1: \begin{abstract}

2: This paper is concerned with the construction, analysis and realization of

3: a numerical method to approximate the solution of high dimensional elliptic

4: partial differential equations.

5: We propose a new combination of an Adaptive Wavelet Galerkin Method (AWGM) and

6: the well-known Hierarchical Tensor (HT) format.

7: The arising HT-AWGM is adaptive both in the wavelet representation of the low

8: dimensional factors and in the tensor rank of the HT representation.

9: 

10: The point of departure is an adaptive wavelet method for the HT format

11: using approximate Richardson iterations from \cite{DB} and an AWGM method

12: as described in \cite{GHS}.

13: HT-AWGM performs a sequence of Galerkin

14: solves based upon a truncated preconditioned conjugate gradient (PCG) algorithm

15: from \cite{ToblerPhd} in combination with a tensor-based

16: preconditioner from \cite{MarkusSobolev}.

17: 

18: Our analysis starts by showing convergence of the truncated

19: conjugate gradient method.

20: The next step is to add routines realizing the adaptive refinement.

21: The resulting HT-AWGM is analyzed concerning convergence and complexity.

22: We show that the performance of the 

23: scheme asymptotically depends only on the desired tolerance

24: with convergence rates depending on the Besov regularity of low dimensional

25: quantities and the low rank tensor structure of the solution.

26: The complexity in the ranks is algebraic with powers of four stemming

27: from the complexity of the tensor truncation. Numerical experiments

28: show the quantitative performance.

29: \end{abstract}

30: