6 papers
On the Faithfulness of Post-Hoc Concept Bottleneck Models
Laines Schmalwasser, Jan Blunk, Niklas Penzel +2
Human decision-making interprets the world through high-level concepts, such as recognizing a bird by its belly color. To bridge the gap between opaque deep learning representation…
TCD-Arena: Assessing Robustness of Time Series Causal Discovery Methods Against Assumption Violations
Gideon Stein, Niklas Penzel, Tristan Piater +1
Causal Discovery (CD) is a powerful framework for scientific inquiry. Yet, its practical adoption is hindered by a reliance on strong, often unverifiable assumptions and a lack of…
Modeling COVID-19 Dynamics in German States Using Physics-Informed Neural Networks
Phillip Rothenbeck, Sai Karthikeya Vemuri, Niklas Penzel +1
The COVID-19 pandemic has highlighted the need for quantitative modeling and analysis to understand real-world disease dynamics. In particular, post hoc analyses using compartmenta…
Locally Explaining Prediction Behavior via Gradual Interventions and Measuring Property Gradients
Niklas Penzel, Joachim Denzler
Deep learning models achieve high predictive performance but lack intrinsic interpretability, hindering our understanding of the learned prediction behavior. Existing local explain…
FastCAV: Efficient Computation of Concept Activation Vectors for Explaining Deep Neural Networks
Laines Schmalwasser, Niklas Penzel, Joachim Denzler +1
Concepts such as objects, patterns, and shapes are how humans understand the world. Building on this intuition, concept-based explainability methods aim to study representations le…
CausalRivers -- Scaling up benchmarking of causal discovery for real-world time-series
Gideon Stein, Maha Shadaydeh, Jan Blunk +2
Causal discovery, or identifying causal relationships from observational data, is a notoriously challenging task, with numerous methods proposed to tackle it. Despite this, in-the-…