EuroVA2024

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/3606994

Browse

Now showing 1 - 15 of 15

Computing Fast and Accurate Decision Boundary Maps
(The Eurographics Association, 2024) Grosu, Cristian; Wang, Yu; Telea, Alexandru; El-Assady, Mennatallah; Schulz, Hans-Jörg
Decision boundary maps (DBMs) are image representations of the behavior of trained machine learning classification models. They are used in examining how the model partitions its data space into decision zones separated by decision boundaries and how this partition is influenced by the training data. However, all current DBM methods require significant computational effort, which precludes their use in interactive visual analytics scenarios. We present FastDBM, a set of techniques for the fast computation of DBMs. Our methods can accelerate any existing DBM algorithm by over one order of magnitude, yield results very similar to the original DBM methods, have a single parameter to set (with good presets), and are simple to implement. We demonstrate our method on various combinations of DBM techniques, datasets, and classification models.
Visualising the Invisible: Exploring Approaches for Visual Analysis of Dynamic Airflow in Geographic Environments Using Sensor Data
(The Eurographics Association, 2024) Zhang, Ying; Williams, Hannah; Schreiber, Falk; Klein, Karsten; El-Assady, Mennatallah; Schulz, Hans-Jörg
Measuring, modelling, and visualising dynamics of fluids, in particular air and water flow, are important in many applications, including engineering, biology, meteorology, and sports. While there are established models of airflow in controlled conditions, we are lacking an understanding of the distinct airflows in natural environments, whose characteristics are highly influential for all forms of aerial activities. There is little data available for fine-scale analysis and representation of airflow outside the lab without the help of sophisticated and laborious airflow measurements. Here, we explore ways to exploit movement data of flying agents for this purpose, propose an approach to model and visualise thermal dynamics as a discrete localised airflow, and demonstrate our approach using data collected from paraglider pilots.
cPro: Circular Projections Using Gradient Descent
(The Eurographics Association, 2024) Buchmüller, Raphael; Jäckl, Bastian; Behrisch, Michael; Keim, Daniel A.; Dennig, Frederik L.; El-Assady, Mennatallah; Schulz, Hans-Jörg
Typical projection methods such as PCA or MDS rely on mapping data onto an Euclidean space, limiting the design of resulting visualizations to lines, planes, or cubes and thus may fail to capture the intrinsic non-linear relationships within data, resulting in inefficient use of two-dimensional space. We introduce the novel projection technique -cPro-, which aligns high-dimensional data onto a circular layout. We apply gradient descent, an adaptable optimization technique to efficiently reduce a customized loss function. We use selected distance measures to reduce high data dimensionality and reveal patterns on a two-dimensional ring layout. We evaluate our approach compared to 1D and 2D MDS and discuss further use cases and potential extensions. cPro enables the design of novel visualization techniques that employ semantic distances on a circular layout.
DimenFix: a Novel Meta-Dimensionality Reduction Strategy for Feature Preservation
(The Eurographics Association, 2024) Luo, Qiaodan; Christino, Leonardo; Milios, Evangelos; Paulovich, Fernando V.; El-Assady, Mennatallah; Schulz, Hans-Jörg
Dimensionality Reduction (DR) methods have become essential tools for the data analysis toolbox. Typically, DR methods combine features of a multi-variate dataset to produce dimensions in a reduced space, preserving some data properties, usually pairwise distances or local neighborhoods. Preserving such properties makes DR methods attractive, but it is also one of their weaknesses. When calculating the embedded dimensions, through usually non-linear strategies, the original feature values are lost and not explicitly represented in the spatialization of the produced layouts, making it challenging to verify the features' contribution to the attained representations. Some strategies have been proposed to tackle this issue, such as coloring the DR layout or generating explanations. Still, they are post-processes, so specific features (values) are not guaranteed to be preserved or represented. This paper proposes DimenFix, a novel meta-DR strategy that explicitly preserves the values of a particular feature or external data (e.g., class, time, or ranking) in one of the embedded dimensions. DimenFix works with virtually any gradient-descent DR method and, in our results, has shown to be capable of representing features without heavily impacting distance or neighborhood preservation, allowing for creating hybrid layouts joining characteristics of scatter plots and DR methods.
Inverting Multidimensional Scaling Projections Using Data Point Multilateration
(The Eurographics Association, 2024) Blumberg, Daniela; Wang, Yu; Telea, Alexandru; Keim, Daniel A.; Dennig, Frederik L.; El-Assady, Mennatallah; Schulz, Hans-Jörg
Current inverse projection methods are often complex, hard to predict, and may require extensive parametrization. We present a new technique to compute inverse projections of Multidimensional Scaling (MDS) projections with minimal parametrization. We use mutilateration, a method used for geopositioning, to find data values for unknown 2D points, i.e., locations where no data point is projected. Being based on a geometrical relationship, our technique is more interpretable than comparable machine learning-based approaches and can invert 2-dimensional projections up to |D|−1 dimensional spaces given a minimum of |D| data points. We qualitatively and quantitatively compare our technique with existing inverse projection techniques on synthetic and real-world datasets using mean-squared errors (MSEs) and gradient maps. When MDS captures data distances well, our technique shows performance similar to existing approaches. While our method may show higher MSEs when inverting projected data samples, it produces smoother gradient maps, indicating higher predictability when inverting unseen points.
cVIL: Class-Centric Visual Interactive Labeling
(The Eurographics Association, 2024) Matt, Matthias; Zeppelzauer, Matthias; Waldner, Manuela; El-Assady, Mennatallah; Schulz, Hans-Jörg
We present cVIL, a class-centric approach to visual interactive labeling, which facilitates human annotation of large and complex image data sets. cVIL uses different property measures to support instance labeling for labeling difficult instances and batch labeling to quickly label easy instances. Simulated experiments reveal that cVIL with batch labeling can outperform traditional labeling approaches based on active learning. In a user study, cVIL led to better accuracy and higher user preference compared to a traditional instance-based visual interactive labeling approach based on 2D scatterplots.
EuroVa 2024: Frontmatter
(The Eurographics Association, 2024) El-Assady, Mennatallah; Schulz, Hans-Jörg; El-Assady, Mennatallah; Schulz, Hans-Jörg
Toward a Structured Theoretical Framework for the Evaluation of Generative AI-based Visualizations
(The Eurographics Association, 2024) Podo, Luca; Ishmal, Muhammad; Angelini, Marco; El-Assady, Mennatallah; Schulz, Hans-Jörg
The automatic generation of visualizations is an old task that, through the years, has shown more and more interest from the research and practitioner communities. Recently, large language models (LLM) have become an interesting option for supporting generative tasks related to visualization, demonstrating initial promising results. At the same time, several pitfalls, like the multiple ways of instructing an LLM to generate the desired result, the different perspectives leading the generation (code-based, image-based, grammar-based), and the presence of hallucinations even for the visualization generation task, make their usage less affordable than expected. Following similar initiatives for benchmarking LLMs, this paper explores the problem of modeling the evaluation of a generated visualization through an LLM. We propose a theoretical evaluation stack, EvaLLM, that decomposes the evaluation effort in its atomic components, characterizes their nature, and provides an overview of how to implement them. One use case on the Llama2-70-b model shows the benefits of EvaLLM and illustrates interesting results on the current state-of-the-art LLM-generated visualizations. The materials are available at this GitHub repository: https://github.com/lucapodo/evallm_llama2_70b.git
Transient Visual Analytics
(The Eurographics Association, 2024) Schulz, Hans-Jörg; Weaver, Chris; El-Assady, Mennatallah; Schulz, Hans-Jörg
Visual Analytics often utilizes progression as a means to overcome the challenges presented by large amounts of data or extensive computations. In Progressive Visual Analytics (PVA), data gets chunked into smaller subsets, which are then processed independently, and subsequently added to a visualization that completes over time. We introduce Transient Visual Analytics (TVA), which complements this incremental addition of data with progressive removal of data as it becomes outdated, starts to clutter the visualization, and generally distracts from the data that is currently relevant to visual analysis. Through combinations of various progressive addition and removal strategies, and supported by suitable analogies for the analyst and the software engineer, TVA captures a variety of visual analysis scenarios and approaches that are not well captured by PVA alone.
Persistent Interaction: User-Generated Artefacts in Visual Analytics
(The Eurographics Association, 2024) Pérez-Messina, Ignacio; Ceneda, Davide; Schetinger, Victor; Miksch, Silvia; El-Assady, Mennatallah; Schulz, Hans-Jörg
While traditional approaches in visual analytics (VA) prioritize insight generation and knowledge discovery, we argue that user-generated artefacts-annotations, model parameters, subset selections, spatializations, and other constructs-constitute a significant outcome of the analytical process. Drawing from theoretical models in VA literature, we introduce persistent interaction as techniques capturing user decisions. These interactions, called operations, provide a formalization of how users attach subjective judgments to datasets, condensing this input into artefacts serving specific purposes within broader workflows. We provide a description and classification of persistent interaction techniques and outcomes, demonstrating their practical implications in VA systems for system design, information transferability, and guidance capabilities.
Extending the Visual Data Exploration Loop towards Trustworthy Machine Learning in the Healthcare Domain
(The Eurographics Association, 2024) Antweiler, Dario; Fuchs, Georg; El-Assady, Mennatallah; Schulz, Hans-Jörg
Integration of machine learning (ML) systems into healthcare settings creates novel opportunities, including pattern recognition in heterogeneous medical datasets, clinical decision support as well as processes automation to save time, advance the quality of care, reduce costs and relieve healthcare staff. Challenges include opaque digital systems, curbed autonomy as well as requirements on communication, interaction and human-machine decision-making. Obstacles involve the interprofessional gap between data scientists and healthcare professionals (HCPs) during model development as well as the lack of trust into ML models. Visual Analytics (VA) enables versatile interactions between users and ML models via adaptable visualizations and has been successfully deployed to improve accuracy, identify bias and increase trust. However, specifically supporting HCPs to gain trust into ML models through VA systems is not sufficiently explored. We propose an extended visual data exploration framework towards trustworthy ML in the healthcare domain for multidisciplinary teams of data scientists, VA experts and HCPs. Additionally, we apply our framework to three real-world use cases for policy development, plausibility testing and model optimization.
