Christian Bartelt

Universität Mannheim ·  Lehrstuhl für Wirtschaftsinformatik

We present a new approach to goal recognition that involves comparing observed facts with their expected probabilities. These probabilities depend on a specified goal g and initial state s0. Our method maps these probabilities and observed facts into a real vector space to compute heuristic values for potential goals. These heuristic values estimat...

Outlier detection is a crucial analytical tool in various fields. In critical systems like manufacturing, malfunctioning outlier detection can be costly and safety-critical. Therefore, there is a significant need for explainable artificial intelligence (XAI) when deploying opaque models in such environments. This study focuses on manufacturing time...

We present a new approach to goal recognition that involves comparing observed facts with their expected probabilities. These probabilities depend on a specified goal g and initial state s0. Our method maps these probabilities and observed facts into a real vector space to compute heuristic values for potential goals. These values estimate the like...

Reinforcement learning (RL) has seen significant success across various domains, but its adoption is often limited by the black-box nature of neural network policies, making them difficult to interpret. In contrast, symbolic policies allow representing decision-making strategies in a compact and interpretable way. However, learning symbolic policie...

Neural networks often assume independence among input data samples, disregarding correlations arising from inherent clustering patterns in real-world datasets (e.g., due to different sites or repeated measurements). Recently, mixed effects neural networks (MENNs) which separate cluster-specific 'random effects' from cluster-invariant 'fixed effects...

Tabular data is prevalent in real-world machine learning applications, and new models for supervised learning of tabular data are frequently proposed. Comparative studies assessing the performance of models typically consist of model-centric evaluation setups with overly standardized data preprocessing. This paper demonstrates that such model-centr...

Neural networks often assume independence among input data samples, disregarding correlations arising from inherent clustering patterns in real-world datasets (e.g., due to different sites or repeated measurements). Recently, mixed effects neural networks (MENNs) which separate cluster-specific 'random effects' from cluster-invariant 'fixed effects...

Decision Trees (DTs) are commonly used for many machine learning tasks due to their high degree of interpretability. However, learning a DT from data is a difficult optimization problem, as it is non-convex and non-differentiable. Therefore, common approaches learn DTs using a greedy growth algorithm that minimizes the impurity locally at each inte...

Explainable Artificial Intelligence (XAI) is critical in unraveling decision-making processes in complex machine learning models. LIME (Local Interpretable Model-agnostic Explanations) is a well-known XAI framework for image analysis. It utilizes image segmentation to create features to identify relevant areas for classification. Consequently, poor...

Background and purpose Mobile stroke units (MSU) have been demonstrated to improve prehospital stroke care in metropolitan and rural regions. Due to geographical, social and structural idiosyncrasies of the German city of Mannheim, concepts of established MSU services are not directly applicable to the Mannheim initiative. The aim of the present an...

We consider generating explanations for neural networks in cases where the network’s training data is not accessible, for instance due to privacy or safety issues. Recently, Interpretation Nets (I\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsf...

Despite the success of deep learning for text and image data, tree-based ensemble models are still state-of-the-art for machine learning with heterogeneous tabular data. However, there is a significant need for tabular-specific gradient-based methods due to their high flexibility. In this paper, we propose , die t-Based ecision Tree nsembles, a...

Despite the success of deep learning for text and image data, tree-based ensemble models are still state-of-the-art for machine learning with heterogeneous tabular data. However, there is a significant need for tabular-specific gradient-based methods due to their high flexibility. In this paper, we propose GRANDE, Gradient-Based Decision Tree Ensem...

Goal recognition is an important problem in many application domains (e.g., pervasive computing, intrusion detection, computer games, etc.). In many application scenarios, it is important that goal recognition algorithms can recognize goals of an observed agent as fast as possible. However, many early approaches in the area of Plan Recognition As P...

The embedding spaces of image models have been shown to encode a range of social biases such as racism and sexism. Here, we investigate specific factors that contribute to the emergence of these biases in Vision Transformers (ViT). Therefore, we measure the impact of training data, model architecture, and training objectives on social biases in the...

The development dynamics of digital innovations for industry, business, and society are producing complex system conglomerates that can no longer be designed centrally and hierarchically in classic development processes. Instead, systems are evolving in DevOps processes in which heterogeneous actors act together on an open platform. Influencing and...

Outlying Aspect Mining (OAM) is the task of identifying a subset of features that distinguish an outlier from normal data, which is important for downstream (human) decision-making. Existing methods are based on beam search in the space of feature subsets. They need to compute outlier scores for all examined subsets, and thus rely on simple outlier...

Decision Trees (DTs) are commonly used for many machine learning tasks due to their high degree of interpretability. However, learning a DT from data is a difficult optimization problem, as it is non-convex and non-differentiable. Therefore, common approaches learn DTs using a greedy growth algorithm that minimizes the impurity locally at each inte...

