Pouya Samanipour successfully defended his Ph.D. Dissertation titled “Analysis of Closed-loop Dynamical Systems Using ReLU NNs”.

This dissertation focuses on computational methods to estimate the largest possible region of attraction of the equilibria of nonlinear systems. To obtain this estimate it proposes algorithms for finding Lyapunov functions and barrier functions.

Many closed-loop dynamical systems can be (approximately) modeled using neural network functions. Pouya’s dissertation focuses on neural networks with rectified linear unit (ReLU) activations, which make them piecewise affine functions. This choice enables rigorous and exact methods to be applied, instead of relying on sampling-based techniques and expensive verification.