Equivariant maps from invariant functions

  • 2022-09-29 18:52:17
  • Ben Blum-Smith, Soledad Villar
  • 0

Abstract

In equivariant machine learning the idea is to restrict the learning to ahypothesis class where all the functions are equivariant with respect to somegroup action. Irreducible representations or invariant theory are typicallyused to parameterize the space of such functions. In this note, we explicate ageneral procedure, attributed to Malgrange, to express all polynomial mapsbetween linear spaces that are equivariant with respect to the action of agroup $G$, given a characterization of the invariant polynomials on a biggerspace. The method also parametrizes smooth equivariant maps in the case that$G$ is a compact Lie group.

 

Quick Read (beta)

loading the full paper ...