Current browse context:
math
Change to browse by:
References & Citations
Mathematics > Optimization and Control
Title: The q-Levenberg-Marquardt method for unconstrained nonlinear optimization
(Submitted on 5 Jul 2021)
Abstract: A q-Levenberg-Marquardt method is an iterative procedure that blends a q-steepest descent and q-Gauss-Newton methods. When the current solution is far from the correct one the algorithm acts as the q-steepest descent method. Otherwise the algorithm acts as the q-Gauss-Newton method. A damping parameter is used to interpolate between these two methods. The q-parameter is used to escape from local minima and to speed up the search process near the optimal solution.
Link back to: arXiv, form interface, contact.