Seminar ‘Formal Verification of Neural Networks’ – Topics

The following list of topics is provisional. Topics may still be added, removed, or altered.

Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networks

Strong mixed-integer programming formulations for trained neural networks

Towards Fast Computation of Certified Robustness for ReLU Networks

Verifiably Safe Exploration for End-to-End Reinforcement Learning

Safe Exploration in Continuous Action Spaces

Probabilistic Safety for Bayesian Neural Networks

Perfectly Parallel Fairness Certification of Neural Networks