Goal recognition is an important problem in many application domains (e.g., pervasive computing, intrusion detection, computer games, etc.). In many application scenarios it is important that goal recognition algorithms can recognize goals of an observed agent as fast as possible and with minimal domain knowledge. Hence, in this paper, we propose a...

In this paper, we propose a new ensemble method, which is called Dynamic Forest, for learning from data streams with varying feature spaces. Unlike traditional online learning where the feature space is static, in varying feature spaces, new features may emerge while others may vanish. This leads to several problems for which state-of-the-art onlin...

Outlier explanation is the task of identifying a set of features that distinguish a sample from normal data, which is important for downstream (human) decision-making. Existing methods are based on beam search in the space of feature subsets. They quickly becomes computationally expensive, as they require to run an outlier detection algorithm from...

Sum-Product Networks (SPNs) are expressive probabilistic models that provide exact, tractable inference. They achieve this efficiency by making use of local independence. On the other hand, mixtures of exchangeable variable models (MEVMs) are a class of tractable probabilistic models that make use of exchangeability of discrete random variables to...

We consider generating explanations for neural networks in cases where the network's training data is not accessible, for instance due to privacy or safety issues. Recently, $\mathcal{I}$-Nets have been proposed as a sample-free approach to post-hoc, global model interpretability that does not require access to training data. They formulate interpr...

Understanding the function learned by a neural network is crucial in many domains, e.g., to detect a model’s adaption to concept drift in online learning. Existing global surrogate model approaches generate explanations by maximizing the fidelity between the neural network and a surrogate model on a sample-basis, which can be very time-consuming. T...

Service interoperability for embedded devices is a mandatory feature for dynamically changing Internet-of-Things and Industry 4.0 software platforms. Service interoperability is achieved on a technical, syntactic, and semantic level. If service interoperability is achieved on all levels, plug-and-play functionality known from USB storage sticks or...

The automatic, sensor-based assessment of human activities is highly relevant for production and logistics, to optimise the economics and ergonomics of these processes. One challenge for accurate activity recognition in these domains is the context-dependence of activities: Similar movements can correspond to different activities, depending on, e.g...

Sum-Product Networks (SPNs) are expressive probabilistic models that provide exact, tractable inference. They achieve this efficiency by making used of local independence. On the other hand, mixtures of exchangeable variable models (MEVMs) are a class of tractable probabilistic models that make use of exchangeability of random variables to render i...

Competitive Multi-Agent Systems (MAS) are inherently hard to control due to agent autonomy and strategic behavior, which is particularly problematic when there are system-level objectives to be achieved or specific environmental states to be avoided. Existing solutions for this task mostly assume specific knowledge about agent preferences, utilitie...

Semantic interoperability for web services is still a problem. Although decentralized solutions such as describing the integration context with a formal mapping language or using a web service description language exist, practitioners rely on implementing software adapters manually. For IoT and Web of Things systems, current scientific solutions fa...

Applications provided by software intensive systems in an Internet of Things environment offer new business opportunities from the industry. An application describes the expected behavior of the software system. Thereby, the steps of a business process (e.g., event booking) are determined using objects from the Internet of Things environment at run...

In this paper, a new application domain for mobility predictions is presented. Based on the application domain new challenges arise in terms of when and how the mobility prediction has to be done. This results in three cases for mobility predictions, i.e. predicting some time beforehand, predicting before departure, and predicting when departed. Se...

In this work, we present iDropout, a new method to adjust dropout, from purely randomly dropping inputs to dropping inputs based on a mix based on the relevance of the nodes and some randomness. We use Deep Taylor Decomposition to calculate the respective relevance of the inputs and based on this, we give input nodes with a higher relevance a highe...

Functional and nonfunctional characteristics of software systems are defined by their architecture. Therefore, research streams such as Internet-of-Things (IoT) or component-based software engineering provide researchers and practitioners with construction guidelines for selected architectural characteristics. Current systems can be categorized in...

Manufacturing systems are currently being equipped with high-level application interfaces (e.g. OPC UA). Although these information models can be standardized by using domain-dependent standards (e.g. PLCopen), the semantics of services is hard to formalize. Especially in the industrial automation domain, current semantic service descriptions (e.g....

Using languages with formalized semantics for automating component integration is a well-established research area. As a consequence, independently developed software systems can interact without the need for manual integration effort in a " plug-and-play " manner. However, such dynamic adaptive system architectures are not widely used in Industria...

Many technological innovations from the research area of dynamic adaptive systems or IT ecosystems are already established in current software systems. Especially cyber-physical systems should benefit by this progress to provide smart applications in ambient environments of private and industrial space. But a proper and methodical engineering of cy...

