Min-Max Optimization for Robust Nonlinear Least Squares Problems

This paper considers min-max optimization for a class of robust nonlinear least squares problems. We show via an example that solving the first order optimality conditions defined by gradients of the objective function can lead to incorrect solutions of min-max problems. We give an explicit formula for the value function of the inner maximization problem. Using the formula, we show that finding a first order ϵ-approximate necessary minimax point of the min-max problem needs at most O(|logϵ| +ϵ^-2) evaluations of the function value and gradients of the objective function. Moreover, we establish error bounds from any solution of the nonlinear least squares problem to the solution set of the robust nonlinear least squares problem.