Sydney M. Katz

Textbook covering algorithms for validating performance of safety-critical decision-making systems using optimization, probability, and formal methods; co-authored and used as course text.

Doctoral dissertation presenting methods for designing safer perception systems by analyzing perception in closed-loop with downstream decision systems.

Paper proposing efficient estimation methods (smoothing bandits) to determine safe perception-system performance characteristics given black-box closed-loop simulators, demonstrated on vision-based aircraft collision avoidance.

Dataset and simulator paper presenting AVOIDDS to enable robust ML research for vision-based detect-and-avoid in aviation contexts.

Work proposing the use of generative models to reduce input dimensionality for verification of image-based neural network controllers, applied to an autonomous aircraft taxi problem.