Currently, we compute error indicators in enriched space (positive by construction), transfer them back to the base space, and then interpolate to ensure that the result is still positive.
Might we be able to avoid this second interpolation using Bernstein polynomials?
Currently, we compute error indicators in enriched space (positive by construction), transfer them back to the base space, and then interpolate to ensure that the result is still positive.
Might we be able to avoid this second interpolation using Bernstein polynomials?
Parent issue: https://github.com/pyroteus/pyroteus/issues/9.