A differentiable homotopy to compute Nash equilibria of \(n\)-person games (Q5942327)

The existing methods of computing Nash equilibria in \(n\)-person (non-cooperative) games use a variety of techniques. For two-person games linear a nonlinear programming methods are available [see \textit{T. Parthasathy} and \textit{T. E. S. Rhagavan}, Some Topics in Two-Person Games, Elsevier (1971; Zbl 0225.90049) for a survey]. In more general situations one can represent Nash equilibria as fixed points of specific operators and compute them as such (a survey of such methods is contained in \textit{M. Todd} [The Computation of Fixed Points and Applications (1974; Zbl 0332.54003)]). Alternatively, one can apply dynamic system related techniques as those presented by \textit{A. S. Antipin} [From optima to equilibria, in: ``Dynamic and Non-Homogeneous Systems'', Yu. S. Popov (ed.), Russian Academy of Sciences, Editorial URSS, Moscow, 35-64 (2000)] or iterative methods as those occurring in the paper of \textit{D. Butnariu} and \textit{E. Resmerita} [``Averaged subgradient methods for constrained convex optimization and Nash equilibria computation'', Optimization (to appear)]. The authors present an algorithm of determining Nash equilibria which exploits the special differentiability properties of the problem data. A special merit of this new method is that among the many equilibria of the game it finds one with very interesting features, namely, the equilibrium which is selected by the tracing procedure of \textit{J. C. Harsanyi} and \textit{R. Selten} [A General Theory of Equilibrium Selection in Games, Cambridge, MIT Press (1988; Zbl 0693.90098)].