Equivariant Learning of Stochastic Fields:
      Gaussian Processes and Steerable Conditional Neural Processes

               Peter Holderrieth * 1 Michael Hutchinson * 1 Yee Whye Teh 1 2

              Abstract
   Motivated by objects such as electric fields or
   fluid streams, we study the problem of learning
   stochastic fields, i.e. stochastic processes whose
   samples are fields like those occurring in physics
   and engineering. Considering general transforma-
   tions such as rotation ...