1: \begin{abstract}

2:   {

3:     Crystallographic groups describe the symmetries of

4:     crystals and other repetitive structures encountered in nature and the sciences.

5:     These groups include the wallpaper and space groups.

6:     We derive linear and nonlinear representations of functions that are (1) smooth and

7:     (2) invariant under such a group.

8:     The linear representation generalizes the Fourier basis to crystallographically invariant

9:     basis functions. We show that such a basis exists for each crystallographic group,

10:     that it is orthonormal in the relevant $\L_2$ space, and recover the standard Fourier

11:     basis as a special case for pure shift groups. The nonlinear representation

12:     embeds the orbit space of the group into a finite-dimensional Euclidean space.

13:     We show that such an embedding exists for every crystallographic group, and that it factors functions

14:     through a generalization of a manifold called an orbifold.

15:     We describe algorithms that, given a standardized description of the group,

16:     compute the Fourier basis and an embedding map. As examples,

17:     we construct crystallographically invariant neural networks, kernel machines,

18:     and Gaussian processes.

19:   }

20: \end{abstract}

21: