1: \begin{abstract}

2: We develop new adaptive algorithms for temporal 

3: integration of nonlinear evolution 

4: equations on tensor manifolds.

5: %

6: These algorithms, which we call step-truncation 

7: methods, are based on performing one time 

8: step with a conventional time-stepping 

9: scheme, followed by a truncation operation 

10: onto a tensor manifold. 

11: %

12: By selecting the rank of the tensor manifold 

13: adaptively to satisfy stability and accuracy 

14: requirements, 

15: %

16: we prove convergence of a wide range 

17: of step-truncation methods, including 

18: explicit one-step and multi-step methods.

19: %

20: These methods are very easy to implement 

21: as they rely only on arithmetic operations 

22: between tensors, which can be performed by 

23: efficient and scalable parallel algorithms.

24: %

25: Adaptive step-truncation methods 

26: can be used to compute numerical solutions 

27: of high-dimensional PDEs, which have become 

28: central to many new areas of application 

29: such optimal mass transport, random dynamical 

30: systems, and mean field optimal control.

31: %

32: Numerical applications are presented and 

33: discussed for a Fokker--Planck equation

34: with spatially dependent drift on a flat 

35: torus of dimension two and four.

36: \keywords{

37: High-dimensional PDEs \and

38: Low-rank tensor manifolds \and

39: Dynamical tensor approximation\and 

40: Tensor train\and 

41: Hierarchical Tucker format.}

42: \end{abstract}

43: