Fig 1 - uploaded by Haleh Karbasforoushan
Content may be subject to copyright.
Context in source publication
Context 1
... multi-dimensional indexing structures have been proposed to improve the multi-dimensional image databases performance. Most of the multi-dimensional indexing structures are based on the principle of hierarchical partitioning of the data space, so that they have a tree-like structure [4]. In these structures, as shown in Fig. 1, data points are stored in data nodes and each directory node points to a set of sub trees. There is a single directory node, which is called the root. The index structures are height-balanced; it means the lengths of the paths between the root and all data nodes are identical [3], ...
Similar publications
A notable problem with current information retrieval systems is that the input queries cannot express user information needs properly. This imprecise representation of the query hampers the effectiveness of the retrieval system. One method to solve this problem is to transform the original query into a more meaningful form. This paper proposes an o...
Citations
... Then, a deep neural network is adopted to learn a set of hierarchi-cal nonlinear transformations, using back-propagation to map sample pairs into other subspace, where each positive sample pair is less than a smaller threshold 1 and that of each negative pair is higher than a larger threshold 2 , so that discriminative information is exploited for the robustness accuracy in unconstrained environmental images Most of them are not strong enough to obtain the nonlinear transformation wherever face images usually lie, as these methods use linear functions to project face representations to a new feature space. To solve this problem, some methods use the kernel trick, which is commonly used to first map face representations to feature space with high dimension and then learn the distance metric in the high dimensional feature space [12,21,47]. But these methods cannot capture the nonlinear functions to solve the problem of scalability. ...
Metric learning aims to learn a distance to measure the difference between two samples, and it plays an important role in pattern recognition tasks. Most of the existing metric learning methods rely on pairs of samples. However, the importance of sample pairs varies greatly because of possible noise and the difference between samples and the decision boundaries. In this paper, we propose a robust hierarchical metric learning (SPHML) framework based on self-paced learning, which can help gain knowledge about the weights of sample pairs and utilize them in an easy or hard manner. Hierarchical nonlinear functions are learned by back-propagation to map sample pairs into a more discriminative feature space. Experimentally, our method achieves very competitive performance when compared with state-of-the-art methods.
... of multi-dimensional objects especially for high-dimensional spaces. In [9] a classification of high-dimensional indexing structures has been proposed. Many clustering algorithms are proposed by now, but only few research studies are conducted to propose clustering methods in the purpose of applying them to indexing of image information [10]. ...
In most practical applications of image retrieval, high-dimensional feature
vectors are required, but current multi-dimensional indexing structures lose
their efficiency with growth of dimensions. Our goal is to propose a divisive
hierarchical clustering-based multi-dimensional indexing structure which is
efficient in high-dimensional feature spaces. A projection pursuit method has
been used for finding a component of the data, which data's projections onto it
maximizes the approximation of negentropy for preparing essential information
in order to partitioning of the data space. Various tests and experimental
results on high-dimensional datasets indicate the performance of proposed
method in comparison with others.
