First page Back Continue Last page Overview Text


Data fusion is originally developed for signal processing in that improved performance can be obtained by combing evidence from several different sources. When it is used in similarity searching, firstly we search for a target structure and rank the database structures in order of decreasing similarity. Then repeat with different representations, coefficients, etc. Next, add the rank positions for a given structure to give an overall fused rank position. And finally, the resulting fused ranking is the output from the search. There are small, but consistent, improvements in performance over use of a single ranking.