Constrained Optimization Problems

In this paper, a brand new augmented Lagrangian penalty characteristic for confined optimization issues is studied. The twin homes of the augmented Lagrangian objective penalty characteristic for restrained optimization issues are proved. under a few conditions, the saddle factor of the augmented Lagrangian objective penalty feature satisfies the first-order Karush-Kuhn-Tucker (KKT) situation. specifically, when the KKT situation holds for convex programming its saddle factor exists. based totally on the augmented Lagrangian goal penalty characteristic, an algorithm is advanced for locating a international strategy to an inequality limited optimization problem and its global convergence is likewise proved below a few conditions.
Augmented Lagrangian penalty capabilities are powerful processes to inequality confined optimization. Their major concept is to convert a limited optimization trouble into a series of unconstrained optimization problems that are less difficult to clear up. Theories on and algorithms of Lagrangian penalty function have been added in Du’s et al. works [1] . Many researchers have tried to locate opportunity augmented Lagrangian features. Many literatures on augmented Lagrangian (penalty) features were posted from each theoretical and practical components (see [2] – [8] ), whose key worries cover 0 hole of dual, lifestyles of saddle factor, exactness, set of rules and so forth.
All augmented Lagrangian functions include two elements, a Lagrangian characteristic Constrained with a Lagrangian parameter and a penalty feature with a penalty parameter (see [2] – [8] ). dual and saddle point is the key worries of augmented Lagrangian function. furthermore, zero gap of Lagrangian characteristic’s dual is true only for convex programming and augmented Lagrangian function. consequently, augmented Lagrangian characteristic algorithms clear up a chain of restricted optimization troubles by taking differential Lagrangian parameters and penalty parameters in [2] [3] [4] [5] . Lucidi [6] and Di Pillo et al. [7] received some consequences of genuine augmented Lagrangian feature, however numerical outcomes had been now not given. R. S. Burachik and C. Y. Kaya gave an augmented Lagrangian scheme for a wellknown optimization hassle, and mounted for this replace primal-dual convergence the augmented penalty technique in [8] . however, on the subject of computation, to apply these methods, masses of Lagrangian parameters or penalty parameters need to be adjusted to solve a few unconstrained optimization dual problems, which make it hard to gain an optimization way to the unique trouble. for this reason, it’s far significant to observe a novel augmented Lagrangian characteristic technique.
In current years, the penalty characteristic approach with an objective penalty parameter has been discussed in [9] – [16] . Burke [12] taken into consideration a extra widespread type. Fiacco and McCormick [13] gave a trendy introduction to sequential unconstrained minimization strategies. Mauricio and Maculan [14] discussed a Boolean penalty method for zero-one nonlinear programming. Meng et al. [15] studied a widespread goal penalty characteristic technique. moreover, Meng et al. studied residences of twin and saddle factors of the augmented Lagrangian goal penalty feature in [16] . here, a new augmented Lagrangian goal penalty characteristic which differs from the one in [16] is studied. some vital outcomes just like those of the augmented Lagrangian goal penalty function in [16] are received.
the main conclusions of this paper consist of that the most useful goal value of the dual problem and the greatest target price of the unique hassle is 0 hole, and saddle factor is equal to the KKT condition of the authentic trouble under the convexity conditions. A global set of rules and its convergence are provided. The the rest of this paper is prepared as follows. In segment 2, an augmented Lagrangian goal penalty feature is described, Constrained its dual properties are proved, and an algorithm to find a worldwide way to the unique hassle (P) with convergence is offered. In phase 3, conclusions are given.