Options
Prof. Dr.
Wrobel, Stefan
Now showing
1  10 of 26

PublicationImplementation and evaluation of an additional GPT4based reviewer in PRISMAbased medical systematic literature reviews( 20240901)
;Landschaft, Assaf ;Mackay, Sina ;Höres, TimmAllendeCid, HéctorBackground: PRISMAbased literature reviews require meticulous scrutiny of extensive textual data by multiple reviewers, which is associated with considerable human effort. Objective: To evaluate feasibility and reliability of using GPT4 API as a complementary reviewer in systematic literature reviews based on the PRISMA framework. Methodology: A systematic literature review on the role of natural language processing and Large Language Models (LLMs) in automatic patienttrial matching was conducted using human reviewers and an AIbased reviewer (GPT4 API). A RAG methodology with LangChain integration was used to process fulltext articles. Agreement levels between two human reviewers and GPT4 API for abstract screening and between a single reviewer and GPT4 API for fulltext parameter extraction were evaluated. Results: An almost perfect GPT–human reviewer agreement in the abstract screening process (Cohen's kappa > 0.9) and a lower agreement in the fulltext parameter extraction were observed. Conclusion: As GPT4 has performed on a par with human reviewers in abstract screening, we conclude that GPT4 has an exciting potential of being used as a main screening tool for systematic literature reviews, replacing at least one of the human reviewers. 
PublicationWasserstein Dropout( 2024)
;Sicking, Joachim ;Pintz, Maximilian AlexanderFischer, AsjaDespite of its importance for safe machine learning, uncertainty quantification for neural networks is far from being solved. Stateoftheart approaches to estimate neural uncertainties are often hybrid, combining parametric models with explicit or implicit (dropoutbased) ensembling. We take another pathway and propose a novel approach to uncertainty quantification for regression tasks, Wasserstein dropout, that is purely nonparametric. Technically, it captures aleatoric uncertainty by means of dropoutbased subnetwork distributions. This is accomplished by a new objective which minimizes the Wasserstein distance between the label distribution and the model distribution. An extensive empirical analysis shows that Wasserstein dropout outperforms stateoftheart methods, on vanilla test data as well as under distributional shift in terms of producing more accurate and stable uncertainty estimates. 
PublicationExplainable production planning under partial observability in highprecision manufacturing( 202310)
;Volbach, PeterConceptually, highprecision manufacturing is a sequence of production and measurement steps, where both kinds of steps require to use nondeterministic models to represent production and measurement tolerances. This paper demonstrates how to effectively represent these manufacturing processes as Partially Observable Markov Decision Processes (POMDP) and derive an offline strategy with stateoftheart Monte Carlo Tree Search (MCTS) approaches. In doing so, we face two challenges: a continuous observation space and explainability requirements from the side of the process engineers. As a result, we find that a tradeoff between the quantitative performance of the solution and its explainability is required. In a nutshell, the paper elucidates the entire process of explainable production planning: We design and validate a whitebox simulation from expert knowledge, examine stateoftheart POMDP solvers, and discuss our results from both the perspective of machine learning research and as an illustration for highprecision manufacturing practitioners. 
PublicationMaximal closed set and halfspace separations in finite closure systems( 20230921)
;Seiffarth, FlorianSeveral concept learning problems can be regarded as special cases of halfspace 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 halfspace separation problem is NPcomplete. 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. 
PublicationRobustness in Fatigue Strength Estimation( 20221202)Fatigue strength estimation is a costly manual material characterization process in which stateoftheart approaches follow a standardized experiment and analysis procedure. In this paper, we examine a modular, Machine Learningbased approach for fatigue strength estimation that is likely to reduce the number of experiments and, thus, the overall experimental costs. Despite its high potential, deployment of a new approach in a reallife lab requires more than the theoretical definition and simulation. Therefore, we study the robustness of the approach against misspecification of the prior and discretization of the specified loads. We identify its applicability and its advantageous behavior over the stateoftheart methods, potentially reducing the number of costly experiments.

