Residual of an inverse problem ($r=Ku-f$)

The residual is the difference between predicted and observed data. Its norm measures data misfit and can determine an iteration stopping rule for regularization of an inverse problem.