30167 Hannover
Research Interests
My research interests revolve around theoretical development and practical application of statistical and Machine Learning methods, particularly in the context of Green AutoML. It is a field dedicated to developing environmentally sustainable and energy-efficient automated machine learning algorithms. I aim to leverage these techniques within industrial contexts to drive innovation across a variety of applications.
Curriculum Vitae
-
Working Experience
2023 - Present
Doctoral Researcher, Leibniz University Hannover2022 - 2023
Doctoral Researcher, Volkswagen AG, Wolfsburg.2022
Analytics Professional, Deloitte Consulding LLC, Dusseldorf.2021 - 2022
Master's degree candidate, Internship and Thesis, IAV GmbH, Gifhorn. -
Education
2023 - Present
Ph.D. Student at the Institute of Artificial Intelligence, Leibniz University Hannover2019 - 2022
M.Sc. , Financial Mathematics, Technische Universität Braunschweig. Thesis: "Novelty Detection via Kernel Mean Embeddings".2016 - 2019
B.Sc. , Financial Mathematics, Bielefeld University. Thesis: "Prediction of Customer Churn Using Machine Learning Algorithms".
Guidelines for the Quality Assessment of Energy-Aware NAS Benchmarks
Abstract
Neural Architecture Search (NAS) accelerates progress in deep learning through systematic refinement of model architectures. The downside is increasingly large energy consumption during the search process. Surrogate-based benchmarking mitigates the cost of full training by querying a pre-trained surrogate to obtain an estimate for the quality of the model. Specifically, energy-aware benchmarking aims to make it possible for NAS to favourably trade off model energy consumption against accuracy. Towards this end, we propose three design principles for such energy-aware benchmarks: (i) reliable power measurements, (ii) a wide range of GPU usage, and (iii) holistic cost reporting. We analyse EA-HAS-Bench based on these principles and find that the choice of GPU measurement API has a large impact on the quality of results. Using the Nvidia System Management Interface (SMI) on top of its underlying library influences the sampling rate during the initial data collection, returning faulty low-power estimations. This results in poor correlation with accurate measurements obtained from an external power meter. With this study, we bring to attention several key considerations when performing energy-aware surrogate-based benchmarking and derive first guidelines that can help design novel benchmarks. We show a narrow usage range of the four GPUs attached to our device, ranging from 146 W to 305 W in a single-GPU setting, and narrowing down even further when using all four GPUs. To improve holistic energy reporting, we propose calibration experiments over assumptions made in popular tools, such as Code Carbon, thus achieving reductions in the maximum inaccuracy from 10.3 % to 8.9 % without and to 6.6 % with prior estimation of the expected load on the device.
Details
- Organisationseinheit(en)
-
Institut für Künstliche Intelligenz
Fachgebiet Maschinelles Lernen
- Externe Organisation(en)
-
Leiden University
Rheinisch-Westfälische Technische Hochschule Aachen (RWTH)
Technische Universität Darmstadt
- Typ
- Aufsatz in Konferenzband
- Seiten
- 50-59
- Anzahl der Seiten
- 10
- Publikationsdatum
- 2025
- Publikationsstatus
- Veröffentlicht
- Peer-reviewed
- Ja
- ASJC Scopus Sachgebiete
- Information systems, Hardware und Architektur, Informationssysteme und -management, Sicherheit, Risiko, Zuverlässigkeit und Qualität, Computernetzwerke und -kommunikation, Artificial intelligence
- Ziele für nachhaltige Entwicklung
- SDG 7 - Erschwingliche und saubere Energie
- Elektronische Version(en)
-
https://doi.org/10.1109/CCGridW65158.2025.00017 (Zugang:
Geschlossen
)
https://doi.org/10.48550/arXiv.2505.15631 (Zugang: Offen )
