PROJECT TITLE :

Indexing Earth Mover’s Distance over Network Metrics

ABSTRACT:

The Earth Mover's Distance (EMD) may be a well-known distance metric for knowledge represented as likelihood distributions over a predefined feature space. Supporting EMD-based similarity search has attracted intensive research effort. Despite the plethora of literature, most existing solutions are optimized for Lp feature areas (e.g., Euclidean area); while in an exceedingly spectrum of applications, the relationships between options are higher captured using networks. During this paper, we study the matter of answering k-nearest neighbor (k-NN) queries below network-based EMD metrics (NEMD). We have a tendency to propose OASIS, a new access technique which leverages the network structure of feature house and permits economical NEMD-based mostly similarity search. Specifically, OASIS employs 3 novel techniques: (i) Vary Oracle, a scalable model to estimate the vary of k-th nearest neighbor beneath NEMD, (ii) Boundary Index, a structure that efficiently fetches candidates within given vary, and (iii) Network Compression Hierarchy, an incremental filtering mechanism that effectively prunes false positive candidates to save lots of unnecessary computation. Through in depth experiments using both synthetic and real data sets, we tend to confirmed that OASIS significantly outperforms the state-of-the-art methods in query processing value.