Parabolic Heuristics and Anisotropic Scaling for Trust-Region Updates in Sequential Linear Programming

Vol 57, 2025 - 340805
Extended Abstracts (EA)
Favorite this paper
How to cite this paper?
Abstract

This paper evaluates a parabolic heuristic for trust-region radius updates in two Sequential Linear Programming architectures for nonlinearly constrained optimization: a unified Ared/Pred method (UNIF) and an Inexact Restoration method (Two-Phase). The heuristic estimates a step-length multiplier through one-dimensional quadratic interpolation and is tested with scalar and componentwise radii. Experiments on small CUTEst instances, restricted to at most 10 variables and at most one constraint, show that the Two-Phase family is consistently more robust than UNIF. The isotropic parabolic variant improves strict KKT coverage and median iterations in both families, whereas anisotropic scaling generally increases cost and is less reliable. Direct comparisons favor the best Two-Phase configuration in iterations, objective value, and execution time.

Share your ideas or questions with the authors!

Did you know that the greatest stimulus in scientific and cultural development is curiosity? Leave your questions or suggestions to the author!

Sign in to interact

Have a question or suggestion? Share your feedback with the authors!

Institutions
  • 1 Instituto Tecnológico de Aeronáutica; Universidade Federal de São Paulo
  • 2 Universidade Federal de São Paulo
Track
  • ONL-Non-linear Optimization
Keywords
Sequential Linear Programming
Trust Region Methods
Inexact Restoration
Nonlinear Optimization