Stefan Wrobel

TMLR Journal 2025 Journal Article

Detecting Systematic Weaknesses in Vision Models along Predefined Human-Understandable Dimensions

• Sujan Sai Gannamaneni
• Rohil Prakash Rao
• Michael Mock
• Maram Akila
• Stefan Wrobel

Slice discovery methods (SDMs) are prominent algorithms for finding systematic weaknesses in DNNs. They identify top-k semantically coherent slices/subsets of data where a DNN-under-test has low performance. For being directly useful, slices should be aligned with human-understandable and relevant dimensions, which, for example, are defined by safety and domain experts as part of the operational design domain (ODD). While SDMs can be applied effectively on structured data, their application on image data is complicated by the lack of semantic metadata. To address these issues, we present an algorithm that combines foundation models for zero-shot image classification to generate semantic metadata with methods for combinatorial search to find systematic weaknesses in images. In contrast to existing approaches, ours identifies weak slices that are in line with predefined human-understandable dimensions. As the algorithm includes foundation models, its intermediate and final results may not always be exact. Therefore, we include an approach to address the impact of noisy metadata. We validate our algorithm on both synthetic and real-world datasets, demonstrating its ability to recover human-understandable systematic weaknesses. Furthermore, using our approach, we identify systematic weaknesses of multiple pre-trained and publicly available state-of-the-art computer vision DNNs.

TMLR Journal 2025 Journal Article

Learning to Rank Features to Enhance Graph Neural Networks for Graph Classification

• Fouad Alkhoury
• Tamas Horvath
• Christian Bauckhage
• Stefan Wrobel

A common strategy to enhance the predictive performance of graph neural networks (GNNs) for graph classification is to extend input graphs with node- and graph-level features. However, identifying the optimal feature set for a specific learning task remains a significant challenge, often requiring domain-specific expertise. To address this, we propose a general two-step method that automatically selects a compact, informative subset from a large pool of candidate features to improve classification accuracy. In the first step, a GNN is trained to estimate the importance of each feature for a given graph. In the second step, the model generates feature rankings for the training graphs, which are then aggregated into a global ranking. A top-ranked subset is selected from this global ranking and used to train a downstream graph classification GNN. Experiments on real-world and synthetic datasets show that our method outperforms various baselines, including models using all candidate features, and achieves state-of-the-art results on several benchmarks.

TCS Journal 2023 Journal Article

Maximal closed set and half-space separations in finite closure systems

• Florian Seiffarth
• Tamás Horváth
• Stefan Wrobel

Several concept learning problems can be regarded as special cases of half-space separation in abstract closure systems over finite ground sets. For the typical scenario that the closure system is given via a closure operator, we show that the half-space separation problem is NP-complete. As a first approach to overcome this negative result, we relax the problem to maximal closed set separation, give a simple generic greedy algorithm solving this problem with a linear number of closure operator calls, and show that this bound is sharp. For a second direction, we consider Kakutani closure systems and prove that they are algorithmically characterized by the greedy algorithm. As a first special case of the general problem setting, we consider Kakutani closure systems over graphs and give a sufficient condition for this kind of closure systems in terms of forbidden graph minors. For a second special case, we then focus on closure systems over finite lattices, give an improved adaptation of the generic greedy algorithm, and present an application concerning subsumption lattices.

AAAI Conference 2022 Conference Paper

Graph Filtration Kernels

• Till Schulz
• Pascal Welke
• Stefan Wrobel

The majority of popular graph kernels is based on the concept of Haussler’s R-convolution kernel and defines graph similarities in terms of mutual substructures. In this work, we enrich these similarity measures by considering graph filtrations: Using meaningful orders on the set of edges, which allow to construct a sequence of nested graphs, we can consider a graph at multiple granularities. A key concept of our approach is to track graph features over the course of such graph resolutions. Rather than to simply compare frequencies of features in graphs, this allows for their comparison in terms of when and for how long they exist in the sequences. In this work, we propose a family of graph kernels that incorporate these existence intervals of features. While our approach can be applied to arbitrary graph features, we particularly highlight Weisfeiler-Lehman vertex labels, leading to efficient kernels. We show that using Weisfeiler-Lehman labels over certain filtrations strictly increases the expressive power over the ordinary Weisfeiler-Lehman procedure in terms of deciding graph isomorphism. In fact, this result directly yields more powerful graph kernels based on such features and has implications to graph neural networks due to their close relationship to the Weisfeiler-Lehman method. We empirically validate the expressive power of our graph kernels and show significant improvements over state-of-the-art graph kernels in terms of predictive performance on various real-world benchmark datasets.

