Abstract:
A continuous algorithm is proposed whereby the saddle points are found of convexconcave
functions specified on a subset of the Euclidean space for the case where the
functions are not assumed to be continuosly differentiable. The algorithm is represented
as a system of differential inclusions. The asymptotic properties of the algorithm are
studied with the aid of tools from convex analysis, the theory of monotone operators,
and the direct Lyapunov method.