We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

math

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Mathematics > Optimization and Control

Title: Optimization Condition and Algorithm of Optimization with Convertible Nonconvex Function

Abstract: The paper introduces several new concepts for solving nonconvex or nonsmooth optimization problems, including convertible nonconvex function, exact convertible nonconvex function and differentiable convertible nonconvex function. It is proved herein many nonconvex functions or nonsmooth (or discontinuous) functions are actually convertible nonconvex functions and convertible nonconvex function operations such as addition, subtraction, multiplication or division result in convertible nonconvex functions. The sufficient condition for judging a global optimal solution to unconstrained optimization problems with differentiable convertible nonconvex functions is proved, which is equivalent to Karush-Kuhn-Tucker(KKT) condition. Two Lagrange functions of differentiable convertible nonconvex function are defined with their dual problems defined accordingly. The strong duality theorem is proved, showing that the optimal objective value of the global optimal solution is equal to the optimal objective value of the dual problem, which is equivalent to KKT condition. An augmented Lagrangian penalty function algorithm is proposed and its convergence is proved. So the paper provides a new idea for solving unconstrained nonconvex or non-smooth optimization problems and avoids subdifferentiation or smoothing techniques by using some gradient search algorithms, such as gradient descent algorithm, Newton algorithm and so on.
Subjects: Optimization and Control (math.OC)
MSC classes: 90C06, 90C25, 90C26, 90C59
Cite as: arXiv:2201.04264 [math.OC]
  (or arXiv:2201.04264v1 [math.OC] for this version)

Submission history

From: Zhiqing Meng [view email]
[v1] Wed, 12 Jan 2022 01:37:08 GMT (37kb)

Link back to: arXiv, form interface, contact.