currently in PHI-PSI when we compute the weighted sum of matrix entries and the weighted sum of sub- scores, all the segments in the database share the same two sets of weights: one set associated with the similarity matrices, the other associated with the window. This is almost certainly a suboptimal algorithm, because the “important properties” and the “important positions” may be different in each case, and even the most appropriate window size for each case may vary. We are experimenting with different heuristics and parameters for finding best matches for an input.