Main Article Content
Abstract
Surmised closest neighbor (ANN) search has made incredible progress in numerous errands. Be that as it may, existing mainstream techniques for ANN search, for example, hashing and quantization strategies, are intended for static databases as it were. They can't deal with well the database with information circulation advancing powerfully, because of the high computational exertion for retraining the model dependent on the new database. In Optimized Product Quantization for Approximate Nearest Neighbor Search paper, we address the issue by building up an online item quantization (online PQ) model and steadily refreshing the quantization codebook that suits to the approaching gushing information. Besides, to additionally reduce the issue of huge scale calculation for the online PQ update, we plan two spending requirements for the model to refresh fractional PQ codebook rather than all. KDD [KNOWLEDGE DISCOVERY DATABASE] dataset is proposed to address item quantization. The suitable mediator clients can be effectively found by utilizing this dataset. KDD dataset is completely founded on CODEBOOK. In this work, the issue is tended to by building up an online item quantization (online PQ) model and steadily refreshing the quantization codebook that suits to the approaching gushing information. An item quantization can produce an exponentially huge codebook at low space.