A Scalable Optimization Mechanism for Pairwise Based Discrete Hashing

Maintaining the pairwise relationship among originally high-dimensional data into a low-dimensional binary space is popular strategy to learn codes. One simple and intuitive method utilize two identical code matrices produced by hash functions approximate real label matrix. However, resulting quartic problem in term of difficult directly solve due non-convex non-smooth nature objective. In this paper, unlike previous optimization methods using various relaxation strategies, we aim original novel alternative mechanism linearize introducing linear regression model. Additionally, find that gradually learning each batch codes sequential mode, i.e. batch, greatly beneficial convergence learning. Based on significant discovery proposed strategy, introduce scalable symmetric discrete hashing algorithm smoothly updates To further improve smoothness, also propose greedy update bit Moreover, extend problems for many other based algorithms. Extensive experiments benchmark single-label multi-label databases demonstrate superior performance over recent state-of-the-art kinds retrieval tasks: similarity ranking order. The source are available https://github.com/xsshi2015/Scalable-Pairwise-based-Discrete-Hashing.