This seminar presents stability analysis as a foundational framework toward safe physical AI. We examine nonlinear system stability through basins of attraction and Transient Stability Analysis (TSA), addressing how systems respond to disturbances and remain within safe operating regions.
The seminar connects dynamical systems, nonlinear optimization, and Morse theory to interpret optimization convergence through the stability structure of multi-basin landscapes. It discusses how stability-informed analysis improves the robustness and stability of learning-based models, while also addressing safe and privacy-preserving AI through adversarial defenses, differential privacy, and federated learning.
By integrating dynamical systems theory, optimization, and topology, this seminar highlights how stability-aware analysis can guide the design of robust and trustworthy AI systems for real-world physical interaction.
