Birgit Hofer
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Birgit Hofer has a total of 32 co-authors. Here are the top co-authors by number of shared publications:
- Franz Wotawa
Graz University of Technology
26 shared publications - Dietmar Jannach
University of Klagenfurt, Klagenfurt, Austria
7 shared publications Franz Wotawa
7 shared publications- Rui Abreu
Meta Platforms, Inc., Menlo Park, USA
5 shared publications - Konstantin Schekotihin
Department of Artificial Intelligence and Cybersecurity, University of Klagenfurt, Klagenfurt, Austria
4 shared publications - Patrick Koch
Applied Informatics, Alpen-Adria Universitat Klagenfurt Fakultat fur Technische Wissenschaften, 392327 Klagenfurt, Carinthia Austria
4 shared publications
Latest Publications
- Open Access
- 20 May 2026
- Published by TIB Open Publishing in International Sustainable Energy Conference - Proceedings
- Vol. 2
Operational Optimization of PV-Assisted Heat Pump Systems Using Reinforcement Learning
Heat pumps are increasingly integrated with PV systems to supply domestic hot water and space heating. End user purchase prices for electricity are typically significantly higher than feed-in tariffs for PV energy. From a financial perspective, it therefore makes sense to increase the self-consumption of local PV electricity and reduce the amount of energy drawn from the grid. One approach is to overheat the thermal energy storage when the heat pump can be operated with PV energy. However, with increasing temperature, heat losses increase and the heat pump’s COP decreases, making excessive overheating uneconomical. The optimal degree varies with time and depends on various factors, such as heat demand, storage capacity and the purchase-to-feed-in price ratio. Using detailed system simulations, we inves-tigate the application of self-adaptive reinforcement learning agents that can dynamically determine the optimal degree of overheating. The agents continuously evaluate environmental and operational inputs to make informed control decisions, enhancing cost-effectiveness. The results of our empirical evaluation demonstrate that the agents reliably learn to avoid unnec-essary hot water overheating and adapt their space heating overheating strategy to the char-acteristics of each system, achieving net electricity costs close to the optimal solution. Across multiple configurations, agents operating in reduced environments—i.e., with fewer state vari-ables—converge more quickly and develop more stable control policies. In contrast, restrict-ing the action space alone does not consistently enhance performance. Although the overall cost reduction potential is modest, amounting to EUR 35–90 per year, the proposed approach has the important advantage of guaranteeing that user comfort is never violated.
- 20 May 2026
- Published by TIB Open Publishing in International Sustainable Energy Conference - Proceedings
- Vol. 2
Open AccessOperational Optimization of PV-Assisted Heat Pump Systems Using Reinforcement Learning
Heat pumps are increasingly integrated with PV systems to supply domestic hot water and space heating. End user purchase prices for electricity are typically significantly higher than feed-in tariffs for PV energy. From a financial perspective, it therefore makes sense to increase the self-consumption of local PV electricity and reduce the amount of energy drawn from the grid. One approach is to overheat the thermal energy storage when the heat pump can be operated with PV energy. However, with increasing temperature, heat losses increase and the heat pump’s COP decreases, making excessive overheating uneconomical. The optimal degree varies with time and depends on various factors, such as heat demand, storage capacity and the purchase-to-feed-in price ratio. Using detailed system simulations, we inves-tigate the application of self-adaptive reinforcement learning agents that can dynamically determine the optimal degree of overheating. The agents continuously evaluate environmental and operational inputs to make informed control decisions, enhancing cost-effectiveness. The results of our empirical evaluation demonstrate that the agents reliably learn to avoid unnec-essary hot water overheating and adapt their space heating overheating strategy to the char-acteristics of each system, achieving net electricity costs close to the optimal solution. Across multiple configurations, agents operating in reduced environments—i.e., with fewer state vari-ables—converge more quickly and develop more stable control policies. In contrast, restrict-ing the action space alone does not consistently enhance performance. Although the overall cost reduction potential is modest, amounting to EUR 35–90 per year, the proposed approach has the important advantage of guaranteeing that user comfort is never violated.
