Der Connectom-Fonds bietet Seed-Funding für gemeinsame Forschung zwischen hessian.AI-Mitgliedern und weiteren Kolleg_innen beteiligter Hochschulen. Er wird aus Mitteln des Landes Hessen finanziert und wird von der Technischen Universität Darmstadt als federführende Partnerin in hessian.AI ausgeschrieben. Aus Mitteln des Fonds werden zeitlich begrenzte Vorhaben im gesamten Spektrum von Forschung, Lehre, Demonstratoren-/Prototypenerstellung oder Anwendung mit bis zu 100.000 EUR je Vorhaben gefördert.
Bis zu 100,000 € pro Projekt
Insgesamt 1.2 Mio €
Förderzeitraum 2025
Die Bewertung und Auswahl der zu fördernden Projektskizzen erfolgt durch eine Auswahlkommission, die sich aus Vertreterinnen und Vertretern aller an hessian.AI beteiligten Hochschulen zusammensetzt. Die eingereichten Skizzen werden auf Basis folgender Kriterien bewertet:
Antragsberechtigt sind professorale hessian.AI-Mitglieder und DEPTH-Nachwuchsgruppenleiter_innen. Kolleg_innen anderer Hochschulen können an den Projekten beteiligt sein.
Einsendungen möglich bis 28.02.2025
Jury, die die Institutionen von hessian.AI repräsentiert
Prof. Dr. Gemma Roig, Goethe-University, Department of Computer Sciences
Dr. Med. Andreas Bucher (KGU)
The project focuses on revolutionizing medical imaging by integrating AI-driven image processing and Explainable AI (xAI). This approach will enable efficient biomarker identification for multiple medical hypotheses, overcoming the time constraints of human analysis. Automated image analysis, or opportunistic screening, enhances the capabilities of radiologists and provides a wider range of diagnostic information beyond traditional methods, which is essential for personalized medicine. The goal is to develop a comprehensive imaging biomarker platform that identifies a wider range of health indicators missed by standard clinical evaluation, thereby revolutionizing diagnostics.
Prof. Dr. Anke Becker, Philipps-Universität Marburg, FB Biologie / Synmikro
Prof. Dr. Bernd Freisleben, Philipps-Universität Marburg, FB Mathematik / Informatik
In this project, called AI4DNA, we focus on developing a new AI-based encoder/decoder (“codec”) approach supported by machine learning (ML) methods capable of adapting to different storage requirements to ensure optimal error tolerance and coding efficiency. In particular, we plan to use deep neural networks (NNs) and leverage recent DNA-related foundation models based on transformer architectures to create novel DNA codecs that will be adaptable to different storage conditions, utilizing a realistic AI-based DNA channel model. Furthermore, we combine fountain codes with the ML methods to obtain a hybrid rule-aware AI approach to provide error correction and data recovery in DNA storage.
Prof. Dr. Gemma Roig, Goethe-University, Department of Computer Sciences
Prof. Dr. Georgia Chalvatzaki, TUDa, Department of Computer Sciences
How can robots learn robust and generalizable object representations that are useful for tasks, such as manipulating objects, grasping an object, avoiding an object during navigation, etc.? One common way to approach this problem is to learn such objects representations while the robot is learning to perform the specific task, or by deploying pretrained visual models which do not necessarily provide relevant features for robotic task learning. Such techniques might result in object representations that are not generalizable to other scenarios and are not easily transferable to other tasks. Another possibility could be to first learn a general object representation from the robot’s perspective, which is then used to adapt to various tasks. Just like babies and toddlers, robots can inject sequential information from the environment with different sensory modalities. As we explored in previous work on computational modeling visual representation learning in objects, self-learning objectives following the slowness principle and the multi-modality co-occurrence with aligning visual input with sparse and noisy language representations lead object representations that contain both categorical and instance information. Moreover, in prior work we have shown how information-theoretic self-supervised objectives can be leveraged to provide sparse object representations that enable predictive modeling and learning for control, while preliminary work on adapting pretrained visual-language models for robot control affirmed that while pretrained models can help speed up learning, they do not lead to robust and generalizable robot task-oriented learning. In this project, we propose to analyze the impact of object representation learning on the ability of the robot to perform certain tasks. How does including language alignment improve performance, and to what degree is it robust to noise and sparsity at the input level while training? We will adapt the framework proposed in. We will first extend the simulation environment from which contains only toys, to other object categories. Then, we will use the same strategy to train visual models, as well as visual language models, with our dataset. It is important to note that for a thorough analysis of the impact of the pretrained model and what factors are determining the generalization to manipulation tasks to the robot, we need control over the dataset that we use for training, the models, and the level of sparsity and noise of the language input for the alignment. The generation of a simulated dataset enables fine-grained adjustment of all these parameters.
Prof. Stefan Roth, TUDa, Department of Computer Sciences
Prof. Simone Schaub-Meyer, TUDa, Department of Computer Sciences
In recent years, deep learning (DL) has established itself as an integral part of artificial intelligence (AI) for computer vision. Increasingly complex deep neural networks (DNNs) and larger amounts of data have allowed researchers to achieve unrivalled results in various areas, such as semantic segmentation and image generation. While this might be seen as a triumph of DL, DNNs still come with the critical limitation that humans do not comprehend how they actually work. As a consequence, they often receive only limited user trust and should not inconsiderately be applied to safety-critical domains, such as autonomous driving or medical imaging. To solve this problem and open up the „black box“, the field of explainable AI (XAI) aims to better understand DNNs and how they function. In particular attribution research, i.e., understanding how relevant each input feature is for the classification of a single data point, has been a major focus of existing work. While this simple setup is a necessary first step, it rarely helps to gain a significantly greater understanding of the model under inspection, especially for more complex image and video analysis tasks beyond classification.
With this initiative, we aim to assemble researchers at the forefront of XAI research and computer vision to jointly discuss the foundation, requirements, and expectations for XAI beyond simple attributions. In an inaugural symposium, we will establish common grounds, question established explanation types, and dis- cover synergies towards XAI methods that truly promote the understanding of DNNs. Potential topics include mechanistic interpretability, the proper evaluation of XAI methods, prototype-based explanations, and XAI beyond classification. The insights and personal connections gained from the symposium will serve as a basis for future collaborations, potential grant applications, and internationally visible workshop proposals.
