A Sparse Embedding and Least Variance Encoding Approach to Hashing

Abstract

Abstract-Hashing is becoming increasingly important in large-scale image retrieval for fast approximate similarity search and efficient data storage. Many popular hashing methods aim to preserve the kNN graph of high dimensional data points in the low dimensional manifold space, which is however difficult to achieve when the number of samples is big. In this paper, we propose an effective and efficient hashing approach by sparsely embedding a sample in the training sample space and encoding the sparse embedding vector over a learned dictionary. To this end, we partition the sample space into clusters via a linear spectral clustering method, and then represent each sample as a sparse vector of normalized probabilities that it falls into its several closest clusters. This actually embeds each sample sparsely in the sample space. The sparse embedding vector is employed as the feature of each sample for hashing. We then propose a least variance encoding model, which learns a dictionary to encode the sparse embedding feature, and consequently binarize the coding coefficients as the hash codes. The dictionary and the binarization threshold are jointly optimized in our model. Experimental results on benchmark datasets demonstrated the effectiveness of the proposed approach in comparison with stateof-the-art methods.Index Terms-hashing, manifold learning, image retrieval, dictionary learning.

Unsupervised Deep Hashing With Pseudo Labels for Scalable Image Retrieval

“…We systematically compare our method with eight state-of-the-art non-deep methods: ITQ [22], PCAH [20], LSH [13], DSH [55], SpH [44], SH [19], AGH [21], and SELVE [56], and three deep methods: DH [39], Deepbit [40], UH-BDNN [41] for retrieval task, all these eleven methods are unsupervised. All of the non-deep methods and UH-BDNN [41] in our experiments use the same VGG [58] fc7 feature as that in our method, and DH [39] and Deepbit [40] are based the same settings like that in their original papers.…”

“…Since the NUS-WIDE is a multi-label dataset, and the number of latent classes is fixed to be 5 (the reason why we choose 5 latent classes will be analysed in the following subsection), longer codes make the intermediate representation non-compact and result in redundant dimensions, which is same as that in PCAH [20]. In addition, this phenomenon of mAP reduction with code length growing also appears in AGH [21] and SELVE [56].…”

Efficient subspace clustering based on self-representation and grouping effect

“…Moreover, a set of high-dimensional data is often drawn from multiple low-dimensional subspaces [7], such as the face images, the point trajectories of moving objects [8] and the texture features of pixel on an image [9][10][11]. Recently, subspace clustering [12] processes this kind of data by following their underlying subspaces to attract increasing attentions.…”

Block Nested Join and Sort Merge Join Algorithms: An Empirical Evaluation

“…For example, many companies usually encounter the troubles for handling tremendous amounts of complex information. It is a necessary factor for they organize and analyze these data [23,25,28,29]. Data warehousing and online analytical processing (OLAP) applications, as very hot technologies, have been effectively contributed to the application of decision support systems (DSS).…”