A new global-local procedure for solving nonlinear systems is proposed
. Each iteration of the procedure is a combination of an approximate N
ewton step and a descending step for minimizing the square-sum objecti
ve function. Starting with an arbitrary initial point, the procedure c
onverges either to a solution of the system or to a local minimizer of
the square-sum. When converging io a solution of the system, the loca
l quadratic rate is achieved.