Approximation of Functionals by Neural Network without Curse of Dimensionality