Global Minimization of Normal Quartic Polynomials Based on Global Descent Directions

Abstract

Abstract.A normal quartic polynomial is a quartic polynomial whose fourth degree term coefficient tensor is positive definite. Its minimization problem is one of the simplest cases of nonconvex global optimization, and has engineering applications. We call a direction a global descent direction of a function at a point if there is another point with a lower function value along this direction. For a normal quartic polynomial, we present a criterion to find a global descent direction at a noncritical point, a saddle point, or a local maximizer. We give sufficient conditions to judge whether a local minimizer is global and give a method for finding a global descent direction at a local, but not global, minimizer. We also give a formula at a critical point and a method at a noncritical point to find a one-dimensional global minimizer along a global descent direction. Based upon these, we propose a global descent algorithm for finding a global minimizer of a normal quartic polynomial when n = 2. For the case n ≥ 3, we propose an algorithm for finding an -global minimizer. At each iteration of a second algorithm, a system of constrained nonlinear equations is solved. Numerical tests show that these two algorithms are promising.

Numerical Solution of Bivariate and Polyanalytic Polynomial Systems

“…More specifically, each polynomial's coefficients are converted to hardware double precision floating point numbers, after which they are multiplied by an integer power of two such that their exponents are approximately centered around zero. [40].…”

Polynomial shape from shading

“…and the gradient is simply ∇F = 2J T r. Minimization of a general multivariate quartic is NPhard [22,28]. However, a useful property of polynomials is that exact line search takes linear time.…”

Solving inverse problems of optical microlithography

“…The objective function is very smooth in this case, which ease application of the gradient-descent methods. While theory of QP is well developed, the polynomial optimization is an area of growing research interest, in particular for quartic polynomials [34].…”