Learning-to-Hash

A webpage dedicated to the latest research on learning-to-hash, including state-of-the-art deep hashing models, all updated on a weekly basis.

Contact Sean Moran about this survey or website. Made with Jekyll and Hyde.

Binary Quantisation Models

Quantisation models Two main categories of quantisation have been proposed for nearest neighbour search: scalar and vector quantisation, which are differentiated by whether the input and output of the quantisation is a scalar or a vector quantity. This page lists models for scalar quantisation. Scalar quantisation is frequently applied to quantise the real-values (projections) resulting from the dot product of the feature vector of each data-point onto the normal vectors to a set of hyperplanes partitioning the feature space. Each dot product yields a scalar value which is then subsequently quantised into binary (0/1) by thresholding. The resulting bits are concatenated to form the hashcode for a data-point.
PaperCodebookOptimisationLearning Type#Thresholds
Yue Cao, Mingsheng Long, Jianmin Wang, Qiang Yang, Philip S. Yu, 2016.Deep Visual-Semantic Hashing for Cross-Modal Retrieval Binary Backpropagation Supervised N/A
Yue Cao, Mingsheng Long, Jianmin Wang, Shichen Liu, 2017.Collective Deep Quantization for Efficient Cross-Modal Retrieval Binary Backpropagation Supervised N/A
W. Kong, W. Li, M. Guo, 2012.Manhattan Hashing for Large-Scale Image Retrieval Natural Binary Code (NBC) K-Means Unsupervised 3+
W. Kong, W. Li, 2012.Double-Bit Quantisation for Hashing Binary Squared Error Minimisation Unsupervised 2
Yunqiang Li, Wenjie Pei, Yufei zha, Jan van Gemert, 2019.Push for Quantization: Deep Fisher Hashing Any Backpropagation Supervised N/A
S. Moran, V. Lavrenko, and M. Osborne, 2013.Neighbourhood Preserving Quantisation for LSH Any Stochastic Search Semi-Supervised 1+
S. Moran, V. Lavrenko, and M. Osborne, 2013.Variable Bit Quantisation for LSH Any Stochastic Search + Binary Integer Linear Program Semi-Supervised Variable
S. Moran, 2016.Learning to Project and Binarise for Hashing-Based Approximate Nearest Neighbour Search Natural Binary Code (NBC) Maximum Margin + Stochastic Search Semi-Supervised Variable
Yang Shi, Xiushan Nie, Xin Zhou, Xiaoming Xi, Yilong Yin, 2019.Variable-Length Quantization Strategy for Hashing Any Greedy Strategy Unsupervised Variable
Z. Wang, L. Duan, J. Lin, X. Wang, T. Huang and W. Gao, 2015.Hamming Compatible Quantization for Hashing Binary Squared Error Minimisation Unsupervised 2
Z. Wang, L. Duan, T. Huang, W. Gao, 2016.Affinity Preserving Quantization for Hashing: A Vector Quantization Approach to Learning Compact Binary Codes Binary K-Means Unsupervised N/A
C. Xiong, W. Chen, G. Chen, D. Johnson, J. Corso, 2014.Adaptive Quantization for Hashing: An Information-Based Approach to Learning Binary Codes Natural Binary Code (NBC) Dynamic Programming Unsupervised Variable