By Alexander M. Rubinov, Xiao-qi Yang

ISBN-10: 1402076274

ISBN-13: 9781402076275

ISBN-10: 1441991727

ISBN-13: 9781441991720

ISBN-10: 1461348218

ISBN-13: 9781461348214

Lagrange and penalty functionality tools offer a robust technique, either as a theoretical device and a computational automobile, for the learn of limited optimization difficulties. even though, for a nonconvex limited optimization challenge, the classical Lagrange primal-dual strategy might fail to discover a mini­ mum as a 0 duality hole isn't consistently assured. a wide penalty parameter is, usually, required for classical quadratic penalty capabilities so that minima of penalty difficulties are a superb approximation to these of the unique restricted optimization difficulties. it truly is recognized that penaity capabilities with too huge parameters reason a disadvantage for numerical implementation. therefore the query arises find out how to generalize classical Lagrange and penalty services, with the intention to receive a suitable scheme for decreasing limited optimiza­ tion difficulties to unconstrained ones that would be compatible for sufficiently large periods of optimization difficulties from either the theoretical and computational viewpoints. a few methods for this kind of scheme are studied during this booklet. considered one of them is as follows: an unconstrained challenge is developed, the place the target functionality is a convolution of the target and constraint features of the unique challenge. whereas a linear convolution ends up in a classical Lagrange functionality, other kinds of nonlinear convolutions result in fascinating generalizations. we will name services that seem as a convolution of the target functionality and the constraint features, Lagrange-type functions.

Show description

Read Online or Download Lagrange-type Functions in Constrained Non-Convex Optimization PDF

Similar linear programming books

Download e-book for iPad: Linear and Integer Programming vs Linear Integration and by Jean-Bernard Lasserre

During this e-book the writer analyzes and compares 4 heavily comparable difficulties, particularly linear programming, integer programming, linear integration, linear summation (or counting). the point of interest is on duality and the process is very novel because it places integer programming in standpoint with 3 linked difficulties, and allows one to outline discrete analogues of famous non-stop duality techniques, and the reason in the back of them.

Read e-book online Advances in Statistical Control, Algebraic Systems Theory, PDF

This volume—dedicated to Michael ok. Sain at the social gathering of his 70th birthday—is a set of chapters masking contemporary advances in stochastic optimum keep watch over idea and algebraic structures idea. Written via specialists of their respective fields, the chapters are thematically equipped into 4 parts:* half I makes a speciality of statistical keep watch over concept, the place the fee functionality is considered as a random variable and function is formed via fee cumulants.

Read e-book online Lagrange-type Functions in Constrained Non-Convex PDF

Lagrange and penalty functionality tools offer a strong process, either as a theoretical instrument and a computational motor vehicle, for the research of limited optimization difficulties. notwithstanding, for a nonconvex limited optimization challenge, the classical Lagrange primal-dual strategy may well fail to discover a mini­ mum as a nil duality hole isn't continuously assured.

Download e-book for iPad: Stochastic Modelling and Control by M. H. A. Davis, R. B. Vinter (auth.)

This publication goals to supply a unified remedy of input/output modelling and of keep an eye on for discrete-time dynamical platforms topic to random disturbances. the implications offered are of broad applica­ bility up to speed engineering, operations study, econometric modelling and plenty of different parts. There are designated ways to mathematical modelling of actual structures: a right away research of the actual mechanisms that contain the method, or a 'black field' strategy according to research of input/output information.

Additional resources for Lagrange-type Functions in Constrained Non-Convex Optimization

Example text

Let us check that a= gu(y). If to the contrary a =f. 24) a< gu(y). First assume that y > b. 24). Hence a Then = gu(y). Consider now the casey = b. which is impossible. We need to check that (a,y) E bd* J, (a',y') > (a,y) ===? (a',y') ¢ U. It has already been proved that a = gu (y). If y' > y, then gu (y') < gu (y) = a~ a'. Since a' > gu(y'), it follows that (a',y') rf. U. If y' = y, then a'> a= gu(y). It means that (a', y') rf. U. Thus the result follows. 25 Let p : lR~ --+ lR+ be a continuous strictly increasing IPHfunction such that p(x) > O,for all x =f.

Thus hp(Y) > 0. 16 that p( x) > 0 for all x Let p : IR~ ---+ IR+ be a continuous IPH function such # 0. Then there exists a number b :::=: 0 such that supp (p,L) = y::::; b,O b, 0 0, for ally > 0. If y E domhp = {y > 0: hp(y) < +oo}, then (hp(y),y) E supp (p,L). Hence, ifdomhp = (O,+oo),thensupp(p,L) = {(a,y): y > O,a 'S hp(y)}. 11) holds with b = 0. Assume that there exists a pointy > 0 such that hp (y) = +oo. It means that (a, y) E supp (p, L ), for all a > 0.

25) and h(y) > 0 for ally > 0. 25) that the equation y = h(y)u has the unique solution Yu for all u > 0. We now show that limsupyu ·u-tO ~c. 27) 44 LAGRANGE-TYPE FUNCTIONS wherec > 0. Sincehisdecreasing, weconcludethath(Yuk)::::; h(c+E) < +oo, hence Yuk = hp(Yuk)uk ::; hp(c + c)uk. 27). 26) holds. 2) Assume that dom hp supp (p,L) = (b, +oo). ,y): (a,y) E supp (p)} p(u, 1) max( sup (a,y):a>O,y«::;b min( au, y), sup a-5:_hp(y),y>b min( au, y)). 28) > 0, Let us check that, for each u min( au, y) sup = b.

Download PDF sample

Lagrange-type Functions in Constrained Non-Convex Optimization by Alexander M. Rubinov, Xiao-qi Yang

by Steven

Lagrange-type Functions in Constrained Non-Convex by Alexander M. Rubinov, Xiao-qi Yang PDF
Rated 4.07 of 5 – based on 35 votes