Promotionen der Fakultät für Informatik im Jahr 2025

21.03.2025

Datsogiannis, Dimitrios
"Smart Detection of Deficiencies and Faults in Automotive Software Releases"
Promotion zum Dr.-Ing.

Abstract:

This study introduces a novel approach for evaluating the Electric and Electronic automotive software development processes and software releases across all V-model phases. The aim is to provide transparency among the stakeholders and deliver precise feedback over the development cycles. The ultimate purpose is to prevent software defects and reproducible conflicts.After a systematic literature review of the complete development, the proposed model enables the stakeholders to evaluate the status, performance, and quality of the software packages by posing targeted questions to the involved stakeholders and evaluating the answers to induct the conclusions. The questions are resulted from the literature as metrics and principles, before being transformed into questions. The questions are selectively proposed by an intelligent Reinforcement Learning agent using Contextual Multi-Armed Bandits (CMAB), which serve as a recommendation system. Each question has a different weight, pre-evaluated by experts for this study, and each answer has a different value. These two parameters define the reward of the agent, which is balancing between exploration-exploitation dilemma.The questions act as software packages containing all vital information about the evaluation, allowing a comprehensive assessment of the system under test.The model of this study is boundlessly scalable in terms of the complexity of the target software or component that can be evaluated, empowering the continuous improvement of its performance as the algorithm learns over time.The evaluation results confirm that the concept is effectively functional under any circumstances, addressing the main challenges of cold start, partial feedback, and data parsing.In summary, this thesis contributes to automotive software development by enhancing transparency and enabling the punctual detection of process deficiencies and software faults.

13.03.2025

Larisch, René
"The effects of inhibitory plasticity and the emerging network dynamics on processing visual information"
Promotion zum Dr. rer. nat.

Abstract:

In recent years, deep neural networks have shown how a brain-inspired approach can lead to outstanding performance in object recognition and detection. Despite their success in visual tasks, these artificial neural networks differ from the biological visual system in several ways:
They have a predominantly feed-forward connectivity structure instead of many lateral and recurrent connections.Understanding how information is processed in the brain can help to improve artificial visual systems and artificial neural networks in general.

There is a large corpus of neural networks in computational neuroscience that model different parts of the biological visual system to investigate the role of excitatory neurons and how different synaptic plasticity rules cause their representation of visual information, a core element for object recognition.
Despite the large number of studies on the visual system, one mechanism that has received little attention is the plasticity of inhibitory neurons and how it influences the emergence of neuronal representations.
Moreover, while these networks succeed in modeling different isolated parts of the visual pathway, such as the retina or the primary visual cortex, only a few models implement multiple areas to investigate how visual information is represented along different areas.

In my dissertation defense, I present a spiking neural network to model the early visual system up to the primary visual cortex (V1).
By extending the network with different key functions of the visual system, I show the importance of inhibition for spatial and temporal representation of visual information.I found that the dynamics resulting from the interplay of excitatory and inhibitory synaptic plasticity are crucial for the development of the representation of different input features.
As this is relevant for object recognition, the usefulness of the input representation is supported by benchmarks of different standard object recognition datasets.
I show that the temporal encoding of visual information is rooted in the very early layers of the visual system, the retina, and the thalamus, and that the selectivity of motion detection in V1 is strongly influenced by temporal dynamics between these early stages and by tuned lateral inhibition within V1.

In summary, my thesis improves the understanding of plasticity dynamics and different circuit motifs in more brain-like networks and supports their potential as artificial visual systems.

16.01.2025

Noura, Mahda
"Goal-Oriented End User Development of Web of Things Compositions"
Promotion zum Dr.-Ing.

Abstract:

Internet of Things (IoT) devices have gained widespread adoption by end users due to their provided benefits. The recent emergence of the Web of Things (WoT) standard has further simplified access to IoT devices. Enabling end users, particularly those lacking technical expertise, to compose the behaviour of their WoT environments to meet their specific needs is desirable. However, this is complex due to the intrinsic characteristics of end users and WoT environments. End users are struggling to compose their WoT environments, primarily because of the required expertise, effort and time. This thesis conducts a systematic analysis of this situation, identifying three main problems that hinders end users from personalizing the behaviour of their smart environments. Insufficient support of interoperability, usability and usefulness are thethree main problems addressed in this thesis. A set of requirements is derived for a suitable Things composition development approach by end users and used to assess the state of the art in the WoT domain. Existing research in Things composition development exhibits shortcomings, particularly in providing dedicated approaches to enable end users to compose heterogeneous Things and facilitate useful functionalities in a usable manner.

This thesis proposes a solution to address the identified shortcomings and empower end users to compose their WoT environments, with a focus on interoperability, usability and usefulness. The proposed solution, called Goal-Oriented End User Development for Web of Things (GrOWTH) provides a holistic framework consisting constructive software engineering principles, formalisms, methods and tools for Things interoperability, end user interaction and Things composition. GrOWTH Things interoperability solutions address the problem of vendor- and technology dependent WoT systems by proposing a semi-automatic knowledge-based approach, allowing end users to personalize Things usage. The end user interaction solutions adopt a goal-oriented approach with a multimodal interface, facilitating more intuitive and natural interactions with WoT environments. Additionally, the Things composition solutions propose an AI planning-based approach for compositions development by lowering composition complexity and speeding up time-consuming activities for end users. The feasibility and applicability of these concepts are demonstrated through various evaluation experiments and empirical user studies. The thesis concludes with an evaluation based on requirements assessment and technical evaluation of the approach with the application scenario, accompanied by an outlook towards future research directions.