Lecture
This lecture introduces the Riemannian Trust Regions (RTR) framework, focusing on the concept of conjugate directions, Newton's method, and the main idea behind RTR. The instructor explains the retraction-based algorithm, the unique minimizer vector, and the conditions for model improvement. The lecture covers the trust-region radius update, the acceptance/rejection criteria for the next iterate, and the requirements for subproblem solving. It also discusses the regularity conditions and the boundedness of the model. Examples and parameters for the maximum radius and threshold are provided.