Prof. Dr. Horst Zisgen, h_da
tbd
Prof. Christin Seifert, University of Marburg, Department of Computer Sciences
Prof. Dr. Gemma Roig, Goethe-University, Department of Computer Sciences
tbd
Prof. Dr. Oliver Hinz, Goethe-University
tbd
Prof. Dr. Bernhard Humm, Hochschule Darmstadt
tbd
Prof. Dr. Anirban Mukhopadhyay, TUDa,
Bei der Entwicklung von KI-gestützter Technologie für das Gesundheitswesen, das ein hohes Risiko birgt und sehr menschlich ist, spielen öffentliches Engagement und offene Gespräche eine zentrale Rolle für die gesellschaftliche Bereitschaft für die Technologie. Daher ist die Verbreitung von Wissen im Gespräch mit den verschiedenen Interessengruppen von entscheidender Bedeutung und sollte nicht als nachträglicher Gedanke behandelt werden. Die globale akademische Gemeinschaft im Bereich der medizinischen KI nutzte jedoch das Potenzial des öffentlichen Engagements nicht in vollem Umfang aus, da es keinen entsprechenden Kanal gab.
Der „AI-ready Healthcare“-Podcast schließt diese Lücke. Podcasts sind ein großartiges Medium für die Verbreitung von Wissen, konstruktiven Argumenten und tiefgründigen, aufschlussreichen Diskussionen. Indem wir uns fest auf die fortschrittliche KI-Technologie stützen, die in der akademischen Gemeinschaft entwickelt wurde, erkunden wir die dynamische Landschaft der KI im Gesundheitswesen. Oft spreche ich zusammen mit meinem Co-Moderator Henry Krumb mit internationalen Akteuren aus ganz unterschiedlichen Bereichen, wie beispielsweise Kolleginnen aus der medizinischen KI, ärztlichen Wissenschaftlerinnen, Industriepartnerinnen, Regulierungsbehörden, Patientenvertreterinnen und globalen Gesundheitsbefürworter*innen, um nur einige zu nennen.Diese Gespräche führen zu tiefen Einblicken in die translationalen Aspekte der KI-Forschung in der klinischen Versorgung, die oft in traditionellen Kommunikationsformen wie begutachteten Fachartikeln nicht thematisiert werden. Der Podcast hat eine doppelte Wirkung: Er erweitert den Horizont technischer Probleme für akademische Forscher und erhöht die Sichtbarkeit der medizinischen KI-Forschung erheblich.
Prof. Sascha Steffen, Frankfurt School of Finance & Management
tbd
Prof. Lucas Böttcher, Frankfurt School of Finance & Management, Computational Science
The goal of personalized medicine is to tailor interventions for maintaining or restoring an individual’s health based on their unique biology and life situation. Key to this vision are computational models known as medical digital twins (MDTs), which integrate a wide range of health-related data and can be dynamically updated. Medical digital twins play a growing role in predicting health trajectories and optimizing the impact of inter- ventions to guide a patient’s health effectively. While MDTs are increasingly adopted in biomedicine, their high dimensionality, multiscale nature, and stochastic characteristics complicate tasks related to optimization and control, such as the development of treatment protocols.
Recent advancements in neural-network control methods show great promise in addressing difficult control problems. However, their application to biomedical problems is still in its early stages. In this project, our goal is to develop dynamics-informed neural network controllers that leverage existing knowledge about the structural characteristics of MDTs to effectively address optimal control problems in medicine. This encompasses tasks like minimizing side effects while removing pathogens. We will illustrate the effectiveness of the proposed control approaches using an MDT focused on pulmonary aspergillosis, a common respiratory fungal infection.
Dr. Anirban Mukhopadhyay, TUDa, FB Elektrotechnik und Informationstechnik
Prof. Dr. Jan Peters, TUDa, FB Computer Sciences
Ultrasound-guided percutaneous needle navigation is a common task for clinicians to perform diagnoses (biopsy) and treatments (ablation, neoadjuvant therapy). Such procedures require skilful movement of the needle in conjunction with the U/S probe. Depending on the angle of insertion, the needle is barely visible on the U/S image. However, this task is challenging with patient and tissue movement and requires two steady hands under massive cognitive load. The increasing incidence of cancer in Europe’s ageing population, combined with the lack of qualified clinical staff, makes the timely diagnosis and treatment of cancer a critical health issue. Our goal is to develop an intelligent robotic assistant that can enable unskilled healthcare professionals to perform routine cancer diagnosis and treatment.
Prof. Hongbin Zhang, TUDa, Department of Materials Sciences
Prof Grace Li Zhang, TUDa, Department of Electrical Engineering and Information Technology
The project aims to develop an adaptive design framework based on neural networks and to use it for the multi-criteria design of advanced materials. To achieve this goal, robust latent space constructions and uncertainty quantification are required. However, these two requirements cannot be directly met with the conventional weight-based neural networks with single-valued predictions. In this project, we will apply error suppression and compensation to improve the robustness of neural networks in dimensionality reduction, and the uncertainties will be modelled and evaluated using Bayesian neural networks based on statistical training. These implementations will be used to perform adaptive multi-objective design of novel functional materials, including permanent magnets, thermoelectric materials and high-entropy alloys, using the existing databases and further developed into a generic framework for future autonomous experiments.
Prof. Dr. Dominik L. Michels, TUDa, FB Computer Science
Prof. Dr. Bernd Freisleben, Philipps-Universität Marburg, FB Mathematics and Computer Science
Prof. Dr. Kristian Kersting, TUDa, Department of Computer Science
Plant protection products are used in agricultural fields to protect crops from pests, diseases and weeds. On the one hand, this favours the cultivation practices of closely timed, monotonous crop rotations, but is accompanied by highly problematic losses in landscape and biodiversity. Furthermore, pesticide residues end up in animal feed and human food and are thus also consumed by consumers. Cereals can currently only be grown to the extent required (climate change, war in Ukraine, etc.) with the use of enormous quantities of plant protection products such as pesticides, herbicides, fungicides and insecticides. As a result, however, around 385 million people are poisoned by pesticides every year. In order to reduce the dramatic consequences of the current form of cereal cultivation for nature, humans and animals, the use of pesticides must be reduced as far as possible. Instead of applying pesticides on a large scale to entire fields, they should only be applied in a targeted manner to those areas where this is really indicated. This requires an automatic early detection of diseases and pests.The aim of this research project is therefore to develop a prototype for an early detection system for pests in cereal crops.Firstly, a flying drone (UAV) will record close-up images of various groups of cereal plants in the field using a high-resolution camera.These images will then be analysed using computer vision for signs of the so-called Hessian fly (lat. Mayetiola destructor, eng.Hessian fly), one of the most problematic pests in cereal cultivation.A suitable insecticide can then be sprayed on the affected areas in a targeted manner (precision spraying).The sub-discipline of machine learning (ML) in AI has suitable deep neural networks (deep learning) that can be used to recognise areas yellowed by the presence of the Hessian fly.
Prof. Dr. Carlo D’Eramo, University of Würzburg and TUDa, FB Computer Science
Prof. Dr. Georgia Chalvatzaki, TUDa, FB Computer Science
Current demographic trends and reports of a shortage of carers make the need for intelligent robots that can act as universal assistants essential.While robot learning promises to equip robots with complex skills through experience and interaction with the environment, most methods are too tailored to individual tasks that do not generalise well in the non-stationary real world.Conversely, humans are constantly learning and building on existing knowledge.Lifelong learning of robots requires that an agent can form representations that are useful for continuous learning of a set of tasks to avoid catastrophic forgetting of previous skills. We propose to investigate a method that enables robots to learn a set of behaviours through multimodal cues that can be easily assembled to synthesise more complex behaviours. We propose to incorporate large pre-trained base models for language and vision into robot-oriented tasks. We will explore the design of novel parameter-efficient residuals for lifelong reinforcement learning (RL), which would allow us to build on previous representations to learn new skills while avoiding the two main problems of task interference and catastrophic forgetting.Crucially, we investigate forward and backward transfer and inference from the perspective of explainability to enable robots to explain non-experts similarities they have found during their training life in different tasks, and even to translate their actions during the execution of a task into natural language explanations.We argue that explainability is a crucial component to increase the trustworthiness of AI robots during their interaction with non-expert users.Therefore, we call this area of work Lifelong Explainable Robot Learning (LExRoL), which opens new avenues of research in the field of lifelong learning and robotics.
Prof. Dr.-Ing. Andreas Koch, TUDa, Department of Embedded Systems and Applications (ESA)
Prof. Dr. Kristian Kersting, TUDa, Department of Computer Sciences
The Artificial Intelligence and Machine Learning LAB (AIMLL) and the Department of Embedded Systems and their Applications (ESA) have been working on various aspects of Sum Product Networks (SPNs) for several years. SPNs are a machine learning model closely related to the class of probabilistic graphical models and allow the compact representation of multivariate probability distributions with efficient inference.These properties allow, for example, neural networks to be augmented in order to increase their accuracy and allow predictions to be evaluated.In the context of SPNs, SPFlow is one of the most relevant software libraries.SPFlow is primarily developed and maintained by AIMLL and allows different types of SPNs to be created and trained quickly and easily. SPFlow offers a wide range of options for creating SPNs: Standard training routines can be used as well as customised training approaches. In addition, SPFlow can be extended so that the models, training and inference can be adapted accordingly. As part of a loose collaboration between AIMLL and ESA, a project dealing with the acceleration of SPN inference was launched back in 2018. The expertise in relation to MLIR and the acceleration of SPNs is to be used to extend the SPNC so that it can also be compiled for AI engines (AIEs) in the future. The overall aim of the project is to develop AI engines as a possible target architecture for SPN inference.In particular, various options for optimising the models are also to be evaluated, as new architectures in particular can often only be used optimally if corresponding peculiarities of the architecture and the corresponding models are exploited.The SAICo project proposal thus optimally combines the compiler/architecture-specific experience of the ESA research group with the model-related expertise of AIMLL.
Dr. Simone Schaub-Meyer, TUDa, Department of Computer Sciences
Dr. Martin Mundt, TUDa, Department of Computer Sciences
Much of the recent progress in artificial intelligence (AI) has been focussed on improving performance numbers or qualitatively appealing examples. This may be due in part to the way we set up a traditional machine learning workflow. Typically, we start by defining a limited, well-defined task, collect data for that task, select a statistical model to learn it, and later typically conclude that an approach is successful if it performs well on the appropriate test data.This process is often repeated several times to improve the result, either by optimising the model (model-centric view) or by collecting or processing more data (data-centric view).However, well-designed real-world systems require more than just the best performance number in a benchmark. Dedicated test sets are limited not only by the fact that they do not take into account the variance of the appearance of new, unknown data during deployment, but also the different ways in which tasks change over time. A single good number in a test set therefore does not reflect the experiences and changes that a system undergoes in the real world. As much as we seem to know how to tune our popular static benchmarks, we seem to know less about how to formulate general learning processes that are able to continually learn from endless streams of data, and to adapt and generalize to modified scenarios, as we humans do. Continual, or lifelong, machine learning addresses the crucial questions that arise when aiming to overcome the limitations of single training cycles, rigid inference engines, and fixed datasets. In contrast to conventional benchmarks for machine learning, the project investigates how learners can continue to use, expand and adapt their knowledge when confronted with changing and novel tasks over time. At the heart of this is the realisation that data selection, models, training algorithms and evaluation benchmarks are not static. Despite their recent popularity, current research has only just begun to understand how we can accommodate these factors in human-like, lifelong learning AI systems. Although ongoing efforts are beginning to consider complex sequences of datasets, they are predominantly focused on image classification tasks.Unfortunately, such tasks greatly simplify learning, e.g. by only performing image-level recognition and assuming that all data is always labelled and that continuous learning mainly consists of recognising new object types.In this initiative, we therefore want to bring together researchers who are leaders in the field of continuous learning and computer vision to jointly lay the foundations for continuous learning beyond classification. In an inaugural symposium, we will identify commonalities and synergies with respect to continuous learning for a variety of relevant computer vision tasks, such as semantic segmentation and learning without classification.
