1: \begin{abstract}

2:    Approximating functions by a linear span of truncated basis sets is a standard procedure for the

3:    numerical solution of differential and integral equations. Commonly used concepts of

4:    approximation methods are well-posed and convergent, by provable approximation orders. On the

5:    down side, however, these methods often suffer from the curse of dimensionality, which limits

6:    their approximation behavior, especially in situations of highly oscillatory target functions.

7:    Nonlinear approximation methods, such as neural networks, were shown to be very efficient in

8:    approximating high-dimensional functions.

9:    %

10:    We investigate nonlinear approximation methods that are constructed by composing standard basis

11:    sets with normalizing flows. Such models yield richer approximation spaces while maintaining the

12:    density properties of the initial basis set, as we show. Simulations to approximate

13:    eigenfunctions of a perturbed quantum harmonic oscillator indicate convergence with respect to

14:    the size of the basis set.

15: \end{abstract}

16: