1: \begin{abstract}

2:     Inferring a diffusion equation from discretely-observed measurements is a

3:     statistical challenge of significant importance in a variety of fields, from

4:     single-molecule tracking in biophysical systems to modeling financial

5:     instruments. Assuming that the underlying dynamical process obeys a

6:     $d$-dimensional stochastic differential equation of the form

7:         $$

8:             \d\x_t=\b(\x_t)\d t+\Sigma(\x_t)\d\w_t,

9:         $$

10:      we propose neural network-based estimators of both the drift $\b$ and the

11:      spatially-inhomogeneous diffusion tensor $D = \Sigma\Sigma^{T}$ and provide

12:      statistical convergence guarantees when $\b$ and $D$ are $s$-H\"older

13:      continuous. Notably, our bound aligns with the minimax optimal rate

14:      $N^{-\frac{2s}{2s+d}}$ for nonparametric function estimation even in the

15:      presence of correlation within observational data, which necessitates

16:      careful handling when establishing fast-rate generalization bounds. Our

17:      theoretical results are bolstered by numerical experiments demonstrating

18:      accurate inference of spatially-inhomogeneous diffusion tensors.

19: \end{abstract}

20: