Order-based Equivalence Degrees for Similarity and Distance Measures
Abstract
In order to help to choose similarity or distance measures for information retrieval systems, we compare the orders these measures induce and quantify their agreement by a degree of equivalence. We both consider measures dedicated to binary and numerical data, carrying out experiments both on artificial and real data sets, and identifying equivalent as well as quasi-equivalent measures that can be considered as redundant in the information retrieval framework.
Domains
Artificial Intelligence [cs.AI]
Origin : Publisher files allowed on an open archive
Loading...