Exploring Relationships between Events in Context
(The Eurographics Association, 2024) Andrienko, Natalia; Andrienko, Gennady; El-Assady, Mennatallah; Schulz, Hans-Jörg
We propose an approach to exploring interrelationships between two or more sequences of events when events occurring in one sequence both affect and are affected by events occurring in another sequences. We present the approach by example of exploring the dynamic relationships between COVID pandemic events and changes in population mobility behaviours across various countries. The key idea is to generate data capturing the temporal context of each event, i.e., what types of events occurred in different sequences within a specified time buffer around this event. An application of 2D space embedding to the context data reveals groups of events occurring in similar contexts. We can investigate the types of events each group consists of and see when and where these events and these contexts took place. By interactive or algorithmic clustering of the context data, we categorise event contexts based on their similarities, which allows us to compute, visualise, explore, and compare summary statistics of the context clusters, as well as exploring their distribution over time and other data dimensions.
Hierarchical Topic Maps for Visual Exploration and Comparison of Documents
(The Eurographics Association, 2024) Tytarenko, Mariia; Shao, Lin; Rutar, Tobias Walter; Bedek, Michael A.; Krenn, Cornelia; Lengauer, Stefan; Schreck, Tobias; El-Assady, Mennatallah; Schulz, Hans-Jörg
Information visualization nowadays provides a large amount of different text visualization techniques that help to summarize and present textual information in an intuitive and comprehensible manner. Despite many advancements, there remains a gap in effectively illustrating the thematic and structural distinction between similar documents in a hierarchical and interactive manner. We present the Hierarchical Topic Maps (HTM), an innovative approach, inspired by Tile Bars, that addresses this gap by illustrating the content distribution across a document hierarchically. Our model incorporates a multi-resolution display feature, enabling users, in particular curators of large document collections, with the need to quickly obtain text document structure, to delve deeper and draw more meaningful conclusions, to assess thematic similarities at multiple levels of detail, as well as facilitate nuanced comparison of textual documents. We demonstrate the effectiveness of both our approach's document exploration and document comparison potential by two exemplary use case scenarios. Our findings suggest that HTM not only simplifies the document overview process but also provides a practical solution for comparing thematic structures, thereby offering contributions to the field of text visualization and visualization analytics.
Tag-Xplore: Interactive Exploration of Annotation Practices in Digital Editions
(The Eurographics Association, 2024) Blum, Michael; Sachdeva, Madhav; Stricker, Yann; Mumenthaler, Rudolf; Bernard, Jürgen; El-Assady, Mennatallah; Schulz, Hans-Jörg
Digital Editions (DE) are scholarly document collections that make research artifacts accessible to both humans and machines in a structured manner, enriched with annotations. However, the interoperability and reusability of DE can be hampered by annotation inconsistencies within DE and heterogeneous annotation practices across DE. We present Tag-Xplore, an interactive and visual exploration tool for annotation practices within and across DE. Tag-Xplore offers multiple coordinated views that provide both attribute-based and document-based access to the huge search space at multiple granularities. The approach also provides rank, filter, and comparison techniques, to further support the exploration. With Tag-Xplore, data curators can validate assumptions based on existing knowledge and generate new insights about annotation practices. We demonstrate the usefulness of Tag-Xplore with two qualitative case studies on attribute ambiguity and outlier documents
Visually Supporting the Assessment of the Incident Management Process
(The Eurographics Association, 2024) Palma, Alessandro; Angelini, Marco; El-Assady, Mennatallah; Schulz, Hans-Jörg
Incident Management (IM) is the process to prevent, protect, and react to incidents affecting an organization and should be well-defined to be prepared in case of alerts. To this aim, security standards define guidelines to manage the incidents and the organizations should comply with them to properly set up a secure-by-design process. Assessing whether an organization is compliant or not with security standards requires a big effort as the main methodologies are based on manual analysis and leveraging automatic approaches to support human decisions is challenging. To facilitate this task, we design IMPAVID, a visual analytics solution to support the assessment of IM process compliance through process mining. The aim is to increase the level of awareness of the security assessor to support her in making informed decisions about actions to improve IM process compliance with regulatory and technical standards. We evaluate the proposed system through a usage scenario based on a publicly available dataset containing data from a real IM log of an IT company.

Browse

Browsing EuroVA2024 by Issue Date

Results Per Page

Sort Options