering the mechanics underlying a system from measurements. The Sparse Identification of Non-Linear Dynamics (SINDy) achieves both, allowing for the discovery of differential equations from time-series data. However, especially in biochemistry, but also in ecology and epidemiology, chemical reaction networks (CRNs) often act as target for (automatic) modeling. Their semantics require differential equations with a characteristic coupling.

In this talk, I will demonstrate how to adapt SINDy to the CRN formalism and present the findings published in our recent paper "Discovering Biochemical Reaction Models by Evolving Libraries", that was voted "best paper" at the CMSB 2024. We found that combining the extension of SINDy to CRNs with an evolutionary algorithm enables the integration of new kinds of prior knowledge and can increase the parsimony of learned models. The talk concludes by expanding the perspective to other automatic modeling methods which may support modeling in the future, e.g., inferring models from textual descriptions using large language models. 

Cultural institutions are digitizing their archives and collections, uncovering a wealth of previously largely unexhibited information. As these collections are presented on the web, the need for innovative interface and visualization approaches to make them digitally explorable is growing. Yet, concepts that go beyond conventional grid and search-based layouts, offering interfaces for unrestricted exploration of artifacts and artworks, remain rare. While humanist researchers emphasize the need for new humanistic visualization methods that are more sensitive to the specificity of cultural heritage data, cultural heritage data is also of interest to information visualization research for questions of representing uncertainty, computational analysis, provenance, bridging close and overview perspectives, serendipity, and engaging interaction concepts. Drawing from experiences with cultural collection visualization at the UCLAB (FH Potsdam), the talk will showcase research challenges, project processes, and outcomes from collaborations with cultural heritage collections, combining these insights with questions of human-computer interaction, design, collaborative methods, and critical practices—all with the aim of empowering people to interactively unfold the generosity of cultural collections.

Monitoring environmental change requires robust methods capable of handling the complexity of temporal-spatial data across diverse ecosystems. In this talk, I present my recent work leveraging machine learning and deep learning techniques applied to remote sensing data to address key challenges in environmental monitoring. Focusing on methods such as segmentation, regression, and representation learning, I demonstrate how these approaches enhance the analysis of satellite imagery and aerial point clouds. Through applications to forests and wetlands, we show how advanced learning techniques can offer critical insights for environmental management.

The goal of radiation therapy is to destroy tumor tissue with ionizing radiation, but patient movements, such as breathing, complicate precise targeting, risking damage to healthy tissue. Safety margins are used to account for this, but they can be minimized by continuously monitoring the tumor's position. While X-ray imaging and implanted markers are commonly used, 3D ultrasound offers real-time, nonionizing visualization of soft tissue, making it a promising alternative for therapy guidance. However, challenges like limited field of view and image quality need to be addressed. Robotic ultrasound systems, which automatically align the transducer, can help by continuously visualizing the target, requiring robust and real-time tracking for optimal treatment delivery.

Due to low image quality, volumetric image characteristics and high-dimensional soft-tissue motion, 4D ultrasound tracking is a challenging task. Only a few methods have been investigated so far, showing that there is a lack in 4D ultrasound tracking research. This work contributes to filling this gap by investigating the usability of representation learning using deep neural networks for the purpose of 4D ultrasound tracking. Training deep neural networks requires a substantial amount of data, but to date, only a limited amount of 4D ultrasound data is publicly accessible. This limited amount of data has been extended in this work in a 4D ultrasound labeling study. A novel 4D ultrasound data set has been made available containing image and landmark data. It has been shown that local image features can be detected and described in a unique and meaningful way in 3D ultrasound images using binary feature descriptors. In addition, it has been investigated that autoencoders are able to map 3D ultrasound patches into latent representations that can be used to identify similar soft-tissue structures and differentiate dissimilar ones. Therefor, different types of autoencoders were developed and investigated.

Different target tracking algorithms were developed. Algorithms working in ultrasound image space as well as in representation space created by autoencoders were implemented and evaluated. It has been shown that 4D ultrasound tracking in representation space can outperform image space-based tracking in terms of runtime while maintaining comparable accuracy. The target tracking methodology proposed in this work is based on unsupervised learning, is real-time capable, robust, and can be generalized across patients and organs, making it promising for ultrasound guided therapy purposes. The applicability of the representation space-based tracking has been shown in an online robotic ultrasound tracking experiment. Hence, this work proposes a novel method for 4D ultrasound tracking that could be integrated into any therapy domain.

For many decisions in our personal and professional lives, computing has made it easy to compile large numbers of options to choose from. With no objectively optimal solution present, identifying the best solution requires our human judgment to trade off conflicting goals. Data visualization is a powerful tool to help us explore and make sense of available courses of action. While many interactive visualizations already live in the context of decision-making, how to design for humans who make decisions with visualized data continues to be a vibrant research area. In this talk, I will share some ideas on how we can describe decision activities, design visualization tools for their support and validate their usefulness under real-world conditions. Through case studies, I will show how our visualizations helped users apply their preferences to balance the levels of performance that are achievable under different conditions. Finally, I will share my thoughts on future research directions that emerge in this area.

For many decisions in our personal and professional lives, computing has made it easy to compile large numbers of options to choose from. With no objectively optimal solution present, identifying the best solution requires our human judgment to trade off conflicting goals. Data visualization is a powerful tool to help us explore and make sense of available courses of action. While many interactive visualizations already live in the context of decision-making, how to design for humans who make decisions with visualized data continues to be a vibrant research area. In this talk, I will share some ideas on how we can describe decision activities, design visualization tools for their support and validate their usefulness under real-world conditions. Through case studies, I will show how our visualizations helped users apply their preferences to balance the levels of performance that are achievable under different conditions. Finally, I will share my thoughts on future research directions that emerge in this area.

Simulation is an important tool for understanding complex real-world situations and exploring different scenarios in a cost-efficient and risk-free manner. They find applications in a wide range of domains, from science (e.g., computational biology or physics) to engineering and the optimisation of work processes. Simulations are themselves complex software systems that require the careful collaboration between domain experts and software engineers, ensuring that the simulation implemented accurately represents the real-world scenario under study. In this talk, I explore how bespoke – domain-specific – modelling languages and automated transformations may help address some of the challenges faced by simulation engineers, including how to ensure adequate stakeholder involvement, how to reuse existing simulations in new contexts, and how to validate simulations and simulation studies. I will use examples from work in computational biology and in modelling patient flow in emergency care.

Deep Learning is nowadays the go-to solution for many machine learning tasks.  However, the inherent randomness within the training process introduces a curious phenomenon: basically equivalent or even the exact same training data with the same algorithm and parameterization may yield disparate outcomes. Despite achieving comparable aggregate performance, individual predictions and internal representations can exhibit significant variance. In this talk, we give a short introduction into this phenomenon from an empirical perspective, covering potential sources of instability, challenges in measuring, and potential application consequences.

Decision Trees (DTs) are commonly used for many machine learning tasks due to their high degree of interpretability. However, learning a DT from data is a difficult optimization problem, as it is non-convex and non-differentiable. Therefore, common approaches learn DTs using a greedy growth algorithm that minimizes the impurity locally at each internal node. Unfortunately, this greedy procedure can lead to inaccurate trees. In this talk, I present a novel approach for learning hard, axis-aligned DTs with gradient descent. The proposed method uses backpropagation with a straight-through operator on a dense DT representation, to jointly optimize all tree parameters. The approach outperforms existing methods on binary classification benchmarks and achieves competitive results for multi-class tasks.

Active Objects are an increasingly popular programming language paradigm, characterized by asynchronous task execution, strong data encapsulation, and explicit synchronization to minimize data races. We discuss, as a representative example, the Active Object language ABS and illustrate how it is used in different simulation scenarios. ABS has extensions for timed as well as for hybrid systems. Together with a high level of abstraction, this makes it an interesting candidate for simulation models.

Interactive visualization has evolved into an established field and is utilized by thousands of people every day. Currently, most visualization tools are optimized for desktop usage, and some for mobile devices. However, with the enormous push of Virtual Reality and Augmented Reality (VR/AR), these display modalities are gaining increasing importance in visualization, giving rise to the emerging areas of immersive and situated analytics. In this talk, I will reflect on the experiences from our 5-year endeavor to bridge Visualization and VR/AR research. Through case studies of our attempts at VR/AR-based visualizations, we will discuss pitfalls, challenges, and the opportunities that this area offers.

Recently, deep learning methods have been extremely successful for domains like natural language processing or computer vision. However, such data-driven methods are still insufficient for domains where training data is scarce, like modeling and prediction of ecosystem dynamics. On the other hand, mechanistic models or simulation models are less dependent on large training data sets, but can be less flexible and accurate when the true system dynamics is not completely captured by the model. In this talk, I will give an overview of research efforts that aim to integrate these two modeling concepts. Specifically, I will present methods for utilizing simulation models for training data generation, knowledge distillation, and neuro-symbolic models. I will give some examples from our own research, including hybrid models for modeling ecosystem dynamics and sensor-based activity recognition.

In my presentation on Digital Twins of the Ocean, I will start with a motivation for this topic and an overview of the general concept. It is followed by illustrative examples in the four areas that jointly enable the idea of a Digital Twin: Acquisition of marine data, (federated) data management, analytics & simulation, and visualization & interaction. By this, the talk gives a summary of the research activities of the chair Maritime Graphics and the maritime departments of Fraunhofer IGD.. 

Inverse rendering is the process of recovering the properties of a scene from images, such as its geometry, lighting, and materials. Neural radiance fields (NeRFs) are a type of deep learning model that can be used for inverse rendering. NeRFs represent a 3D scene as a continuous function that maps from 3D points to RGB colors and opacity values.

In this short talk,  I will give an introduction to inverse rendering and NeRFs. I will discuss the basic principles of inverse rendering, and how NERFs can be used to solve this problem. 

In this short presentation, I will cover the basic ideas of handling texts in neural networks, handling sequences and finally building transformer models. First, I will show how to represent and reason about textual data within neural networks, by introducing word2vec models - how to build and how to use them.  Then, I will show how to handle sequence data in recurrent architectures. Finally, I will introduce transformer models, by describing their general architecture, underlying ideas and assumptions and show how to train them.

The talk "A Zoo of Models" will give an overview of simulation studies conducted at the modeling and simulation group. The simulation studies address research questions as diverse as analyzing differences of mitochondria in healthy and less healthy cells, illuminating the interplay of osteoblasts and osteoclasts in their response to electrical stimulation, studying environmental influences on cod metabolism, or exploring possible effects of specific governmental regulations on recreational fishery. Extracts of simulation models and their provenance, including different data sources and simulation experiments, will be presented. The simulation studies will illuminate the role of different modeling approaches in answering the research questions and dealing with uncertainties in studying intra-cellular, inter-cellular, ecological, and socio-ecological systems.

Transitional Interfaces stellen eine zunehmend bedeutsame und bislang wenig erforschte Klasse von Benutzungsschnittstellen für die Mensch-Daten-Interaktion dar. Diese Benutzungsschnittstellen kombinieren die traditionelle Interaktion und Visualisierung mittels PCs, großen Touchscreens oder Mobilgeräten nahtlos mit Augmented und Virtual Reality. In Transitional Interfaces können zu jedem Zeitpunkt Mitglieder eines Teams individuell zwischen unterschiedlichen Positionen (bzw. "Kontexten") auf dem Realitäts-Virtualitäts-Kontinuum wechseln, um das für die aktuelle Aufgabe persönlich am besten geeignete Bedien- und Visualisierungskonzept zu wählen. Mein Vortrag wird zunächst den Begriff der "Mensch-Daten-Interaktion" bzw. Human-Data Interaction (HDI) vorstellen und welche Schlüsselrolle die menschliche Interaktion mit Visualisierungen dabei spielt. Dann werde ich die "Transitional Interfaces" als neues Forschungsthema innerhalb der HDI einführen, das verschiedene Aspekte der vorgenannten Themen in sich vereint.

Devesh Singh recently obtained his Master’s degree in Data and Knowledge Engineering from Otto-von-Guericke University Magdeburg. He now joined the DZNE Rostock as a PhD student working on deep learning explainability under the supervision of Dr. Martin Dyrba.

 In this VAC colloquium, he will present his two past projects:

• The Data Science for Social Good (DSSG) Short Term Fellowship with DFKI and TU Kaiserslautern was a full-time summer fellowship program which trained data scientists to work on machine learning projects with social impact. Devesh was working with data from the Paraguayan national public procurement agency (Dirección Nacional de Contrataciones Públicas (DNCP)) to detect fraudulent public procurement processes which might be an indicator of corruption and other inconsistencies. He worked on training NLP models to automatically analyze the data. The realized explainable model was four times better than the DNCP’s existing process.
   
• External Master Thesis at SICK Hamburg.
  SICK is a German manufacturing company whose industrial sensor solutions are used in logistical centers and large-scale warehouses. To further advance their existing image classification models, deep-learning based generative Image-to-Image (I2I) translation methods were explored under the larger setup of synthetic data generation. Devesh worked with two I2I approaches: CycleGANs and Contrastive Unpaired Translation (CUT). The realized I2I translation model was consistent with the semantic information given in the input images as well as it was able to produce textures from the desired target domain for the generated images. 

Samaneh Zolfaghari is currently doing her last year of PhD in the Mathematics and Computer Science department of the University of Cagliari, Italy. She started her PhD in the Autumn of 2019 and her research concentration is on sensor-based behavior and locomotion monitoring for healthcare applications. 

She obtained her M.S. and B.S. in Computer Engineering (Software) from Alzahra University and Shahid Rajaee University in 2016, and 2013, respectively in Iran. Her research interests are smart environments, AI and machine learning applications in health care, and human factors in pervasive computing. Currently, she’s serving as a visiting scholar at the University of Rostock working on the kitchen task assessment dataset, in order to detect challenging behaviors in cooking activities performed by people with cognitive impairments. The challenge here is to identify the actions the person is executing, and the goals he/she is following, to detect errors in behavior, and to reason about the causes of these errors.

Researchers are now able to engineer synthetic genetic circuits for a range of applications in the environmental, medical, and energy domains. Crucial to the success of these efforts is the development of methods and tools for genetic design automation (GDA). While inspiration can be drawn from experiences with electronic design automation (EDA), design with a genetic material poses several challenges. In particular, genetic circuits are composed of very noisy components making their behavior more asynchronous, analog, and stochastic in nature. This talk presents our research in the development of GDA tools that leverage our past experiences in asynchronous circuit synthesis and formal verification. These tools enable synthetic biologists to construct models, efficiently analyze and visualize them, and synthesize a genetic circuit from a library of parts. Each step of this design process utilizes standard data representation formats enabling the ready exchange of results.

Runtime systems are fundamental components of modern applications that provide the necessary support for executing programs and sometimes allow for advanced introspection capabilities. In this talk, I will provide an overview of recent and current research activities on runtime systems carried out by the High Performance and Dependable Computing Systems (HPDCS) research group at the University of Rome Tor Vergata. In the Software Engineering field, we show how a symbolic/concrete runtime environment allows behavioural similarity analysis, enabling effective test program prioritization and reducing the time-to-test. In High-Performance Computing, we show how a simple programming model allows for significant speedup of Discrete-Event Simulation models by relying on asynchronous/non-blocking algorithms. Similarly, we show that it is possible to improve the performance of task-based applications by explicitly addressing the detection of priority inversions of tasks. Concerning Cyber Security, we show how to carry out transparent system-wide detection of side-channel attacks leveraging off-the-shelf hardware components called Performance Monitoring Units. We also show how runtime environments could promote Anti-Tampering frameworks to make general applications harder to reverse engineer.

It is a truism in healthcare to state that the process of diagnosis is extremely important to patient care, increasing the likelihood of a positive outcome for the patient. However, diagnosis depends on information availability, which depends on the availability of sensors capable of eliciting the information the clinician needs. Direct observation by a clinician is naturally a starting point, but clinician time is limited and expensive. The use of large diagnostic tools offers powerful insight, but do not represent a panacea. Both of these can only provide a snapshot in time, in large part because many aspects of the patient’s lived experience simply do not show up in the timescale observed. As a result, there is considerable interest in the effectiveness of relatively small, cheap sensors, such as wearable accelerometers, as an additional data source to support the processes of diagnosis and evaluation of treatment.

The SPHERE project has explored the possibility of making use of multiple small, cheap sensors in the home to describe participants’ everyday life, testing systems in well over sixty homes over several years. The project used data drawn from a heterogeneous sensor network to support the detection of information such as activities of daily living, sleep and personal hygiene.

In this seminar, we look at the strengths and limitations of single- and multi-sensor approaches. We explore the challenges of heterogeneous (multi-sensor) system deployment, and the many difficulties that are involved in deployment of complex sensor networks in participant homes, from the effect of architecture on signal reception and networking to the identification of minimally-invasive methods of fixing removable sensors in place. We discuss the issues involved in recruiting participants, particularly the need to sensitively negotiate participant acceptance of privacy-invasive technologies such as silhouette cameras during the design process, and to identify and adopt privacy-enhancing mechanisms wherever appropriate.

We then discuss the challenges of the deployment process, particularly system monitoring, reliability engineering and repair, with a focus on whole-cohort management using configuration management systems and the urgent need to develop COVID-safe system engineering methods to safeguard vulnerable populations. Finally, we discuss lessons learned from analysing multi-sensor pervasive healthcare data alongside subject-expert clinicians, and suggest some recommendations from our own experience.

Self-adaptation equips a computing system with a feedback loop that enables it dealing with uncertainties during operation, such as sudden changes in the availability of resources or fluctuating workloads. One of the key challenges of engineering self-adaptive systems is ensuring that the system complies with the adaptation goals.  To that end, we defined ActivFORMS (Active FORmal Models for Self-adaptation). ActivFORMS spans four main stages of the life cycle of a feedback loop: design, deployment, runtime adaptation, and evolution. In this talk we give an overview of ActivFORMS and zoom in on one of its distinct features: the use of statistical model checking at runtime to select adaptation options that realise the adaptation goals with a required level of accuracy and confidence. We use an IoT application for building security monitoring that is deployed in Leuven as illustrative case. 

Die Covid-19 Pandemie hat EntscheidungsträgerInnen aus Politik und Gesundheitswesen vor beispiellose Herausforderungen gestellt. Primäres Ziel war und ist dabei die Grenzen der Gesundheitssysteme nicht zu überschreiten, um eine entsprechende Versorgung der Bevölkerung gewährleisten zu können. Aber mit zunehmendem Verlauf der Pandemie spielen auch andere Outcomes eine wichtige Rolle. Durch jahrelange Expertise und Forschung, in Kooperation mit der dwh GmbH und der TU Wien, verfügt die Forschungsplattform DEXHELPP über Werkzeuge (u.a. mit GEPOC und dem ABT-Framework), mit denen sowohl die Effizienz und Wirksamkeit einzelner Maßnahmen bzw. von Maßnahmenpaketen als auch deren Auswirkungen auf die zur Verfügung stehenden Ressourcen des Gesundheitssystems berechnet werden können. Der Vortrag "Datenbasis und Simulationsmodelle - notwendige Grundlagen für bessere Entscheidungen in der COVID19 Krise" wird dabei insbesondere auf die Rolle der Daten als Input, für die Kalibrierung und Validierung von Simulationsmodellen eingehen.

Weitere Informationen

The amount of biological data is continuously growing, the reason why the data analysis methods have to evolve accordingly. Throughout the project AVATARS various biological data about rapeseed such as genomes, transcriptomes, metabolomes, and phenomes are generated and visualized in virtual reality. Since all biological data domains are related, AVATARS aims to support a holistic understanding of several and complex data sets with immersive analytics.

In a first step, need-finding interviews were conducted to identify core and pressing data analysis needs of twelve biologists. Based on the insights, a first prototype has been developed. A tabular representation in VR is used to align genomes of several phenotypes. The prototype supports the user to get an overview of up to 5 million data values, to access details on demand and in real-time. Initial user feedback indicates that immersive analytics is beneficial to analyze large data sets.

Das übergeordnete Ziel unserer designorientierten Visualisierungsforschung ist es, grafische Benutzeroberflächen zu entwickeln, die die ergebnisoffene Erkundung komplexer Daten - z.B. des digitalen Kulturerbes - anregen. In unserem Vortrag stellen wir Konzepte, Methoden sowie Frage- und Zielstellungen vor, die hinter unserer Arbeit stehen.

Im ersten Teil gehen wir auf theoretische und praktische Herangehensweisen unserer Forschung in Potsdam ein. Philosophische Konzepte wie der Flaneur und die Falte können dabei helfen, lange bestehende Dichotomien wie die Unterscheidung zwischen Überblick und Detail oder die Trennung von Interaktion und Repräsentation zu hinterfragen. Zu unserem Methodenkanon zählen wir insbesondere das Co-Design, in dem für ein Projekt relevante Personen direkt in den Gestaltungs- und Forschungsprozess eingebunden werden. In diesem Modus können Konzepte für prototypische Interfaces und Visualisierungen entstehen, die umfassende Bestände wie Zeichnungen, Münzen, Handbibliotheken, Kunstsammlungen und Künstlernachlässe über zoombare, filterbare oder multiperspektivische Arrangements explorieren lassen.

Insbesondere in den Geisteswissenschaften führt die Visualisierung von Daten zu einer Reihe von Herausforderungen durch Dateneigenschaften wie Unsicherheit, Mehrdeutigkeit, Lokalität und Subjektivität. Diese Eigenschaften werden durch den Einsatz von statistischen und datenanalytischen Prozessen und Visualisierungsformen mit dem Ziel von Reduktion von Gesamtkomplexität häufig nicht ausreichend repräsentiert.

Im zweiten Teil des Vortrags geht es darauf aufbauend um die Erforschung von Interaktionen, Visualisierungen und animierten Übergängen zur Bewegung zwischen unterschiedlichen Abstraktions- und Granularitätsstufen von Daten. Im Bezug darauf wird die Forschungsfrage aufgeworfen, wie interaktiv situierte Detailerhöhung durch integrierte Visualisierungstechniken reichhaltigere, kontext-erhaltende Explorationserfahrungen ermöglichen kann.