Abstract: | There are many papers dealing with the approximate solution of linear problems where only partial information is available. Two types of information have been considered: linear and discontinuous nonlinear. In particular, we know that discontinuous nonlinear information is far more powerful than linear information. In this paper we study continuous nonlinear information for linear problems, and we prove that: • -it is no more powerful than linear information in the worst case setting, • -it is much more powerful than linear information in the average case setting. |