Infinite-dimensional operator ridgelet transform

6. Examples with genuinely infinite-dimensional inputs🔗

This chapter is Section 7 of the manuscript together with Appendix E. A function f on H is cylindrical if f=\widetilde f\circ L for a finite-rank linear map L; all examples are non-cylindrical, so finite width is never obtained by truncating the input. We write \phi_v for the centred one-dimensional Gaussian density of variance v>0, with \phi_0=\delta_0, and \Phi(u)=e^{-u^2/2} for the Gaussian activation.

  1. 6.1. Preliminaries
  2. 6.2. A Gaussian target with a closed-form transform
  3. 6.3. Gaussian-parameter networks on the sequence space
  4. 6.4. Neural-operator layers