Lagrange Multiplier Local Necessary and Global Sufficiency Criteria for Some Non-Convex Programming Problems

B. Muraleetharan,S. Selvarajan,S. Srisatkunarajah,K. Thirulogasanthar

DOI: https://doi.org/10.48550/arXiv.1804.02270

2018-03-19

Abstract:In this paper we consider three minimization problems, namely quadratic, $\rho$-convex and quadratic fractional programing problems. The quadratic problem is considered with quadratic inequality constraints with bounded continuous and discrete mixed variables. The $\rho$-convex problem is considered with $\rho$-convex inequality constraints in mixed variables. The quadratic fractional problem is studied with quadratic fractional constraints in mixed variables. For all three problems we reformulate the problem as a mathematical programming problem and apply standard Karush Kuhn Tucker necessary conditions. Then, for each problem, we provide local necessary optimality condition. Further, for each problem a Lagrangian multiplier sufficient optimality condition is provided to identify global minimizer among the local minimizers. For the quadratic problem underestimation of a Lagrangian was employed to obtain the desired sufficient conditions. For the $\rho$-convex problem we obtain two sufficient optimality conditions to distinguish a global minimizer among the local minimizers, one with an underestimation of a Lagrangian and the other with a different technique. A global sufficient optimality condition for the quadratic fractional problem is obtained by reformulating the problem as a quadratic problem and then utilizing the results of the quadratic problem. Examples are provided to illustrate the significance of the results obtained.

Optimization and Control

What problem does this paper attempt to address?

The problems that this paper attempts to solve mainly focus on the establishment of local necessary conditions and global sufficient conditions for three non - convex optimization problems, as follows:

1. **Quadratic Programming Problem**:
- This problem considers the minimization of a quadratic function with quadratic inequality constraints, where the variables include continuous and discrete variables.
- The paper provides local necessary conditions by rewriting the problem as a mathematical programming problem and applying the standard Karush - Kuhn - Tucker (KKT) necessary conditions.
- Further, the paper provides a global sufficient condition for identifying the global minimum among local minima by underestimating the Lagrangian function.

2. **ρ - Convex Programming Problem**:
- This problem considers the minimization of a function with ρ - convex inequality constraints, where the variables are also mixed variables.
- The paper also provides local necessary conditions by rewriting the problem as a mathematical programming problem and applying the KKT necessary conditions.
- The paper proposes two global sufficient conditions: one is by underestimating the Lagrangian function, and the other is by a different technical method, for identifying the global minimum among local minima.

3. **Quadratic Fractional Programming Problem**:
- This problem considers the minimization of a function with quadratic fractional constraints, where the variables are also mixed variables.
- The paper provides a global sufficient condition for identifying the global minimum among local minima by rewriting the problem as a quadratic problem and using the results of the quadratic problem.

The main objective of the paper is to help researchers and engineers more effectively identify the global optimal solution when dealing with non - convex optimization problems, especially in applications in fields such as electronic circuit design, computational chemistry, and combinatorial optimization.