Discover how projected Hessian updates drive efficient, reliable optimization in constrained problems.
This book explains sequential quadratic programming (SQP) methods that use an approximation to second-derivative information projected onto the constraint tangent space. It focuses on maintaining a positive definite representation in the reduced space, which helps stabilize updates and enable standard quasi-Newton techniques. The text covers how active and working sets are identified and how the projected Hessian guides subproblem formulations and iterations.
- How the projected Hessian enters optimality conditions and how to form and update it in practice
- Ways to manage the working set, including dropping criteria and their impact on convergence
- Update rules and safeguards that keep the projected Hessian well-behaved and numerically stable
- Convergence results for a restricted class of problems and practical insights from numerical results
Ideal for readers who want a rigorous, implementation-conscious view of SQP methods and the role of projected Hessian information in constrained optimization.