Hoang-Dung Tran
Papers
7
Total Citations
452
H-Index
6
About
Hoang-Dung Tran is a leading researcher in the safety verification and formal analysis of neural network-controlled systems, a critical area for deploying AI in safety-critical applications like autonomous driving and cyber-physical systems. His most impactful work tackles the fundamental challenge of ensuring that neural networks behave reliably under all possible inputs. Tran introduced the concept of "maximum sensitivity" for multilayer perceptrons, enabling rigorous output reachable set estimation and safety verification—a contribution that has garnered over 270 citations. He further advanced the field with simulation-guided reachability algorithms that scale to neural network control systems, achieving 80 citations for his 2020 paper. His parallelizable reachability analysis algorithms for feed-forward networks (49 citations) offer practical, efficient solutions for real-time verification. Beyond neural networks, Tran has contributed to model-order reduction for large-scale linear systems and hybrid systems verification tools like C2E2 and HyST. His work bridges the gap between theoretical guarantees and practical deployment, making him a key figure in trustworthy AI and formal methods for control systems.
Research Focus
Key Achievements
Top Papers
- 1Output Reachable Set Estimation and Verification for Multilayer Neural Networks270 citations · 2018
- 2
- 3
- 4
- 5Large-Scale Linear Systems from Order-Reduction9 citations · 2018
- 6
- 7