Iterative quantization: A procrustean approach to learning binary codes

By Yunchao Gong, Svetlana Lazebnik. 2011

1,206 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Media Technology

In Representation & Self-Supervised Learning

Citations in the corpus

Billion-Scale Similarity Search with GPUs (2017, background): “The most popular vector compression methods can be classiﬁed into either binary codes [18, 22], or quantization methods [25, 37].”
Unbounded cache model for online language modeling with open vocabulary (2017, neutral): “Several works have built on this initial binarization technique, such as spectral hashing [54], or Iterative Quantization (ITQ) [19].”
FastText.zip: Compressing text classification models (2016, methodology): “Many subsequent works have improved this initial binarization technique, such as spectal hashing (Weiss et al., 2009), or Iterative Quantization (ITQ) (Gong & Lazebnik, 2011), which learns a rotation matrix minimizing the quantization loss of the binarization.”
Analyzing the Performance of Multilayer Neural Networks for Object Recognition (2014, background): “Such codes are important for practical applications in large-scale image retrieval and mobile image analysis [8,24].”
A Multi-View Embedding Space for Modeling Internet Images, Tags, and Their Semantics (2012, methodology): “C V ] 2 S ep 2 Several promising recent approaches for modeling images and associated text (Gong and Lazebnik, 2011; Hardoon et al., 2004; Hwang and Grauman, 2010, 2011; Rasiwasia et al., 2010) rely on canonical correlation analysis (CCA), a classic technique that maps two views, given by visual…”
Accelerating Very Deep Convolutional Networks for Classification and Detection (2015, background): “This problem belongs to a broader 273 category of procrustes problems [28] that have been 274 adopted for various data reconstruction problems 275 [34], [35], [36].”
Deep Convolutional Ranking for Multilabel Image Annotation (2013, background): “Modeling Internet images and their corresponding textural information (e.g., sentences, tags) have been of great interest in the vision community [2, 10, 12, 15, 30, 29, 34].”
Deep Hashing Network for Unsupervised Domain Adaptation (2017, methodology): “Results and Discussion: We applied the precision-recall curves and the mean average precision (mAP) measures to evaluate the efficacy of the hashing methods, similar to previous research [7, 11, 21].”
Supervised Discrete Hashing (2015, methodology): “shing methods in large-scale image retrieval. 1. Introduction Hashing has attracted considerable attention of researchers in computer vision, machine learning, information retrieval and related areas [8,10,18,33,35,39]. Hashing techniques encode documents, images, videos or other sorts of data by a ”
Large-Scale Supervised Multimodal Hashing with Semantic Correlation Maximization (2014, methodology): “To tackle this issue, (Gong and Lazebnik 2011) and (Kong and Li 2012b) proposed orthogonal rotation learning methods to reduce the quantization error while preserving the orthogonality constraints.”
ImageNet Auto-Annotation with Segmentation Propagation (2014, methodology): “In this paper we employ the efﬁcient binary coding scheme called “Iterative Quantization” (ITQ) [27] to circumvent this issue.”
Comparative Evaluation of Binary Features (2012, background): “Binarized descriptors rely on hashing techniques to reduce high-dimensional, real-value parameterizations into compact binary codes [22–25].”
Correlation Hashing Network for Efficient Cross-Modal Retrieval (2016, methodology): “Furthermore, DCMH adopts Itera- tive Quantization (ITQ) [Gong and Lazebnik, 2011] to generate binary codes, which may be not robust to outlier bits when the codes are unbalanced.”
Polysemous Codes (2016, background): “Many variants have been subsequently proposed, such as spectral hashing [52] or ITQ [21] – see also [43,55,50] for representative works.”
Locally Optimized Product Quantization for Approximate Nearest Neighbor Search (2014, background): “At the other extreme, mapping data points to compact binary codes is not only efficient in space but may achieve fast exhaustive search in Hamming space [10, 16].”
Deep learning of binary hash codes for fast image retrieval (2015, methodology): “Unsupervised hashing methods use unlabeled data to learn a set of hash functions [6, 29, 8].”
Isotropic Hashing (2012, methodology): “Hence, using the same number of bits for different projected dimensions with unequal variances is unreasonable, which has also been argued by many researchers [7, 12, 21].”
Learning Compact Binary Face Descriptor for Face Recognition (2015, methodology): “Representative feature learning methods include sparse autoencoders [5], denoising auto-encoders [47], restricted Boltzmann machine [18], convolutional neural networks [21], independent subspace analysis [24], and reconstruction independent component analysis [27].”