A new game-theoretic approach tackles non-convex constrained optimization.
problem Non-convex constrained optimization with non-differentiable constraints.
method Proxy-Lagrangian formulation, two-player game approach.
result Classifier size is significantly reduced to m+1 models.