- Open Access
- 15 April 2026
- Published by Elsevier in Sustainable Cities and Society: Advances
- Vol. 2 (2)
Detecting soft faults in residential heat pumps via regression-based residual analysis
- 15 April 2026
- Published by Elsevier in Sustainable Cities and Society: Advances
- Vol. 2 (2)
Open AccessDetecting soft faults in residential heat pumps via regression-based residual analysis
- Open Access
- 21 August 2025
- Published by Elsevier in Artificial Intelligence
- Vol. 348
Choosing abstraction levels for model-based software debugging: A theoretical and empirical analysis for spreadsheet programs
- 21 August 2025
- Published by Elsevier in Artificial Intelligence
- Vol. 348
Open AccessChoosing abstraction levels for model-based software debugging: A theoretical and empirical analysis for spreadsheet programs
- Open Access
- 16 January 2025
- Published by Elsevier in Journal of Systems and Software
- Vol. 222
Best practices for evaluating IRFL approaches
- 16 January 2025
- Published by Elsevier in Journal of Systems and Software
- Vol. 222
Open AccessBest practices for evaluating IRFL approaches
- Open Access
- 20 March 2024
- Published by MDPI AG in Electronics
- Vol. 13 (6) , 1146
Reducing the Length of Dynamic and Relevant Slices by Pruning Boolean Expressions
Dynamic and relevant (backward) slicing helps programmers in the debugging process by reducing the number of statements in an execution trace. In this paper, we propose an approach called pruned slicing, which can further reduce the size of slices by reasoning over Boolean expressions. It adds only those parts of a Boolean expression that are responsible for the evaluation outcome of the Boolean expression to the set of relevant variables. We empirically evaluate our approach and compare it to dynamic and relevant slicing using three small benchmarks: the traffic collision avoidance system (TCAS), the Refactory dataset, and QuixBugs. Pruned slicing reduces the size of the TCAS slices on average by 10.2%, but it does not reduce the slice sizes of the Refactory and QuixBugs programs. The times required for computing pruned dynamic and relevant slices are comparable to the computation times of non-pruned dynamic and relevant slices. Thus, pruned slicing is an extension of dynamic and relevant slicing that can reduce the size of slices while having a negligible computational overhead.
- 20 March 2024
- Published by MDPI AG in Electronics
- Vol. 13 (6) , 1146
Open AccessReducing the Length of Dynamic and Relevant Slices by Pruning Boolean Expressions
Dynamic and relevant (backward) slicing helps programmers in the debugging process by reducing the number of statements in an execution trace. In this paper, we propose an approach called pruned slicing, which can further reduce the size of slices by reasoning over Boolean expressions. It adds only those parts of a Boolean expression that are responsible for the evaluation outcome of the Boolean expression to the set of relevant variables. We empirically evaluate our approach and compare it to dynamic and relevant slicing using three small benchmarks: the traffic collision avoidance system (TCAS), the Refactory dataset, and QuixBugs. Pruned slicing reduces the size of the TCAS slices on average by 10.2%, but it does not reduce the slice sizes of the Refactory and QuixBugs programs. The times required for computing pruned dynamic and relevant slices are comparable to the computation times of non-pruned dynamic and relevant slices. Thus, pruned slicing is an extension of dynamic and relevant slicing that can reduce the size of slices while having a negligible computational overhead.
- Open Access
- 16 March 2023
- Published by Elsevier in Journal of Systems and Software
- Vol. 201
Explaining software fault predictions to spreadsheet users
- 16 March 2023
- Published by Elsevier in Journal of Systems and Software
- Vol. 201
Open AccessExplaining software fault predictions to spreadsheet users
- Open Access
- 10 October 2022
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Boosting Spectrum-Based Fault Localization for Spreadsheets with Product Metrics in a Learning Approach
Faults in spreadsheets are not uncommon and they can have significant negative consequences in practice. Various approaches for fault localization were proposed in recent years, among them techniques that transferred ideas from spectrum-based fault localization (SFL) to the spreadsheet domain. Applying SFL to spreadsheets proved to be effective, but has certain limitations. Specifically, the constrained computational structures of spreadsheets may lead to large sets of cells that have the same assumed fault probability according to SFL and thus have to be inspected manually. In this work, we propose to combine SFL with a fault prediction method based on spreadsheet metrics in a machine learning (ML) approach. In particular, we train supervised ML models using two orthogonal types of features: (i) variables that are used to compute similarity coefficients in SFL and (ii) spreadsheet metrics that have shown to be good predictors for faulty formulas in previous work. Experiments with a widely-used corpus of faulty spreadsheets indicate that the combined model helps to significantly improve fault localization performance in terms of wasted effort and accuracy.
- 10 October 2022
- Published by Association for Computing Machinery (ACM)
- p. 1-5
Open AccessBoosting Spectrum-Based Fault Localization for Spreadsheets with Product Metrics in a Learning Approach
Faults in spreadsheets are not uncommon and they can have significant negative consequences in practice. Various approaches for fault localization were proposed in recent years, among them techniques that transferred ideas from spectrum-based fault localization (SFL) to the spreadsheet domain. Applying SFL to spreadsheets proved to be effective, but has certain limitations. Specifically, the constrained computational structures of spreadsheets may lead to large sets of cells that have the same assumed fault probability according to SFL and thus have to be inspected manually. In this work, we propose to combine SFL with a fault prediction method based on spreadsheet metrics in a machine learning (ML) approach. In particular, we train supervised ML models using two orthogonal types of features: (i) variables that are used to compute similarity coefficients in SFL and (ii) spreadsheet metrics that have shown to be good predictors for faulty formulas in previous work. Experiments with a widely-used corpus of faulty spreadsheets indicate that the combined model helps to significantly improve fault localization performance in terms of wasted effort and accuracy.
- Open Access
- 24 August 2022
- Published by Springer Nature in Automated Software Engineering
- Vol. 29 (2) , 1-29
Detecting non-natural language artifacts for de-noising bug reports
Textual documents produced in the software engineering process are a popular target for natural language processing (NLP) and information retrieval (IR) approaches. However, issue tickets often contain artifacts such as code snippets, log outputs and stack traces. These artifacts not only inflate the issue ticket sizes, but also can this noise constitute a real problem for some NLP approaches, and therefore has to be removed in the pre-processing of some approaches. In this paper, we present a machine learning based approach to classify textual content into natural language and non-natural language artifacts at line level. We show how data from GitHub issue trackers can be used for automated training set generation, and present a custom preprocessing approach for the task of artifact removal. The training sets are automatically created from Markdown annotated issue tickets and project documentation files. We use these generated training sets to train a Markdown agnostic model that is able to classify un-annotated content. We evaluate our approach on issue tickets from projects written in C++, Java, JavaScript, PHP, and Python. Our approach achieves ROC-AUC scores between 0.92 and 0.96 for language-specific models. A multi-language model trained on the issue tickets of all languages achieves ROC-AUC scores between 0.92 and 0.95. The provided models are intended to be used as noise reduction pre-processing steps for NLP and IR approaches working on issue tickets.
- 24 August 2022
- Published by Springer Nature in Automated Software Engineering
- Vol. 29 (2) , 1-29
Open AccessDetecting non-natural language artifacts for de-noising bug reports
Textual documents produced in the software engineering process are a popular target for natural language processing (NLP) and information retrieval (IR) approaches. However, issue tickets often contain artifacts such as code snippets, log outputs and stack traces. These artifacts not only inflate the issue ticket sizes, but also can this noise constitute a real problem for some NLP approaches, and therefore has to be removed in the pre-processing of some approaches. In this paper, we present a machine learning based approach to classify textual content into natural language and non-natural language artifacts at line level. We show how data from GitHub issue trackers can be used for automated training set generation, and present a custom preprocessing approach for the task of artifact removal. The training sets are automatically created from Markdown annotated issue tickets and project documentation files. We use these generated training sets to train a Markdown agnostic model that is able to classify un-annotated content. We evaluate our approach on issue tickets from projects written in C++, Java, JavaScript, PHP, and Python. Our approach achieves ROC-AUC scores between 0.92 and 0.96 for language-specific models. A multi-language model trained on the issue tickets of all languages achieves ROC-AUC scores between 0.92 and 0.95. The provided models are intended to be used as noise reduction pre-processing steps for NLP and IR approaches working on issue tickets.
- Open Access
- 30 June 2022
- Published by Elsevier in Journal of Systems and Software
- Vol. 192
A systematic literature review on benchmarks for evaluating debugging approaches
- 30 June 2022
- Published by Elsevier in Journal of Systems and Software
- Vol. 192
Open AccessA systematic literature review on benchmarks for evaluating debugging approaches