Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
This paper introduces a novel metric framework to quantify the robustness of formulation-dependent properties in mathematical programming by modeling formulations as typed collections of constraint functions and defining distance via optimal matching and norm discrepancies. As a key application, it defines and analyzes the pointwise LICQ radius, representing the distance from a feasible formulation to one failing LICQ at a specific point. The work derives an explicit formula for this radius and characterizes the impact of adding single constraints.
The LICQ radius quantifies how close a formulation is to violating the Linear Independence Constraint Qualification, revealing subtle formulation sensitivities.
Mathematical programming formulations are routinely modified through algebraic reformulations, perturbations of constraint functions, and the addition or removal of constraints. Some of these modifications preserve the feasible set, whereas others reflect newly imposed modeling requirements and therefore change the optimization problem itself. This paper develops a metric framework for quantifying the robustness of formulation-dependent properties. Formulations are modeled as typed collections of constraint functions. Within each fixed typed-cardinality class, their distance is defined by optimally matching constraints of the same type and measuring the resulting discrepancies in a norm on the underlying function space. As a concrete demonstration of the framework, we study the Linear Independence Constraint Qualification (LICQ). We introduce the pointwise LICQ radius, defined as the distance from a feasible formulation to the set of formulations that remain feasible at a prescribed point but fail LICQ there. Under suitable assumptions, we derive an explicit formula for this radius and characterize the effect of adding a single equality or inequality constraint.