David Steinmann
YOU?
Author Swipe
Navigating Shortcuts, Spurious Correlations, and Confounders: From Origins via Detection to Mitigation Open
Shortcuts, also described as Clever Hans behavior, spurious correlations, or confounders, present a significant challenge in machine learning and AI, critically affecting model generalization and robustness. Research in this area, however,…
Neural Concept Binder Open
The challenge in object-based visual reasoning lies in generating concept representations that are both descriptive and distinct. Achieving this in an unsupervised manner requires human users to understand the model's learned concepts and,…
View article: A new troglomorphic species of Larca (Pseudoscorpiones, Larcidae) from Colorado
A new troglomorphic species of Larca (Pseudoscorpiones, Larcidae) from Colorado Open
A new species of Larca is described from dry habitats in a cave in central Colorado. Like other cave-dwelling species of Larca , the new species Larca boulderica sp. nov. , shows relatively modest morphological adaptations, such as pale co…
View article: United We Pretrain, Divided We Fail! Representation Learning for Time Series by Pretraining on 75 Datasets at Once
United We Pretrain, Divided We Fail! Representation Learning for Time Series by Pretraining on 75 Datasets at Once Open
In natural language processing and vision, pretraining is utilized to learn effective representations. Unfortunately, the success of pretraining does not easily carry over to time series due to potential mismatch between sources and target…
Right on Time: Revising Time Series Models by Constraining their Explanations Open
Deep time series models often suffer from reliability issues due to their tendency to rely on spurious correlations, leading to incorrect predictions. To mitigate such shortcuts and prevent "Clever-Hans" moments in time series models, we i…
Learning by Self-Explaining Open
Much of explainable AI research treats explanations as a means for model inspection. Yet, this neglects findings from human psychology that describe the benefit of self-explanations in an agent's learning process. Motivated by this, we int…
Learning to Intervene on Concept Bottlenecks Open
While deep learning models often lack interpretability, concept bottleneck models (CBMs) provide inherent explanations via their concept representations. Moreover, they allow users to perform interventional interactions on these concepts b…
One Explanation Does Not Fit XIL Open
Current machine learning models produce outstanding results in many areas but, at the same time, suffer from shortcut learning and spurious correlations. To address such flaws, the explanatory interactive machine learning (XIL) framework h…
Tidal Flushing Rather Than Non-Point Source Nitrogen Pollution Drives Nutrient Dynamics in A Putatively Eutrophic Estuary Open
The effects of nonpoint source nutrients on estuaries can be difficult to pinpoint, with researchers often using indicator species, monitoring, and models to detect influence and change. Here, we made stable isotope measurements of nitroge…
Machines Explaining Linear Programs Open
There has been a recent push in making machine learning models more interpretable so that their performance can be trusted. Although successful, these methods have mostly focused on the deep learning methods while the fundamental optimizat…
Next Generation of LFD Testing – Smartphone-Based Mycotoxin Analysis with RIDA®SMART APP Open
Background Grain mills, feed mills, ethanol plants, flour producers, but also mobile laboratory services need simple and reliable possibilities to analyze grain for mycotoxins during harvest in the field. Objective The RIDA®SMART APP is a …
Cave millipedes of the United States. XV. Coloradesmus gen. nov. (Diplopoda, Polydesmida, Macrosternodesmidae), and four new species from caves in Colorado, USA Open
Coloradesmus , gen. nov., is established in the family Macrosternodesmidae based on Speodesmus aquiliensis Shear, 1984, comb. nov. and includes four new species: Coloradesmus hopkinsae sp. nov. , C. manitou sp. nov. , C. beckleyi sp. nov. …
FIGURE 1. Limnodrilus sulphurensis n in Limnodrilus sulphurensis n. sp., from a sulfur cave in Colorado, USA, with notes on the morphologically similar L. profundicola (Clitellata, Naididae, Tubificinae) Open
FIGURE 1. Limnodrilus sulphurensis n. sp. from Sulphur Cave, exterior and somatic characters. A. Anterior ends of two whole, unmounted worms. B. Ventral chaetal bundle in IV, from a mature worm (n=19, including 10 partially developed). C. …