PublicationA Fast Heuristic for Computing Geodesic Closures in Large Networks( 20221106)
;Seiffarth, FlorianMotivated by the increasing interest in applications of graph geodesic convexity in machine learning and data mining, we present a heuristic for approximating the geodesic convex hull of node sets in large networks. It generates a small set of (almost) maximal outerplanar spanning subgraphs for the input graph, computes the geodesic closure in each of these graphs, and regards a node as an element of the convex hull if it belongs to the closed sets for at least a user specified number of outerplanar graphs. Our heuristic algorithm runs in time linear in the number of edges of the input graph, i.e., it is faster with one order of magnitude than the standard algorithm computing the closure exactly. Its performance is evaluated empirically by approximating convexity based coreperiphery decomposition of networks. Our experimental results with large realworld networks show that for most networks, the proposed heuristic was able to produce close approximations significantly faster than the standard algorithm computing the exact convex hulls. For example, while our algorithm calculated an approximate coreperiphery decomposition in 5 h or less for networks with more than 20 million edges, the standard algorithm did not terminate within 50 days. 
PublicationData Ecosystems: A New Dimension of Value Creation Using AI and Machine Learning( 20220722)Machine learning and artificial intelligence have become crucial factors for the competitiveness of individual companies and entire economies. Yet their successful deployment requires access to a large volume of training data often not even available to the largest corporations. The rise of trustworthy federated digital ecosystems will significantly improve data availability for all participants and thus will allow a quantum leap for the widespread adoption of artificial intelligence at all scales of companies and in all sectors of the economy. In this chapter, we will explain how AI systems are built with data science and machine learning principles and describe how this leads to AI platforms. We will detail the principles of distributed learning which represents a perfect match with the principles of distributed data ecosystems and discuss how trust, as a central value proposition of modern ecosystems, carries over to creating trustworthy AI systems.

PublicationA generalized WeisfeilerLehman graph kernel( 20220427)
;Schulz, Till Hendrik ;Welke, PascalAfter more than one decade, WeisfeilerLehman graph kernels are still among the most prevalent graph kernels due to their remarkable predictive performance and time complexity. They are based on a fast iterative partitioning of vertices, originally designed for deciding graph isomorphism with onesided error. The WeisfeilerLehman graph kernels retain this idea and compare such labels with respect to equality. This binary valued comparison is, however, arguably too rigid for defining suitable graph kernels for certain graph classes. To overcome this limitation, we propose a generalization of WeisfeilerLehman graph kernels which takes into account a more natural and finer grade of similarity between WeisfeilerLehman labels than equality. We show that the proposed similarity can be calculated efficiently by means of the Wasserstein distance between certain vectors representing WeisfeilerLehman labels. This and other facts give rise to the natural choice of partitioning the vertices with the Wasserstein kmeans algorithm. We empirically demonstrate on the WeisfeilerLehman subtree kernel, which is one of the most prominent WeisfeilerLehman graph kernels, that our generalization significantly outperforms this and other stateoftheart graph kernels in terms of predictive performance on datasets which contain structurally more complex graphs beyond the typically considered molecular graphs. 
PublicationVisual Analytics for HumanCentered Machine Learning( 20220125)
;Andrienko, Natalia ;Andrienko, Gennady ;Adilova, LinaraWe introduce a new research area in visual analytics (VA) aiming to bridge existing gaps between methods of interactive machine learning (ML) and eXplainable Artificial Intelligence (XAI), on one side, and human minds, on the other side. The gaps are, first, a conceptual mismatch between ML/XAI outputs and human mental models and ways of reasoning, and second, a mismatch between the information quantity and level of detail and human capabilities to perceive and understand. A grand challenge is to adapt ML and XAI to human goals, concepts, values, and ways of thinking. Complementing the current efforts in XAI towards solving this challenge, VA can contribute by exploiting the potential of visualization as an effective way of communicating information to humans and a strong trigger of human abstractive perception and thinking. We propose a crossdisciplinary research framework and formulate research directions for VA. 
PublicationMultiAgent Neural Rewriter for Vehicle Routing with Limited Disclosure of Costs( 2022)
;Paul, NathalieWe interpret solving the multivehicle routing problem as a team Markov game with partially observable costs. For a given set of customers to serve, the playing agents (vehicles) have the common goal to determine the teamoptimal agent routes with minimal total cost. Each agent thereby observes only its own cost. Our multiagent reinforcement learning approach, the socalled multiagent Neural Rewriter, builds on the singleagent Neural Rewriter to solve the problem by iteratively rewriting solutions. Parallel agent action execution and partial observability require new rewriting rules for the game. We propose the introduction of a socalled pool in the system which serves as a collection point for unvisited nodes. It enables agents to act simultaneously and exchange nodes in a conflictfree manner. We realize limited disclosure of agentspecific costs by only sharing them during learning. During inference, each agents acts decentrally, solely based on its own cost. First empirical results on small problem sizes demonstrate that we reach a performance close to the employed ORTools benchmark which operates in the perfect cost information setting.
 «
 1 (current)
 2
 3
 »