New Permutation Dissimilarity Measures For Proximity Searching

Proximity searching consists in retrieving the most similar objects to a given query from a database. To do so, the usual approach consists in using an index in order to improve the response time of online queries. Recently, the permutation based algorithms (PBA) were presented, and from then on, this technique has been very successful. In its core, the PBA uses a metric between permutations, typically Spearman Footrule or Spearman Rho. Until now, several proposals based on the PBA have been developed and all of them uses one of those metrics. In this paper, we present a new family of dissimilarity measures between permutations. According to our experimental evaluation, we can reduce up to 30% the original technique costs, while preserving its exceptional answer quality. Since our dissimilarity measures can be applied in any state-of-the-art PBA variant, the impact of our proposal is significant for the similarity search community.