ICRA Conference 2015 Conference Paper

Whole-body self-calibration via graph-optimization and automatic configuration selection

• Daniel Maier 0001
• Stefan Wrobel
• Maren Bennewitz

In this paper, we present a novel approach to accurately calibrate the kinematic model of a humanoid based on observations of its monocular camera. Our technique estimates the parameters of the complete model, consisting of the joint angle offsets of the whole body including the legs, as well as the camera extrinsic and intrinsic parameters. We cast the parameter estimation as a least-squares optimization problem. In the error function, we consider the residuals between camera observations of end-effector markers and their projections into the image based on the estimate of the calibration parameters. Furthermore, we developed an approach to automatically select a subset of configurations for the calibration process that yields a good trade-off between the number of observations and accuracy. As the experiments with a Nao humanoid show, we achieve an accurate calibration for this low-cost platform. Further, our approach to configuration selection yields substantially better optimization results compared to randomly chosen viable configurations. Hence, our system only requires a reduced number of configurations to achieve accurate results. Our optimization is general and the implementation, which is available online, can easily be applied to different humanoids.

TCS Journal 2010 Journal Article

Listing closed sets of strongly accessible set systems with applications to data mining

• Mario Boley
• Tamás Horváth
• Axel Poigné
• Stefan Wrobel

We study the problem of listing all closed sets of a closure operator σ that is a partial function on the power set of some finite ground set E, i. e. , σ: F → F with F ⊆ P ( E ). A very simple divide-and-conquer algorithm is analyzed that correctly solves this problem if and only if the domain of the closure operator is a strongly accessible set system. Strong accessibility is a strict relaxation of greedoids as well as of independence systems. This algorithm turns out to have delay O ( | E | ( T F + T σ + | E | ) ) and space O ( | E | + S F + S σ ), where T F, S F, T σ, and S σ are the time and space complexities of checking membership in F and computing σ, respectively. In contrast, we show that the problem becomes intractable for accessible set systems. We relate our results to the data mining problem of listing all support-closed patterns of a dataset and show that there is a corresponding closure operator for all datasets if and only if the set system satisfies a certain confluence property.

ICML Conference 2006 Conference Paper

Efficient co-regularised least squares regression

• Ulf Brefeld
• Thomas Gärtner 0001
• Tobias Scheffer
• Stefan Wrobel

In many applications, unlabelled examples are inexpensive and easy to obtain. Semi-supervised approaches try to utilise such examples to reduce the predictive error. In this paper, we investigate a semi-supervised least squares regression algorithm based on the co-learning approach. Similar to other semi-supervised algorithms, our base algorithm has cubic runtime complexity in the number of unlabelled examples. To be able to handle larger sets of unlabelled examples, we devise a semi-parametric variant that scales linearly in the number of unlabelled examples. Experiments show a significant error reduction by co-regularisation and a large runtime improvement for the semi-parametric approximation. Last but not least, we propose a distributed procedure that can be applied without collecting all data at a single site.

ICML Conference 2004 Conference Paper

A comparative study on methods for reducing myopia of hill-climbing search in multirelational learning

• Lourdes Peña Castillo
• Stefan Wrobel

IJCAI Conference 2003 Conference Paper

Learning Minesweeper with Multinational Learning

• Lourdes Pena Castillo
• Stefan Wrobel

Minesweeper is a one-person game which looks deceptively easy to play, but where average human performance is far from optimal. Playing the game requires logical, arithmetic and probabilistic reasoning based on spatial relationships on the board. Simply checking a board state for consistency is an NP-complete problem. Given the difficulty of hand-crafting strategies to play this and other games, AI researchers have always been interested in automatically learning such strategies from experience. In this paper, we show that when integrating certain techniques into a general purpose learning system (Mio), the resulting system is capable of inducing a Minesweeper playing strategy that beats the winning rate of average human players. In addition, we discuss the necessary background knowledge, present experimental results demonstrating the gain obtained with our techniques and show the strategy learned for the game.

JMLR Journal 2002 Journal Article

Finding the Most Interesting Patterns in a Database Quickly by Using Sequential Sampling

• Tobias Scheffer
• Stefan Wrobel

Many discovery problems, e. g. subgroup or association rule discovery, can naturally be cast as n -best hypotheses problems where the goal is to find the n hypotheses from a given hypothesis space that score best according to a certain utility function. We present a sampling algorithm that solves this problem by issuing a small number of database queries while guaranteeing precise bounds on the confidence and quality of solutions. Known sampling approaches have treated single hypothesis selection problems, assuming that the utility is the average (over the examples) of some function --- which is not the case for many frequently used utility functions. We show that our algorithm works for all utilities that can be estimated with bounded error. We provide these error bounds and resulting worst-case sample bounds for some of the most frequently used utilities, and prove that there is no sampling algorithm for a popular class of utility functions that cannot be estimated with bounded error. The algorithm is sequential in the sense that it starts to return (or discard) hypotheses that already seem to be particularly good (or bad) after a few examples. Thus, the algorithm is almost always faster than its worst-case bounds.

ICML Conference 2001 Conference Paper

Incremental Maximization of Non-Instance-Averaging Utility Functions with Applications to Knowledge Discovery Problems

• Tobias Scheffer
• Stefan Wrobel

IJCAI Conference 1991 Conference Paper

Towards a Model of Grounded Concept Formation

• Stefan Wrobel