Sequential algorithms for observation selection

Some signal reconstruction problems allow for flexibility in the selection of observations and, hence, the signal formation equation. In such cases, w e have the opportunity to determine the best combination of observations be fore acquiring the data. We present and analyze two classes of sequential a lgorithms to select observations-sequential backward selection (SBS) and se quential forward selection (SFS), Although both are suboptimal, they perfor m consistently well. We analyze the computational complexity of various for ms of SBS and SFS and develop upper bounds on the sum of squared errors (SS E) of the solutions obtained by SBS and SFS.