Curriculum Vitae

Lukas Koller joined the Cyber-Physical-Systems Group as a Ph.D. candidate under the supervision of Prof. Dr.-Ing. Matthias Althoff in late 2023. Lukas has received both, his master's degree and his bachelor's degree in Informatics from the Technical University of Munich in 2023 and 2021, respectively.

His research interests include formal verification of neural networks and training of robust neural networks. He is a member of the DFG-funded project "Data-driven process modelling in metal forming technology".

Offered Thesis Topics

I am always looking for motivated students eager to explore the training and verification of neural networks. If you are interested in pursuing a thesis in this area, please reach out via email with your CV, transcript of records, and a brief statement of your research interests. You can find specific thesis proposals and practical course topics listed below for inspiration.

Ongoing:

• [MT] Set-Based Sufficient Subset Training for Robust and Interpretable Neural Networks

Finished:

• [BT] Robust Reinforcement Learning using Set-Based Training
• [BT] Evaluating the Robustness of Neural Networks with Adversarial Attacks
• [BT] Exploiting Monotonicity for Training and Verification of Neural Networks
• [MT] Set-Based Learning of Neural Barrier Certificates for Safety Verification of Dynamic Systems
• [BT] Branching strategies for Zonotope-based Neural Network Verification
• [BT] Structured Composition of Monotonic Neural Networks for Scalable Verification

Teaching

Practical Course – Formal Methods for AI-Enabled Cyber-Physical Systems (IN2106, IN4269)

• Implementation of algorithms for formal neural network verification in CORA 
  [WiSe 23/24: Topics] [SoSe 24: Topics] [WiSe 24/25] [SoSe 25] [WiSe 25/26: Topics]

Seminar – Cyber-Physical Systems (IN0014, IN2107, IN4813)

• Literature research on training robust neural networks [SoSe 24: Topics]
• Literature research on conformance checking for neural networks [WiSe 24/25: Topics]
• Literature research on formal verification of neural networks with linear bounds [SoSe 25: Topics]
• Literature research on recent advances in neural network verification [WiSe 25/26: Topics]

Lecture – Formal Methods for Cyber-Physical Systems (IN2383) [WiSe 23/24] [WiSe 24/25] [WiSe 25/26]