Dr. Anirban Mukhopadhyay, TUDa, Department of Electrical Engineering and Information Technology
Prof. Dr. Heinz Koeppl, TUDa, Department of Electrical Engineering and Information Technology
The trend towards increasingly resource-intensive models is at odds with the goal of democratising deep learning for all. Favourable access to AI technology and a low barrier to entry for participation are necessary to promote and fully exploit the potential of participation around the world. The possibility of widespread collaboration encourages the collection of diverse data, which ultimately enables complex problems to be solved. This includes frugal digital health innovations that could provide access to healthcare for the last billion people. Federated learning is a viable solution, but is inaccessible due to high network bandwidth and computing requirements for clinics that cannot afford such infrastructure.The recently emerging field of neural cellular automata (NCA), which are models that converge to a defined target only through local communication, is in contrast to this as they are lightweight in terms of parameters and computational requirements.However, training NCAs in a centralised environment is already difficult and federated training of NCAs has never been attempted. We propose to (1) combine the expertise on self-organisation with the experience from the development of the first NCA for medical image segmentation to develop a novel lightweight federated NCA learning and inference algorithm. (2) The developed algorithms will be tested within the established collaboration with Peter Wild from the University Hospital Frankfurt for the segmentation of histopathology images. (3) As an extreme example, the capability of federated learning with NCA over a VAST network of low-cost computing devices is demonstrated.
Prof. Dr. Lambert Alff, TUDa, FB Material Sciences
Prof. Dr.-Ing. Andreas Koch, TUDa, Department of Embedded Systems and Applications (ESA)
Prof. Dr. Christian Hochburgen, TUDa, FB ETIT
Künstliche Intelligenz wird quasi ubiquitär in die verschiedensten Lebensbereiche Eingang finden. Dies bedeutet aber auch gleichzeitig, dass der Aspekt der Energieeffizienz des damit verbundenen Rechenaufwands immer wichtiger werden wird. Daher ist die Entwicklung der für die KI eingesetzten Computerarchitekturen ein wichtiges Forschungsfeld. Ein wesentliches neues Bauteil für KI-angepasste Rechnerarchitekturen ist der sogenannte Memristor. Es gibt mehrere materialwissenschaftliche Ansätze, mit denen (hochgradig energieeffiziente) Memristoren realisiert werden können, aus denen aber unterschiedliches Bauteilverhalten resultiert bzw. realistische Einsatzszenarien noch gar nicht erforscht wurden. Dieses Projekt zielt darauf ab, die für Memristoren notwendige Kette vom Material über das Bauteil hin zur Schaltung für bestimmte Anwendungen der KI zusammenzubringen und gemeinsame Forschungsprojekte im Sinne dieses interdisziplinären und holistischen Ansatzes zu befördern.
Prof. Dr. Gemma Roig, Goethe-University, FB Informatik
The Algonauts project is an ongoing project that aims to explore human and machine intelligence using the latest algorithmic tools (Cichy et al., 2019). In this way, the Algonauts project serves as a catalyst to bring together researchers from the fields of biological and machine intelligence on a common platform to exchange ideas and advance both fields in the form of challenges and joint workshops (http://algonauts.csail.mit.edu/). Here we propose to leverage the ongoing success of the Algonauts project to motivate young talents, including high school and undergraduate students, to become the future leaders in the field of AI from an interdisciplinary perspective. To this end, we aim to build a demonstrator that shows how AI models, in particular artificial deep neural networks, can be used to reveal the workings of the brain that lead to human behaviour, and how the insights gained can be used to guide the design of brain-inspired AI models that could have desirable properties similar to human cognition, such as modularity of functions. This could shed light on how to support more transparent and explainable behaviour of model decisions, as well as how to develop models that are robust to perturbations and noise. The demonstrator is intended to be interactive and provide a user-friendly interface to go through the three main steps of using AI models to understand the human brain. Step 1 is to collect brain data from people viewing images or videos, step 2 is to select existing AI models or create your own model to explain the brain data, and the third step is to compare the two to gain insights into what is happening inside the brain while people are viewing the stimuli. To this end, we will integrate into the core of the demonstrator our laboratory toolbox called Net2Brain, whose purpose is to integrate AI models to predict brain data (Bersch et al., 2022). We will improve and further develop it so that it can later be made available to the scientific community. An important goal is the integration of AI models developed in hessian.AI, e.g. those with human learning characteristics, such as continuous learning models (Dr Mundt) and embodied models from robotics (Prof Chalvatzaki), as well as interpretability and explainability algorithms.
Prof. Dr. Antje Jahn, Darmstadt University of Applied Sciences, Department of Mathematics and Natural Sciences
Prof. Dr. Sebastian Döhler, Darmstadt University of Applied Sciences, Department of Mathematics and Natural Sciences
Prof. Dr. Gunter Grieser, Darmstadt University of Applied Sciences, Department of Computer Science
Prof. Dr. Bernhard Humm, Darmstadt University of Applied Sciences, Department of Computer Science
In machine learning (ML), open-source and free software packages for common ecosystems such as R and Python are responsible for the transfer into practice and the entry point for many beginners into the data science field. Advanced users make the informed choice of their tools through the information from academia, where systematic evaluations of different packages and implementations support this choice for a specific application purpose. The overall goal of this project is to make knowledge about new methods for high-dimensional and right-censored data available to beginners and experienced users. Hochdimensionale Daten treten zum Beispiel im Gebiet des Text Mining oder der genetischen Statistik auf. Basierend auf aktuellen Forschungen zu Goodness-of-Fit (GOF) Tests für hochdimensionale Daten soll ein R-Paket erstellt werden, welches die Anwendung dieser Methoden für die genannten Gebiete vereinfacht. Rechtszensierte Daten treten häufig in medizinischen Daten oder im Gebiet des predictive maintenance auf. Examples include predictions of survival probabilities under various medical interventions or the prediction of optimal maintenance times for technical devices. Right-censored data requires special data science methods, for which there is sometimes insufficient support for an informed selection of implementations. This support is provided in this project. The transfer of the results achieved is to take place in the form of a multimedia campaign – consisting of a short video channel, a video channel and a blog – under the aspect of the ‘third mission’.