The manufacturing of composite structures is expensive due to high material cost and the amount of manual labor needed. Current manufacturing technologies, e.g. filament winding, braiding or fiber placement technologies are offering a possibility of automated manufacturing to lower these costs. Nevertheless, these technologies are limited when it c...

In the early phases, software engineers use whiteboards and flip charts to create and discuss their ideas and later they transform manually the hand drawn pictures into machine readable models. During this transformation important sketch information, like the history of origin or some elements, will be lost. To solve this problem, we present a new...

Software is not self-supporting. It is executed by hardware and interacts with its environment. So-called software systems are complicated hierarchical systems. They are carefully engineered by competent engineers. In contrast, complex systems, like biological ecosystems, railway systems and the Internet itself, have never been developed and teste...

Most of the time developers make extensive use of software tools in a software development process to support them in their day-to-day work. One of the first and most important phases of this process is the design phase, but within this phase intuitive and easy to use tools, which support the creative but also collaborative workflow (parallel/distr...

Due to distributed development of complex technical systems like machine tools, different system components are modeled and simulated in independent program suits. Several standards specify exchange of model data, but communication during concurrent simulations is not standardized yet. Therefore, the SimBus (Simulation Bus) was developed to close t...

Most of the time developers make massive use of software tools in a software development process to support them in their day-to-day work. One of the first and most important phases is the design phase but within this phase tools are missing which support the creative but also collaborative workflow (parallel/distributed). At the moment engineers i...

Dynamic adaptive middleware solutions for component-based development have become very important for creating complex applications in recent years. Many different middleware systems have been developed. In addition, decentralized middleware systems have been developed for special areas such as ambient intelligence or generic middleware systems for...

The initial investment costs for a production plant are enormous – in many cases up to 100Mio. €, and the entire life-cycle is up to 30 years. In order to preserve existing investments and to retain the competitiveness of the production plant it continuously has to be developed and improved further. In consequence, basic modules (software, electric...

Software Engineering ist Teamarbeit. Schon im Vorfeld der Programmierung modelliert und verfeinert ein Team aus Softwareingenieuren und Fachseite gemeinschaftlich (kollaborativ) die zu realisierende Software. Zur Beschreibung von Software haben sich weitgehend standardisierte Modellierungssprachen wie die UML aber auch eine Reihe domänenspezifische...

Today the distribution of development locations, the co-evolution of models and the concurrency of work are typical for collaborative modeling in software projects. Software engineering teams demand modeling techniques at several abstraction levels to manage the complexity of software descriptions. Besides, software models are applied more and more...

Model driven engineering is one answer to increasing demands on software development and maintenance. Today's software systems are often large, complex but also safety-critical and should be highly adaptable in life cycle. The efficient development of large and complex software systems needs a high degree of collaboration in the design and specific...

In vielen industriellen Anwendungsbereichen des System- und Softwareengineering ist eine hochgradig parallele Entwicklungsarbeit notwendig. Zudem spielen formale, grafische Modellierungssprachen eine immer gr¨oßere Rolle. Ein Beispiel daf¨ur ist die ASCET-Modellierung von Motorsteuerger¨ate-Funktionen bei der Volkswagen AG. Hierbei stellt sich die...

Global software engineering has become a fact in many companies due to real necessity in practice. In contrast to co-located projects global projects face a number of additional software engineering challenges. Among them quality management has become much more difficult and schedule and budget overruns can be observed more often. Compared to co-lo...

Specification by models plays a decisive role, during the development process of complex systems. The division and concurrency of labor in teams is a further characteristic of such development. Therefore an efficient configuration and variant management of resulting documents is essential. In practice, a lot of established configuration management...

Whether an object-oriented or an agile approach – the new V-Modell XT fulfils many requirements. By changes in structure, it offers an overall view of the system development and it solves the tightrope walk between power and simplicity by tailoring a project perfectly. As is usual with a development proc-ess, the V-Modell XT expects creating and ma...

In these days the trend "everything, every time, everywhere" becomes more and more apparent. As consequence of this trend everyone has a lot of small or invisible devices in his direct environment, e.g. mobile phones, PDAs, or music players. Also some network technologies to connect the different devices like (W)LAN or Bluetooth moved mainstream.To...

Das V-Modell XT hat das etablierte V-Modell 97 als verbindlicher Standard zur Durchführung von IT-Projekten der öffentlichen Hand abgelöst. In diesem Beitrag ziehen wir ein erstes Mal Bilanz über die aktuell laufenden Pilotprojekte. Wir geben einen kurzen Überblick über die Charakteristiken der einzelnen Projekte und deren Umfang. Wir skizzieren di...

Ambient-intelligence (AmI) systems raise a series of new challenges in software and system development: Mobility, adaptability and heterogeneity are new concerns that have to be addressed. Many of these concerns are common and therefore should be addressed by a common AmI infrastructure instead of each individual application. This position paper pr...