A lower bound for the controlling parameters of the exact penalty functions

The purpose of this paper is to present new exact penalty functions and discuss their properties. A lower bound on the controlling parameters is given, for which above this value, the optimum of the exact penalty function coincides with the optimum of the nonlinear programming problem.This work was supported by the National Research Council of Canada under Grant A4414.     

Stationarity and superlinear convergence of an algorithm for univariate locally lipschitz constrained minimization

This paper introduces an algorithm for minimizing a single-variable locally Lipschitz function subject to a like function being nonpositive. The method combines polyhedral and quadratic approximation, a new type of penalty technique and a safeguard in such a way as to give convergence to a stationary point. The convergence is shown to be superlinear under somewhat stronger assumptions that allow both nonsmooth and nonconvex cases. The algorithm can be an effective subroutine for solving line search subproblems called for by multivariable optimization algorithms. Research sponsored, in part, by the Air Force Office of Scientific Research, Air Force Systems Command, USAF, under Grant Number AFOSR-83-0210. The U.S. Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright notation thereon.     

Progressive global random search of continuous functions

A sequential random search method for the global minimization of a continuous function is proposed. The algorithm gradually concentrates the random search effort on areas neighboring the global minima. A modification is included for the case that the function cannot be exactly evaluated. The global convergence and the asymptotical optimality of the sequential sampling procedure are proved for both the stochastic and deterministic optimization problem.The research is sponsored in part by the Air Force under Grant AFOSR-72-2371.     

Modeling NMR chemical shifts: a comparison of charge models for solid state effects on N chemical shift tensors

This paper presents results from applying different point charge models to take into account intermolecular interactions to model the solid state effects on the ¹⁵N NMR chemical shifts tensors. The DFT approach with the BLYP gradient corrected exchange correlation functional has been used because it can include electron correlation effects at a reasonable cost and is able to reproduce ¹⁵N NMR chemical shifts with reasonable accuracy. The results obtained with the point charge models are compared with the experimental data and with results obtained using the cluster model, which includes explicitly neighboring molecular fragments. The results show that the point charge models can take into account solid state effects at a cost much lower than the cluster methods.     

Linear programming via a quadratic penalty function

We use quadratic penalty functions along with some recent ideas from linearl ₁ estimation to arrive at a new characterization of primal optimal solutions in linear programs. The algorithmic implications of this analysis are studied, and a new, finite penalty algorithm for linear programming is designed. Preliminary computational results are presented.Research supported by grant No. 11-0505 from the Danish Natural Science Research Council SNF.     

Ratios of affine functions

Aratio of affine functions is a function which can be expressed as the ratio of a vector valued affine function and a scalar affine functional. The purpose of this note is to examine properties of sets which are preserved under images and inverse images of such functions. Specifically, we show that images and inverse images of convex sets under such functions are convex sets. Also, images of bounded, convex polytopes under such functions are bounded, convex polytopes. In addition, we provide sufficient conditions under which the extreme points of images of convex sets are images of extreme points of the underlying domains. Of course, this result is useful when one wishes to maximize a convex function over a corresponding set. The above assertions are well known for affine functions. Applications of the results include a problem that concerns the control of stochastic eigenvectors of stochastic matrices.     

On the Minimization of Difference Functions

Simple necessary optimality conditions are formulated for a function f of the form f _ g–h, where g and h are nonsmooth functions. Related sufficient conditions are given for local minimization and global minimization.     

Local properties of algorithms for minimizing nonsmooth composite functions

This paper considers local convergence and rate of convergence results for algorithms for minimizing the composite functionF(x)=f(x)+h(c(x)) wheref andc are smooth buth(c) may be nonsmooth. Local convergence at a second order rate is established for the generalized Gauss—Newton method whenh is convex and globally Lipschitz and the minimizer is strongly unique. Local convergence at a second order rate is established for a generalized Newton method when the minimizer satisfies nondegeneracy, strict complementarity and second order sufficiency conditions. Assuming the minimizer satisfies these conditions, necessary and sufficient conditions for a superlinear rate of convergence for curvature approximating methods are established. Necessary and sufficient conditions for a two-step superlinear rate of convergence are also established when only reduced curvature information is available. All these local convergence and rate of convergence results are directly applicable to nonlinearing programming problems.This work was done while the author was a Research fellow at the Mathematical Sciences Research Centre, Australian National University.     

Testing optimality for quadratic 0–1 unconstrained problems

This paper analyses a necessary and sufficient optimality condition for quadratic pseudo-Boolean unconstrained problems. It is proved that in general testing any necessary and sufficient optimality condition is a difficult task for anyNP-hard problem. An-optimality condition is derived together with an approximation scheme to test it.This work has been supported by the Progetto Finalizzato Trasporti 2, 93.01799.PF74.Professor Paolo Carraresi died unexpectedly on March 5, 1994. At the time of his death this paper had been completed. While undertaking the final revision, the other two authors were reminded just how much they were indebted to Professor Carraresi after many years of common work together.