Universität Stuttgart
Permanent URI for this communityhttps://elib.uni-stuttgart.de/handle/11682/1
Browse
6 results
Search Results
Item Open Access IDEA - towards an interactive tool that supports creativity sessions in automotive product development(2023) Kaschub, Verena Lisa; Wechner, Reto; Krautmacher, Lara; Diers, Daniel; Bues, Matthias; Lossack, Ralf; Kloos, Uwe; Riedel, OliverItem Open Access Decoding mental effort in a quasi-realistic scenario : a feasibility study on multimodal data fusion and classification(2023) Gado, Sabrina; Lingelbach, Katharina; Wirzberger, Maria; Vukelić, MathiasHumans’ performance varies due to the mental resources that are available to successfully pursue a task. To monitor users’ current cognitive resources in naturalistic scenarios, it is essential to not only measure demands induced by the task itself but also consider situational and environmental influences. We conducted a multimodal study with 18 participants (nine female, M = 25.9 with SD = 3.8 years). In this study, we recorded respiratory, ocular, cardiac, and brain activity using functional near-infrared spectroscopy (fNIRS) while participants performed an adapted version of the warship commander task with concurrent emotional speech distraction. We tested the feasibility of decoding the experienced mental effort with a multimodal machine learning architecture. The architecture comprised feature engineering, model optimisation, and model selection to combine multimodal measurements in a cross-subject classification. Our approach reduces possible overfitting and reliably distinguishes two different levels of mental effort. These findings contribute to the prediction of different states of mental effort and pave the way toward generalised state monitoring across individuals in realistic applications.Item Open Access Quantum support vector machines of high-dimensional data for image classification problems(2023) Vikas Singh, RajputThis thesis presents a comprehensive investigation into the efficient utilization of Quantum Support Vector Machines (QSVMs) for image classification on high-dimensional data. The primary focus is on analyzing the standard MNIST dataset and the high-dimensional dataset provided by TRUMPF SE + Co. KG. To evaluate the performance of QSVMs against classical Support Vector Machines (SVMs) for high-dimensional data, a benchmarking framework is proposed. In the current Noisy Intermediate Scale Quantum (NISQ) era, classical preprocessing of the data is a crucial step to prepare the data for classification tasks using NISQ machines. Various dimensionality reduction techniques, such as principal component analysis (PCA), t-distributed stochastic neighbor embedding (tSNE), and convolutional autoencoders, are explored to preprocess the image datasets. Convolutional autoencoders are found to outperform other methods when calculating quantum kernels on a small dataset. Furthermore, the benchmarking framework systematically analyzes different quantum feature maps by varying hyperparameters, such as the number of qubits, the use of parameterized gates, the number of features encoded per qubit line, and the use of entanglement. Quantum feature maps demonstrate higher accuracy compared to classical feature maps for both TRUMPF and MNIST data. Among the feature maps, one using 𝑅𝑧 and 𝑅𝑦 gates with two features per qubit, without entanglement, achieves the highest accuracy. The study also reveals that increasing the number of qubits leads to improved accuracy for the real-world TRUMPF dataset. Additionally, the choice of the quantum kernel function significantly impacts classification results, with the projected type quantum kernel outperforming the fidelity type quantum kernel. Subsequently, the study examines the Kernel Target Alignment (KTA) optimization method to improve the pipeline. However, for the chosen feature map and dataset, KTA does not provide significant benefits. In summary, the results highlight the potential for achieving quantum advantage by optimizing all components of the quantum classifier framework. Selecting appropriate dimensionality reduction techniques, quantum feature maps, and quantum kernel methods is crucial for enhancing classification accuracy. Further research is needed to address challenges related to kernel optimization and fully leverage the capabilities of quantum computing in machine learning applications.Item Open Access Combining brain-computer interfaces with deep reinforcement learning for robot training : a feasibility study in a simulation environment(2023) Vukelić, Mathias; Bui, Michael; Vorreuther, Anna; Lingelbach, KatharinaDeep reinforcement learning (RL) is used as a strategy to teach robot agents how to autonomously learn complex tasks. While sparsity is a natural way to define a reward in realistic robot scenarios, it provides poor learning signals for the agent, thus making the design of good reward functions challenging. To overcome this challenge learning from human feedback through an implicit brain-computer interface (BCI) is used. We combined a BCI with deep RL for robot training in a 3-D physical realistic simulation environment. In a first study, we compared the feasibility of different electroencephalography (EEG) systems (wet- vs. dry-based electrodes) and its application for automatic classification of perceived errors during a robot task with different machine learning models. In a second study, we compared the performance of the BCI-based deep RL training to feedback explicitly given by participants. Our findings from the first study indicate the use of a high-quality dry-based EEG-system can provide a robust and fast method for automatically assessing robot behavior using a sophisticated convolutional neural network machine learning model. The results of our second study prove that the implicit BCI-based deep RL version in combination with the dry EEG-system can significantly accelerate the learning process in a realistic 3-D robot simulation environment. Performance of the BCI-based trained deep RL model was even comparable to that achieved by the approach with explicit human feedback. Our findings emphasize the usage of BCI-based deep RL methods as a valid alternative in those human-robot applications where no access to cognitive demanding explicit human feedback is available.Item Open Access Automatisierte Generierung von Trainingsdaten für die Informationsextraktion aus deutschen Geschäftsdokumenten auf Basis von Sprachmodellen(2023) Burkhardt, JannikGenerative KI hat seit der Veröffentlichung von ChatGPT im Dezember 2022 enorme Popularität erlangt. Ihr Potenzial ist immens und schon heute wird diese neue Technik in viele Produkte und Anwendungen integriert. In dieser Arbeit wird untersucht, welchen Einfluss automatisiert annotierte Trainingsdaten und von ChatGPT generierte Trainingsdaten auf das Finetuning von Sprachmodellen haben, wenn nur wenige handannotierte Daten vorhanden sind. Die mit den Methoden verbundenen Vorteile und Hindernisse werden am Beispiel der Relation Extraction aus deutschen Geschäftsdokumenten in Erfahrung gebracht. Es wird gezeigt, dass die Daten von ChatGPT von Fehlern bereinigt werden müssen, diese Daten dann jedoch die Leistung des Sprachmodells signifikant verbessern gegenüber einem Sprachmodell, das nur auf wenigen handannotierten Daten basiert.Item Open Access Training robust and generalizable quantum models(2024) Berberich, Julian; Fink, Daniel; Pranjić, Daniel; Tutschku, Christian; Holm, Christian