Browsing by Author "Leitte, Heike"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Decision Boundary Visualization for Counterfactual Reasoning
    (Eurographics ‐ The European Association for Computer Graphics and John Wiley & Sons Ltd., 2023) Sohns, Jan‐Tobias; Garth, Christoph; Leitte, Heike; Hauser, Helwig and Alliez, Pierre
    Machine learning algorithms are widely applied to create powerful prediction models. With increasingly complex models, humans' ability to understand the decision function (that maps from a high‐dimensional input space) is quickly exceeded. To explain a model's decisions, black‐box methods have been proposed that provide either non‐linear maps of the global topology of the decision boundary, or samples that allow approximating it locally. The former loses information about distances in input space, while the latter only provides statements about given samples, but lacks a focus on the underlying model for precise ‘What‐If'‐reasoning. In this paper, we integrate both approaches and propose an interactive exploration method using local linear maps of the decision space. We create the maps on high‐dimensional hyperplanes—2D‐slices of the high‐dimensional parameter space—based on statistical and personal feature mutability and guided by feature importance. We complement the proposed workflow with established model inspection techniques to provide orientation and guidance. We demonstrate our approach on real‐world datasets and illustrate that it allows identification of instance‐based decision boundary structures and can answer multi‐dimensional ‘What‐If'‐questions, thereby identifying counterfactual scenarios visually.
  • Thumbnail Image
    Item
    EuroVis 2019 CGF 38-3: Frontmatter
    (The Eurographics Association and John Wiley & Sons Ltd., 2019) Gleicher, Michael; Viola, Ivan; Leitte, Heike; Gleicher, Michael and Viola, Ivan and Leitte, Heike
  • Thumbnail Image
    Item
    Topology-based Feature Detection in Climate Data
    (The Eurographics Association, 2019) Kappe, Christopher P.; Böttinger, Michael; Leitte, Heike; Bujack, Roxana and Feige, Kathrin and Rink, Karsten and Zeckzer, Dirk
    The weather and climate research community needs to analyze increasingly large datasets, mostly obtained by observations or produced by simulations. Ensemble simulation techniques, which are used to capture uncertainty, add a further dimension to the multivariate time-dependent 3D data, even tightening the challenge of finding relevant information in the data for answering the respective research questions. In this paper we propose a topology-based method to support the visual analysis of climate data by detecting regions with particularly strong local minima or maxima and highlighting them with colored contours. Combined with preceding clustering of the data fields, typical spatial patterns characterizing the climate variability are detected and visualized. We demonstrate the utility of our method with a study of global temperature anomalies of a 150-years ensemble simulation consisting of 100 members.