It may happen that the simulation for a solution is not converged, which could lead to evaluation metrics which are artificially good. Already, we force evaluation metrics to be computed using 64-bit mode. Should other steps be taken? E.g. compute eval metrics for more wavelengths, or with different simulation settings?
It may happen that the simulation for a solution is not converged, which could lead to evaluation metrics which are artificially good. Already, we force evaluation metrics to be computed using 64-bit mode. Should other steps be taken? E.g. compute eval metrics for more wavelengths, or with different simulation settings?