Berkant Turan
I am working on improving the interpretability of AI, focusing on the Merlin-Arthur Classifier. This intriguing model uses two feature selectors to generate meaningful saliency maps. My objectives are to fine-tune its architecture, explore its adversarial components, and evaluate its effectiveness with complex datasets.
📬 Contact
- office
- Room 3024 at ZIB
- turan (at) zib.de
- homepage
- b-turan.github.io
- languages
- German, English, and Turkish
🎓 Curriculum vitae
- since 2023
- Member of BMS
- since 2022
- Researcher at ZIB
- 2022 to 2022
- Research Assistant at ZIB
- May 2022
- M.Sc. in Scientific Computing at TUB
- Oct 2018
- B.Sc. in Engineering Science at TUB
đź“ť Publications and preprints
Preprints
- GĹ‚uch, G., Turan, B., Nagarajan, S. G., and Pokutta, S. (2024). The Good, the Bad and the Ugly: Watermarks, Transferable Attacks and Adversarial Defenses.
[arXiv]
[BibTeX]
Conference proceedings
- Wäldchen, S., Sharma, K., Turan, B., Zimmer, M., and Pokutta, S. (2024). Interpretability Guarantees with Merlin-Arthur Classifiers. Proceedings of the International Conference on Artificial Intelligence and Statistics.
[arXiv]
[BibTeX]
🔬 Projects
Preserving global vegetation is crucial for addressing and mitigating climate change. Accurate, up-to-date forest health data is essential. AI4Forest aims to develop advanced AI methods to monitor forests using satellite imagery, including radar and optical data. The project will create scalable techniques for detailed, high-resolution maps of the globe, e.g., to monitor canopy height, biomass, and to track forest disturbances.
Existing approaches for interpreting Neural Network classifiers that highlight features relevant for a decision are based solely on heuristics. We introduce a theory that allows us to bound the quality of the features without assumptions on the classifier model by relating classification to Interactive Proof Systems.
đź’¬ Talks and posters
Poster presentations
- Jul 2024
- Unified Taxonomy in AI Safety: Watermarks, Adversarial Defenses, and Transferable Attacks
Workshop on Theoretical Foundations of Foundation Models (TF2M) @ ICML 2024, Vienna - May 2024
- Interpretability Guarantees with Merlin-Arthur Classifiers
27th AISTATS Conference, València - Jul 2023
- Extending Merlin-Arthur Classifiers for Improved Interpretability
The 1st World Conference on eXplainable Artificial Intelligence
đź“… Event Attendance
- May 2025
- 7th DOxML Conference, Kyoto
👨‍🏫 Teaching
- winter 2019
- Tutor for Statics and Elementary Strength of Materials at TUB
- summer 2018
- Tutor for Energy-Based Methods in Mechanics at TUB
- summer 2018
- Tutor for Continuum Mechanics at TUB
- winter 2018
- Tutor for Kinematics and Dynamics at TUB
- summer 2017
- Tutor for Statics and Elementary Strength of Materials at TUB
- winter 2017
- Tutor for Energy-Based Methods in Mechanics at TUB
- winter 2017
- Tutor for Continuum Mechanics at TUB
- summer 2016
- Tutor for Kinematics and Dynamics at TUB
- summer 2015
- Tutor for Statics and Elementary Strength of Materials at TUB