Lukas Gosch

Lukas Gosch is a PhD student at the Technical University of Munich, affiliated with the TUM School of Computation, Information and Technology and the Department of Computer Science. He is part of the DAML research group under the supervision of Prof. Stephan Günnemann and the relAI graduate school.

• Research Focus: Robustness in machine learning, graph neural networks (GNNs), combinatorial optimization, adversarial verification, and efficient ML.
• Education: M.Sc. in Computational Science (2018-2021, University of Vienna), B.Sc. in Physics (2013-2017, Vienna University of Technology).

His work investigates how to certify and improve the robustness of neural networks, particularly against label/data poisoning and backdoor attacks. He leverages techniques like neural tangent kernels and mixed-integer programming to derive theoretical guarantees on model behavior. Recent papers analyze robustness plateaus and semantic-aware adversarial examples.

Recent Scientific Recognition:

• Best Paper Award @ NeurIPS 2024 AdvML Frontiers Workshop
• Selected Oral Talk @ NeurIPS TSRML 2022
• Best Master's Thesis Award @ Austrian Society for Operations Research 2021
• Performance Scholarship @ University of Vienna 2020