Neural networks and the solution of nonlinear equations.

The statistical nature of the real solutions of a set of αN random nonlinear equations constructed from N variables is examined, using analytical methods developed for neural networks. The nonlinearity of the equations is characterized by powers of the variables. The maximal α under which the equations have a real solution is found for various cases. The application of the results to physical systems is also discussed