1: \begin{abstract}

2: We analyze the convergence of compressive sensing based

3: sampling techniques for the efficient evaluation of 

4: functionals of solutions for a class of 

5: high-dimensional, affine-parametric, linear operator equations

6: which depend on possibly infinitely many parameters.

7: The proposed algorithms are based on

8: so-called ``non-intrusive''

9: sampling of the high-dimensional parameter

10: space, reminiscent of Monte-Carlo sampling.

11: In contrast to Monte-Carlo, however, a functional of the parametric solution

12: is then computed via compressive sensing methods from samples of functionals of

13: the solution. 

14: A key ingredient in our analysis of independent interest

15: consists in a generalization of recent results on the approximate 

16: sparsity of generalized polynomial chaos representations (gpc)

17: of the parametric solution families,

18: in terms of the gpc series with respect to tensorized \Tscheb polynomials.

19: In particular, we establish sufficient conditions on the 

20: parametric inputs to the parametric operator equation

21: such that the \Tscheb coefficients of gpc expansion are contained 

22: in certain weighted $\ell_p$-spaces for $0<p\leq 1$.

23: Based on this we show that

24: reconstructions of the parametric solutions computed from

25: the sampled problems converge, with high probability,  

26: at the $L_2$, resp.\ $L_\infty$ convergence rates afforded 

27: by best $s$-term approximations of the parametric solution up to logarithmic factors.

28: \end{abstract}

29: