A Newton iteration for differentiable set-valued maps (Q1931549)

The authors are interested in solving the generalized equation \(f(x)+F(x)\ni 0\) and take advantage of some recent developments in this area, e.g. ([\textit{J. F. Bonnans}, Appl. Math. Optimization 29, No. 2, 161--186 (1994; Zbl 0809.90115)]; [\textit{A. L. Dontchev}, C. R. Acad. Sci., Paris, Sér. I 322, No. 4, 327--331 (1996; Zbl 0844.47034)]; [\textit{D. Azé} and \textit{C. C. Chou}, Math. Oper. Res. 20, No. 4, 790--800 (1995; Zbl 0860.41030)]; [\textit{S. Dias} and \textit{G. Smirnov}, Nonlinear Anal., Theory Methods Appl., Ser. A, Theory Methods 75, No. 3, 1219--1230 (2012; Zbl 1236.49068)]). Two new main results are as follows. Under the assumptions that the single-valued function \(f\) is Fréchet differentiable and the set-valued mapping \(F\) strictly \(H\)-differentiable, both of them acting between two Banach spaces \(X\) and \(Y\), and under additional assumptions on the derivatives of \(f\) and \(F\), the existence of a sequence \(\{x_n\}\) generated by the Newton type iteration constructed on the basis of linearization of both \(f\) and \(F\) converging \(Q\)-linearly to a solution is proved. After strengthening the assumptions, the \(Q\)-quadratic convergence of the method is obtained.