Efficient high-dimensional retrieval in structured P2P networks

Known by its exceptional scalability and flexibility, Peer-to-peer (P2P) technique is arguably one of the most important mechanisms for sharing massive data (e.g. media data). It has been challenging to support similarity search in structured P2P networks, though it provides efficient indexing for exact search. In this paper, we present an efficient indexing technique to support complex similarity retrieval on high-dimensional data by improving existing approach Multi-dimensional Rectangulation with Kd-trees (MURK). In order to make search more user-centric, relevance feedback techniques are also investigated. To the best of our knowledge, it is the first attempt of utilizing relevance feedback in structured P2P networks. Simulations for content based music retrieval with multiple acoustic features have been conducted to investigate the properties and efficiency of the proposed approach.