1: \begin{abstract}

2: In this paper we present nonparametric estimators for coefficients in stochastic 

3: differential equation if the data are described by independent, identically distributed 

4: random variables. The problem is formulated as  a nonlinear ill-posed operator equation 

5: with a deterministic forward operator described by the Fokker-Planck equation. 

6: We derive convergence rates of the risk for penalized maximum likelihood estimators 

7: with convex penalty terms and for Newton-type methods. The assumptions of our 

8: general convergence results are verified for estimation of the drift coefficient. 

9: The advantages of log-likelihood compared to quadratic data fidelity 

10: terms are demonstrated in Monte-Carlo simulations. 

11: \end{abstract}

12: