Dynamic Partition Forest: An Efficient and Distributed Indexing Scheme for Similarity Search based on Hashing

The similarity search over large-scale feature-rich data(e.g. image, video or text) is a fundamental problem and has become increasingly important in data mining research. Hashing based methods, especially Locality Sensitive Hashing(LSH), have been widely used for fast Approximate Nearest Neighbor search(ANNs). However, there are still two flaws in existing methods: (1) The state-of-the-art distribution scheme sacrificed too much accuracy for …