“…Theorem 1.1 establishes under suitable conditions that for every δ ∈ (0, ∞) there exists c ∈ R such that the solution X ∈ C([0, T ], R d ) of the differential equation X(t) = ξ + t 0 E[F (X(r), Z 0 )] dr, t ∈ [0, T ], (cf. Lemma 3.3), can be approximated by the recursive MLP approximation schemes in (1) with a root mean square error of size ε ∈ (0, 1] and a computational effort that is bounded by c ε −(2+δ) . The computational effort is quantified by the numbers RV n,m , n, m ∈ N. The function F : R d × S → R d is required to be (B(R d ) ⊗ S)/B(R d ) -measurable and Lipschitz continuous in the first variable, uniformly in the second variable, and we assume that E[ F (ξ, Z 0 ) 2 ] < ∞.…”