Prof. Dr. Michael Guckert, THM, Department of Mathematics, Natural Sciences and Data Processing
Prof. Dr. Thomas Farrenkopf, THM, Department of Mathematics, Natural Sciences and Data Processing
Prof. Dr. Nicolas Stein, THM, Department of Mathematics, Natural Sciences and Data Processing
Prof. Holger Rohn, THM, Department of Industrial Engineering and Management
Prof. Dr. Udo Fiedler, THM, Department of Industrial Engineering and Management
Prof. Dr. Carsten Stroh, THM, Department of Industrial Engineering and Management
Small and medium-sized enterprises (SMEs) are under increasing pressure to innovate these days. They operate with limited financial and human resources and therefore have to operate complex production structures, often with one-off and small batch production. Effizienzsteigerung in der Produktion besitzt häufig existentielle Bedeutung. Limited capacities also force them to make efficient use of the available production resources while at the same time meeting the increasing quality requirements of the markets.
Artificial intelligence (AI) in production can be used in a wide range of corporate processes and bring about lasting effects. Eine systematische, automatisierte Erhebung von Daten, die während der Fertigung unmittelbar in den Maschinen anfallen, erlaubt eine konsequente Anwendung von KI-Algorithmen und unterstützt genauere Vorhersagen der tatsächlichen Ressourcennutzung. Aus den Daten gewonnene Erkenntnisse können Prognosen von Ausbringungsmengen und -qualitäten oder Maschinenverfügbarkeiten ermöglichen. Unmittelbare Effekte einer solchen intelligenten Maschinen- und Prozessüberwachung sind eine höhere Liefertreue, eine effizientere Auslastung der Ressourcen im Unternehmen (inkl. Energie- und Ressourceneffizienz) und eine erhöhte Transparenz über den Zustand der eingesetzten Fertigungsanlagen. In order to advance the level of AI maturity in SMEs, the high potential of the technology is to be illustrated with the help of demonstrators in a real laboratory environment. The impetus already given for the introduction of AI can be used as a lever and systematic implementation can be started in addition to the already known application possibilities. In addition to the processes, the operational use in a laboratory environment is also shown using demonstrators.
Prof. Dr. Melanie Siegel, Darmstadt University of Applied Sciences, Department of Computer Science
Prof. Dr. Bernhard Humm, Darmstadt University of Applied Sciences, Department of Computer Science
The original idea of social media was to enable people to exchange information and opinions as openly as possible and thus support communication. This idea of social participation is being massively disrupted by current developments: where an open exchange of opinions on political issues was possible, the forums are increasingly being flooded with hate and threats of violence. Where free access to information was the goal, false factual claims are increasingly being made and in some cases automatically disseminated. Texts, images and videos are used and semantically linked to each other. It is becoming increasingly difficult for children and young people in particular to categorise information. There are two basic ways in which toxic information can be recognised: intrinsically by analysing and evaluating published content or extrinsically by evaluating such content in the context of other information. One must be able to classify a post as harmless banter or opinion, insult, discrimination, or even threat.
In addition, a distinction must be made between a harmless private false claim, a socially relevant false claim that should at least be commented on journalistically, and even acts of disinformation that are relevant under criminal law. Automatic processes can help with categorisation, as the DeTox project has already shown. However, the topics and language of toxic content are constantly changing, so it is necessary for the models (automatic processes – intelligent systems) to be regularly retrained. In the case of models based on neural networks, however, further training can lead to previously trained content being overwritten and the models no longer functioning on the original (old) data (‘catastrophic forgetting’). Complete retraining is usually not practicable due to the high model complexity and the associated high computational effort. False reports are not only made up of language (text). In order to transfer opinions, images and texts are often combined from a different context and placed in a new, non-existent context. Dies macht eine menschliche und automatische Erkennung besonders schwierig. Therefore, approaches are needed that analyse the text and the image in context.
Prof. Dr. Oliver Hinz, Goethe University Frankfurt, Department of Economics
Prof. Dr. Matthias Blonski, Goethe University Frankfurt, Department of Economics
The aim of the project at the interface between AI and microeconomics is to understand how cooperative behaviour changes with repeated interaction with learning machines instead of humans. The following research questions are considered: How does the willingness to cooperate in the repeated prisoner’s dilemma change when one of the players is replaced by an artificially intelligent algorithm? How does this willingness depend on the expected duration of the game and the knowledge of the human about the identity of the opponent? Do any differences in cooperative behaviour result from the changed nature of the opponent (human or machine) or from deviating strategic behaviour?
Prof. Dr. Dominik Heider, Philipps-Universität Marburg, FB Mathematics/Computer Science
Prof. Dr. Thorsten Papenbrock, Philipps-Universität Marburg, FB Mathematics/Computer Science
Prof. Dr. Bernd Freisleben, Philipps-Universität Marburg, Mathematics/Computer Science
The COVID pandemic has exposed the weaknesses of healthcare systems worldwide and the immense pressure doctors are under. In addition, the WHO estimates that there will be a shortage of 12.9 million healthcare professionals by 2035. The Virtual Doc project aims to support medical staff through the use of advanced sensor technologies and state-of-the-art artificial intelligence (AI) methods. The virtual doctor performs various medical tasks on a patient in an intelligent examination cubicle. The sensors in the cabin measure non-invasive parameters (e.g. BMI, heart rate, pulse) and the computer infrastructure interactively records the patient’s medical history to avoid invasive measurements. Clinical parameters are made available to physicians, including advanced disease predictions based on machine learning models for specific (or as yet unknown) conditions (e.g. type 2 diabetes mellitus (T2DM)). In this way, the virtual doctor can relieve medical staff of these tasks, freeing up capacity for treatment, emergencies and care. With this project proposal, we want to expand our existing prototype of the virtual doctor with additional sensors and analysis modules and eliminate potential sources of error. We also want to strengthen collaboration in this multi-faceted project by involving other research groups and their expertise in the development of the virtual doctor. The extent to which such an AI-supported preliminary examination makes sense and is accepted by the population will be investigated in parallel with the help of a survey (cooperation with Prof Dr Michael Leyer, Department of Economics, University of Marburg) and an on-site test in a double cabin at Bochum University Hospital (cooperation with Prof Dr Ali Canbay, UK RUB).
Prof. Dr. Kawa Nazemi, Hochschule Darmstadt, Department of Computer Sciences
Prof. Dr. Bernhard Humm, Hochschule Darmstadt, Department of Computer Sciences
New technologies, as well as already existing but unused technologies, have the potential to sustainably increase the innovative capacity of companies and secure their future success. However, if these relevant technologies and the associated new areas of application are not identified early enough, competitors may establish themselves in these fields ahead of time. Furthermore, neglected new technologies carry the risk of disrupting the corresponding market when they are introduced, potentially displacing unprepared companies. A valid analysis and prediction of potential future technologies is therefore more important than ever.The VAVTech project aims to develop a visual analysis system that enables people to recognize relevant technologies as early as possible and predict their potential trajectory. Scientific publications will serve as the data foundation for the analysis system, as they present respective technologies at a very early stage, making them suitable for early technology detection.The system will primarily combine neural networks and interactive visualizations, allowing companies, startups, and strategic consultants to analyze and predict the potential of new and largely unknown technologies. The neural network will be developed in a modular way to ensure its transferability to other domains. As part of the project, a functional demonstrator will be created using real-world data, laying the foundation for further work in the field of strategic foresight through the application of artificial intelligence methods.The demonstrator will serve multiple purposes: acquiring third-party funding, networking with other AI researchers, and increasing the visibility of the research through visualizations.
Prof. Dr. Barbara Klein, Frankfurt University of Applied Sciences, Department of Social Work and Health
Prof. Dr. Martin Kappes, Frankfurt University of Applied Sciences, Department Computer Sciences and Engineering
The UNESCO Recommendation on the Ethics of Artificial Intelligence establishes globally accepted standards for AI technologies, to which 193 member states have committed. Ethical guidelines are linked to human rights obligations, with a focus on so-called „blind spots,“ such as AI and gender, education, sustainability, and others. For Germany, there is significant need for action in the areas of equal treatment and diversity within AI development teams. Diversity is considered one of the prerequisites for ensuring that such considerations are appropriately reflected in AI programming.The field of artificial intelligence in Germany, among other things, requires a higher proportion of women to avoid future social biases and gender inequalities caused by unconscious biases in algorithms.This aligns with the UNESCO Recommendation, which was adopted on November 23, 2021, as the first globally negotiated legal framework for the ethical development and use of AI. Gerade im Gesundheitswesen und der Medizin sind Frauen unzureichend berücksichtig, was zu fatalen Auswirkungen in der medizinischen Versorgung führt, wenn z. B. Medikamente nur mit Männern getestet wurden. Die Zugänge für Frauen in klassische Männerdomänen wie den IT-Sektor sind oft immer noch schwierig. The goal of the initiative is therefore a three-day workshop (Women AI Days) to connect national female experts and analyze needs, such as strengthening the proportion of women and making research and work areas visible to young talents. Through accompanying social media efforts, a publication, and subsequent public lectures at Frankfurt UAS, the content will be made known to the public, with a particular focus on the state of Hesse.
Prof. Dr.-Ing. Andreas Koch, TUDa, Department of Embedded Systems and Applications (ESA)
Prof. Dr. Carsten Binnig, TUDa, Data Management
Sum-Product-Networks (SPNs) gehören zur Klasse der graphischen probabilistischen Modelle und erlauben die kompakte Repräsentation von multivariaten Wahrscheinlichkeitsverteilungen. Während das Fachgebiet ESA vornehmlich die Beschleunigungsmöglichkeiten von SPNs untersucht hat, hat das FG DM sich damit befasst, für welche Anwendungsfälle im Bereich der Datenbanken SPNs genutzt werden können. Hierzu zählt z.B. die Cardinality Estimation. Sie kann verwendet werden, um die Ergebnisgrößen von Datenbank-Anfragen vorherzusagen und damit die Anfragenbearbeitung von Datenbank Management Systems (DBMS) zu optimieren. Das Gesamtziel des Projekts ist die generelle Beschleunigung der Cardinality Estimation mittels RSPNs (Relationale SPNs), die Automatisierung des Entwicklungs- und Trainingsprozesses der RSPNs sowie darüber hinaus die Untersuchung der potentiellen Nutzbarkeit im Kontext von großen Datenbanken. Die Erweiterung des SPNC, sowie die Bereitstellung entsprechender Trainingsprozesse versprechen in der Kombination hochinteressante, praktisch relevante Forschungsergebnisse, die auch in die anderen Projekten in den beiden beteiligten FGs eingehen können.
Dr. Markus Mühling PU Marburg, FB Mathematics & Computer Science
Prof. Dr. Nina Farbig, PU Marburg, FB Biology
Prof. Dr. Bernd Freisleben, PU Marburg, Dept. of Mathematics & Computer Science, Distributed Systems and Intelligent Computing
In diesem Projekt konzentrieren wir uns auf die automatische Erkennung von Vogelarten in Audioaufnahmen. Um die derzeitigen Überwachungsprogramme für die biologische Vielfalt zu verbessern, wird AI4Birds Tonaufnahmen in einem Waldökosystem verwenden, um neuartige Transformatormodelle zu entwickeln, die auf Selbstbeobachtung basieren, um Vogelarten in Klanglandschaften zu erkennen. Daher steht die Nachhaltigkeit im Hinblick auf die biologische Vielfalt im Mittelpunkt des Projekts. Die Nachhaltigkeit der Fortführung von AI4Birds durch die Einwerbung zusätzlicher finanzieller Mittel ist sehr wahrscheinlich; es ist geplant, AI4Birds zu nutzen, um die Finanzierungsmöglichkeiten im Rahmen der Bundesprogramme zur Förderung der biologischen Vielfalt auszuloten. Darüber hinaus planen wir, unsere Ergebnisse in Microsofts „AI for Earth“-Initiative einzubringen.
Prof. Dr. Martin Hirsch, Artificial Intelligence in Medicine, UMR and Director of the Institute for Artificial Intelligence at UKGM Marburg
Prof. Dr. Thomas Nauss, Department of Mathematics, Natural Sciences and Data Processing, FG Business Informatics – Artificial Intelligence, PU Marburg
Im Mai 2021 vereinbarten die Hessische Ministerin für Digitale Strategie und Entwicklung und der VDE die Einrichtung eines ersten bundesweiten „AI Quality & Testing Hub“ (AIQTH1). Im Umfeld von hessian.AI und dem Zentrum für Verantwortungsvolle Digitalisierung (ZEVEDI) soll damit die Qualität und Vertrauenswürdigkeit von KI-Systemen durch Standardisierung und Zertifizierung in den modellhaften Themenfeldern „Mobilität“, „Finanzen“ und „Gesundheit“ gefördert und für die Bevölkerung überprüfbar und glaubwürdig gemacht werden. Ziel des Projektes ist es, über das EU-Programm DIGITAL EUROPE den modellhaften Themenbereich „Gesundheit“ des AIQTH von hessian.AI zu stärken und damit die Chance zu nutzen, diese Einrichtung in Hessen zu etablieren.
Prof. Dr. Lambert Alff, TUDa, FB Material Sciences
Prof. Dr.-Ing. Andreas Koch, TUDa, Department of Embedded Systems and Applications (ESA)
Prof. Dr. Christian Hochburgen, TUDa, FB ETIT
Künstliche Intelligenz wird quasi ubiquitär in die verschiedensten Lebensbereiche Eingang finden. Dies bedeutet aber auch gleichzeitig, dass der Aspekt der Energieeffizienz des damit verbundenen Rechenaufwands immer wichtiger werden wird. Daher ist die Entwicklung der für die KI eingesetzten Computerarchitekturen ein wichtiges Forschungsfeld. Ein wesentliches neues Bauteil für KI-angepasste Rechnerarchitekturen ist der sogenannte Memristor. Es gibt mehrere materialwissenschaftliche Ansätze, mit denen (hochgradig energieeffiziente) Memristoren realisiert werden können, aus denen aber unterschiedliches Bauteilverhalten resultiert bzw. realistische Einsatzszenarien noch gar nicht erforscht wurden. Dieses Projekt zielt darauf ab, die für Memristoren notwendige Kette vom Material über das Bauteil hin zur Schaltung für bestimmte Anwendungen der KI zusammenzubringen und gemeinsame Forschungsprojekte im Sinne dieses interdisziplinären und holistischen Ansatzes zu befördern.
Prof. Dr. Leopoldo Molina-Luna, TUDa, FB Material Sciences
Prof. Dr. Kristian Kersting, TUDa, FB Computer Sciences
Viele Entwickler von KI-Algorithmen verfügen nicht über das nötige Hintergrundwissen, um mit dem neuesten Stand der Forschung in einem naturwissenschaftlichen Bereich wie z. B. den Materialwissenschaften Schritt zu halten. Auf der anderen Seite verlässt sich die Materialwissenschaft bei der Bestimmung der Parameter für die Entwicklung von KI-Algorithmen in der Regel auf „educated guess“ und zahlen wenig bis gar keine Gebühren. Es besteht eine Wissenslücke zwischen der Informatik und den Materialwissenschaften, und es bedarf eines verstärkten Austauschs auf grundlegender Ebene. Das Projekt schafft eine Plattform für die Umsetzung und Konsolidierung eines umfassenden regelmäßigen Austauschs zwischen allen interessierten Parteien. Es stärkt die Vorbereitungsaktivitäten für einen IRTG-Antrag im Bereich der on operando TEM für Memristoren und ML-basierte Datenanalyseroutinen.
Prof. Dr. Bernhard Humm, h_da, FB Computer Sciences
Prof. Dr. Andrea Krajewski, h_da, FB Media
Zur human-centered AI gehört u.a. die angemessene Erklärung von Entscheidungen oder Empfehlungen, die das KI-System z.B. mittels Machine Learning ausspricht (Stichwort “Explainable AI”). User Experience (UX) befasst sich hingegen mit der Entwicklung von Produkten, insbesondere IT-Systemen, die ein bestmögliches Nutzungserlebnis bieten sollen. In diesem Vorhaben sollen für drei unterschiedliche prototypische KI-Systeme, die im Rahmen des BMBF-geförderten Projekts “Kompetenzzentrum für Arbeit und Künstliche Intelligenz (KompAKI)” entwickelt werden, innovative UX-Konzepte konzipiert, abgestimmt, implementiert und evaluiert werden. Eines der KI-Systeme beschäftigt sich mit der Bereitstellung von Machine Learning (ML) für breite Nutzerkreise mit und ohne Programmierkenntnisse. Zwei KI-Systeme sind für die betriebliche Nutzung in der verarbeitenden Industrie (Industrie 4.0) vorgesehen. Dieses Vorhaben ergänzt andere KI-Initiativen in idealer Weise und fördert die Vernetzung zwischen hessian.AI-Partnern und unterschiedlichen Disziplinen.
Dr. Florian Stock, TUDa, Department of Mechanical Engineering, Department of Automotive Engineering (FZD)
Prof. Dr. Andreas Koch, TUDa, Department of Embedded Systems and Applications (ESA)
Im Fokus der Forschungen zu autonomen Fahren stehen bisher ganz klar Autos, nur wenige Projekte haben dagegen andere Verkehrsmittel im Blick. Um dem abzuhelfen, fördert Hessian.AI mit dem Projekt SpeedTram innovative interdisziplinäre Forschung, die sich mit dem autonomen/assistierten Fahren von Straßenbahnen befasst. Das Fachgebiet für Fahrzeugtechnik (FZD) und das Fachgebiet für Eingebettete Systeme und Anwendungen (ESA) der TU Darmstadt untersuchen darin die beschleunigte Ausführung der für die Automatisierung in und von Assistenzsystemen für Straßenbahnen erforderlichen Algorithmen des maschinellen Lernens. Dabei werden reale Daten bearbeitet, die auf einem Versuchsträger des lokalen Nahverkehrsunternehmens HEAG während des Betriebs aufgezeichnet wurden. Die Auswertung dieses wachsenden, mittlerweile mehr als 140 TB umfassenden Datensatzes, war mit den bestehenden Verfahren nicht mehr sinnvoll möglich. Durch die Arbeiten in SpeedTram konnten die beiden zeitaufwändigsten Schritte der Datenanalyse, nämlich die Objekterkennung auf Basis neuronaler Netze sowie die Verarbeitung der LIDAR-Sensordaten, jeweils um die Faktoren drei und 24 beschleunigt werden. SpeedTram leistet einen wichtigen Beitrag, das Innovationspotential der automatisierten Straßenbahnführung zu heben und für künftige Anwendungen nutzbar zu machen.
Dr. Nicolas Frieß, PU Marburg, FB Geography, Environmental Informatics
Prof. Dr. Bernd Freisegen, PU Marburg, Dept. of Mathematics & Computer Science, Distributed Systems and Intelligent Computing
Prof. Dr. Thomas Nauss, PU Marburg, FB Geography, Environmental Informatics
Biodiversität ist wichtig für verschiedene Ökosystemleistungen, die die Grundlage des menschlichen Lebens bilden. Der derzeitige Rückgang der biologischen Vielfalt erfordert eine Transformation von der manuellen periodischen Bewertung der biologischen Vielfalt hin zu einer automatisierten Echtzeit-Überwachung. Fledermäuse sind eine der am weitesten verbreiteten terrestrischen Säugetierarten und dienen als wichtige Bioindikatoren für den Zustand von Ökosystemen. Üblicherweise werden Fledermäuse durch Aufzeichnung und Analyse ihrer Echoortungsrufe überwacht. In diesem Projekt, AI4Bats, präsentieren wir einen neuartigen KI-basierten Ansatz zur Detektion von Fledermaus-Echoortungsrufen und zur Erkennung von Fledermausarten sowie zur Erkennung von Fledermausverhalten in Audiospektrogrammen. Er basiert auf einer neuronalen Transformer-Architektur und beruht auf Selbstaufmerksamkeitsmechanismen. Unsere Experimente zeigen, dass unser Ansatz aktuelle Ansätze zur Detektion von Fledermaus-Echoortungsrufen und zur Erkennung von Fledermausarten in mehreren öffentlich verfügbaren Datensätzen übertrifft. Während unser Modell zur Detektion von Fledermaus-Echoortungsrufen eine durchschnittliche Präzision von bis zu 90,2% erreicht, erzielt unser Modell zur Erkennung von Fledermausarten eine Genauigkeit von bis zu 88,7% für 14 in Deutschland vorkommende Fledermausarten, von denen einige selbst für menschliche Experten schwer zu unterscheiden sind. AI4Bats legt den Grundstein für Durchbrüche bei der automatisierten Überwachung von Fledermäusen im Bereich der Biodiversität, deren potenzieller Verlust wahrscheinlich zu den bedeutendsten Herausforderungen gehört, die die Menschheit in naher Zukunft bewältigen muss.
Dr. Joachim Bille, TH Mittelhessen, Head of Department FTN
Prof. Dr. Michael Guckert, TH Mittelhessen, Department of Mathematics, Natural Sciences and Data Processing, FG Business Informatics – Artificial Intelligence
Prof. Holger Rohn, TH Mittelhessen, Department of Industrial Engineering and Management, FG Life Cycle Management & Quality Management, Makerspace Friedberg
Ziel des Projekts AI@School war die Entwicklung eines Demonstrators zur anschaulichen Vermittlung von Grundkenntnissen der Künstlichen Intelligenz, der Schülerinnen und Schülern einen möglichst frühen und niedrigschwelligen Zugang zu KI-Themen ermöglichen sollte. Der Demonstrator soll zum einen geeignete Beispiele und Exponate zur anschaulichen Wissensvermittlung beinhalten, zum anderen soll ein interaktiver Einführungskurs zur Wissensvermittlung unter Anwendung der Exponate und Beispiele erarbeitet werden. Aufbauend auf diesen Angeboten soll ebenso eine prototypische Lehreinheit auf Leistungskursniveau entwickelt werden. Die Projektergebnisse sollen dauerhaft Implementierung bei hessian.AI implementiert werden; zudem ist mittel- bis langfristig ein hessenweiter Transfer des Konzepts an passende Institutionen in den anderen Landesteilen geplant.
Prof. Dr. Georgia Chalvatzaki, TUDa, FB Informatik, iROSA: Robot Learning of Mobile Manipulation for Intelligent Assistance
Prof. Dr. Iryna Gurevych, TUDa, FB Computer Science, Ubiquitous Knowledge Processing Lab
Ziel dieses Projekts war es, die Verbindungen zwischen hochrangigen Befehlen in natürlicher Sprache und Robotermanipulation zu untersuchen. Menschen sind in der Lage, Befehle in natürlicher Sprache effektiv zu abstrahieren und zu zerlegen, z. B. „Mach mir einen Kaffee“, aber eine solche Aktion ist für einen Roboter nicht detailliert genug, um sie auszuführen. Das Problem der Aufgabenausführung in der Robotik wird in der Regel als Aufgaben- und Bewegungsplanungsproblem angegangen, bei dem ein Aufgabenplaner das abstrakte Ziel in eine Reihe von logischen Aktionen zerlegt, die von einem Bewegungsgenerator in tatsächliche Aktionen in der Welt umgesetzt werden müssen. Die Verbindung zwischen abstrakter logischer Handlung und realer Beschreibung (z.B. hinsichtlich der genauen Position von Objekten in der Szene) macht die Aufgaben- und Bewegungsplanung zu einem sehr anspruchsvollen Problem. In diesem Projekt haben wir uns diesem Problem aus drei verschiedenen Richtungen genähert, indem wir Teilprobleme des Themas im Hinblick auf unser letztendliches Ziel, das Erlernen von Manipulationsplänen mit langem Zeithorizont unter Verwendung des gesunden Menschenverstands und von Szenegraphen, untersucht haben:
Projektergebnisse aus 2. und 3. sollen in naher Zukunft auf einer großen Konferenz zum maschinellen Lernen veröffentlicht werden. Arbeiten aus iii werden im Rahmen einer aktuellen Zusammenarbeit von iROSA und UKP fortgeführt.
