Semantic Consistency Based Similarity Search Hashing Algorithm In Big Data

With the advent of the big data era,the data of the Internet expands sharply.Among which,fast neighbor search technology,which has great practical significance,encounters difficulties.To address the problem,hashing algorithm has become hot issue of research in the neighbor search field due to its efficiency.The basic idea of the hash algorithm is to map the high-dimensional data in the feature space into the low-dimensional binary code in the Hamming space,while preserving the similarity in the feature space.This paper first studies the Anchor Graph Hashing(AGH)algorithm,which denotes data by nodes in the graph of the feature space.Furthermore,the similarity between the nodes can be approximated by using the similarity between the nodes and the anchor nodes.Thus,the computationally prohibitive step of building large graph Laplacians can be overcome by using low rank similarity matrices.However,Anchor Graph Hashing algorithm measures the similarity between the nodes through Euclidean distance in the feature space,and the semantic similarity is not considered.In this paper,we propose Anchor Graph Hashing with Semantically Consistent Graph algorithm to learn the distance metric by exploiting semantic information based on a semi-supervised learning model.As a result,the optimized transformation matrix is obtained to improve the performance of the AGH algorithm.In order to speed up the convergence speed of learning,we employ the stochastic gradient descent method.Finally,we conduct experiments in the large-scale public image database to verify the performance of our proposed algorithm.The result shows that our proposed algorithm is superior to the AGH algorithm in terms of the search accuracy.