An SRN network was trained to predict the measurements of a dynamical system. After training, the output of the SRN could be used to rectify noise in the measurements. Importantly, such rectification is possible without explicit knowledge of the system dynamics.