With the rapid advances in computation and communication, there is a growing interest in verifying the performance of cyber-physical systems (e.g., connected and autonomous vehicles) using real-time system data. Such a practice, however, can jeopardize privacy since these data may… Read More

Deep neural networks are expressive functional templates to approximate unknown dynamics thanks to their universal approximation property. However, their expressiveness can also cause overfitting, especially when training samples are lacking. Our recent ACC ’22 paper deals with the issue by… Read More