Mixed Integer Quadratic Programming (MIQP): Mathematical Optimization with Discrete-Continuous Variables

Mixed Integer Quadratic Programming (MIQP) represents a critical class of mathematical optimization problems with significant real-world applications, including portfolio optimization, scheduling challenges, and engineering design. MIQP combines linear constraints, quadratic objective functions, and variables that include both continuous and integer types, making it more complex than traditional linear programming or pure integer programming.

MIQP Basic Structure The standard form of an MIQP problem typically comprises a quadratic objective function subject to linear equality or inequality constraints, with some variables restricted to integers. The mathematical formulation generally follows: Objective function: Minimize or maximize a quadratic expression, e.g., ( x^T Q x + c^T x ) Constraints: ( A x \leq b ) or ( A_{eq} x = b_{eq} ) Variable restrictions: Certain components of ( x ) must be integers

Solution Methods for MIQP Branch and Bound: The most common approach, which partitions the feasible solution space iteratively and computes bounds to approximate the optimal solution. Implementation often involves tree traversal and pruning based on objective value comparisons. Cutting Plane Method: Adds linear constraints to progressively reduce the feasible region, suitable for specific MIQP structures. Algorithms like Gomory cuts can be integrated to strengthen formulations. Heuristic Algorithms: Techniques such as genetic algorithms or simulated annealing address large-scale problems but do not guarantee global optimality. These are useful for initial feasible solutions or when exact methods are computationally prohibitive.

Practical Applications and Optimization Tools Commercial and open-source solvers (e.g., Gurobi, CPLEX, SCIP) efficiently handle MIQP problems. In practice, selecting appropriate solvers and employing modeling techniques—such as relaxation (e.g., ignoring integer constraints temporarily) or decomposition methods (e.g., Benders decomposition)—can dramatically improve solving efficiency. Code implementation often involves defining variable types (continuous/integer) and configuring solver parameters for tolerance and convergence.

Despite its high computational complexity, MIQP remains a vital tool across industries, especially in scenarios requiring discrete decisions combined with continuous optimization. Through careful modeling and algorithmic selection, it effectively addresses complex real-world optimization challenges.