From one graph to many: Ensemble transduction for content-based database retrieval

Segmentation is a fundamental task in medical image analysis. Current state-of-the-art Convolutional Neural Networks on medical image segmentation capture local context information using fixed-shape receptive fields and feature detectors with position-invariant weights, which limits the robustness to the variance of input, such as medical objects of variant sizes, shapes, and domains. In order to capture global context information, we propose High-order Attention (HA), a novel attention module with adaptive receptive fields and dynamic weights. HA allows each pixel to has its own global attention map that models its relationship to all other pixels. In particular, HA constructs the attention map through graph transduction and thus captures high relevant context information at high-order. Consequently, feature maps at each position are selectively aggregated as a weighted sum of feature maps at all positions. We further embed the proposed HA module into an efficient encoder-decoder structure for medical image segmentation, namely High-order Attention Network (HANet). Extensive experiments are conducted on four benchmark sets for three tasks, i.e., REFUGE and Drishti-GS1 for optic disc/cup segmentation, DRIVE for blood vessel segmentation, and LUNA for lung segmentation. The results justify the effectiveness of the new attention module for medical image segmentation.

Hyperspectral band selection (BS) is an important technique to reduce data dimensionality for the classification applications of hyperspectral remote sensing images (HSIs). Recently, searching-based BS methods have received increasing attention for their ability to select the best subset of bands while preserving the essential information of the original data. However, existing searching-based BS methods neglect effective exploitation of the spatial and spectral prior information inherent in the data, thus limiting their performance. To address this problem, in this study, a novel unsupervised BS method called Spectral–Spatial Iterative Greedy Algorithm (SSIGA) is proposed. Specifically, to facilitate efficient local search using spectral information, SSIGA conducts clustering on all the bands by employing a K-means clustering method with balanced cluster size constraints and constructs a K-nearest neighbor graph for each cluster. Based on the nearest neighbor graphs, SSIGA can effectively explore the neighborhood solutions in local search. In addition, to efficiently evaluate the discriminability and information redundancy of the solution given by SSIGA using the spatial and spectral information of HSIs, we designed an effective objective function for SSIGA. The value of the objective function is derived by calculating the Fisher score for each band in the solution based on the results of the superpixel segmentation performed on the target HSI, as well as by computing the average information entropy and mutual information of the bands in the solution. Experimental results on three publicly available real HSI datasets demonstrate that the SSIG algorithm achieves superior performance compared to several state-of-the-art methods.

In this paper, we study the problem of finding the shortest path in stochastic graphs and propose an iterative algorithm for solving it. This algorithm is based on distributed learning automata (DLA), and its objective is to use a DLA for finding the shortest path from the given source node to the given destination node whose weight is minimal in expected sense. At each stage of this algorithm, DLA specifies edges needed to be sampled. We show that the given algorithm finds the shortest path with minimum expected weight in stochastic graphs with high probability which can be close to unity as much as possible. We compare the given algorithm with some distributed learning automata-based iterative algorithms, a particle swarm optimization-based algorithm, an ant colony-based algorithm, a Q-learning-based algorithm, and an actor–critic-based algorithm for finding the shortest path. Computer experiments show that the proposed algorithm requires fewer edge samples to find the shortest path than the previously introduced DLA-based algorithms.

Tensorial Multi-view Clustering (TMC), a prominent approach in multi-view clustering, leverages low-rank tensor learning to capture high-order correlation among views for consistent clustering structure identification. Despite its promising performance, the TMC algorithms face three key challenges: 1). The severe computational burden makes it difficult for TMC methods to handle large-scale datasets. 2). Estimation bias problem caused by the convex surrogate of the tensor rank. 3). Lack of explicit balance of consistency and complementarity. Being aware of these, we propose a basic framework Efficient and Scalable Tensorial Multi-View Subspace Clustering (ESTMC) for large-scale multi-view clustering. ESTMC integrates anchor representation learning and non-convex function-based low-rank tensor learning with a Generalized Non-convex Tensor Rank (GNTR) into a unified objective function, which enhances the efficiency of the existing subspace-based TMC framework. Furthermore, a novel model ESTMC-C with the proposed Enhanced Tensor Rank (ETR), Consistent Geometric Regularization (CGR), and Tensorial Exclusive Regularization (TER) is extended to balance the learning of consistency and complementarity among views, delivering divisible representations for the clustering task. Efficient iterative optimization algorithms are designed to solve the proposed ESTMC and ESTMC-C, which enjoy time-economical complexity and exhibit theoretical convergence. Extensive experimental results on various datasets demonstrate the superiority of the proposed algorithms as compared to state-of-the-art methods.

A challenging problem of image retrieval is the similarity learning between images. To improve similarity search in Content-Based Image Retrieval (CBIR), many studies on distance metric learning have been published. Despite their success, most existing methods are limited in two aspects: (i) they usually attempt to learn a linear distance metric, which limits their capacity of measuring similarity for complex applications; (ii) they are often designed for learning metrics on unique-modal data, which could be suboptimal for similarity learning on multimedia objects with multiple feature representations. To overcome these limitations, in this paper, we investigate the online kernel-based multimodal similarity learning, which aims to integrate multiple kernels for learning multimodal similarity functions, and conduct experiments to evaluate the performance of the proposed method for CBIR on several different image datasets. The experiment results are encouraging and verify the effectiveness and the superiority of the proposed method.

In this paper, we first study the average sum rate of sparse code multiple access (SCMA) systems, where a general scenario is considered under the assumption that the distances between the mobile users and the base station are not necessarily identical. Closed-form analytical results are derived for the average sum rate based on which an optimal factor graph matrix is designed for maximizing the capacity of the SCMA systems. Moreover, we propose a low-complexity iterative algorithm to facilitate the design of the optimal graph matrix. Finally, Monte Carlo simulations are provided to corroborate the accuracy of the theoretical results and the efficiency of the proposed iterative algorithm.

Manifold learning and feature selection have been widely studied in face recognition in the past two decades. This paper focuses on making use of the manifold structure of datasets for feature extraction and selection. We propose a novel method called Joint Sparse Locality Preserving Projections (JSLPP). In order to preserve the manifold structure of datasets, we first propose a manifold-based regression model by using a nearest-neighbor graph, then the \( L_{2,1} \)-norm regularization term is imposed on the model to perform feature selection. At last, an efficient iterative algorithm is designed to solve the sparse regression model. The convergence analysis and computational complexity analysis of the algorithm are presented. Experimental results on two face datasets indicate that JSLPP outperforms six classical and state-of-the-art dimensionality reduction algorithms.

Feature selection is a basic and important step in real applications, such as face recognition and image segmentation. In this paper, we propose a new weakly supervised multi-view feature selection method by utilizing pairwise constraints, i.e., the pairwise constraint-guided multi-view feature selection (PCFS for short) method. In this method, linear projections of all views and a consistent similarity graph with pairwise constraints are jointly optimized to learning discriminative projections. Meanwhile, the l2,0-norm-based row sparsity constraint is imposed on the concatenation of projections for discriminative feature selection. Then, an iterative algorithm with theoretically guaranteed convergence is developed for the optimization of PCFS. The performance of the proposed PCFS method was evaluated by comprehensive experiments on six benchmark datasets and applications on cancer clustering. The experimental results demonstrate that PCFS exhibited competitive performance in feature selection in comparison with related models.

Patent equivalents: Problems of retrieval and patent family tracing in databases

A cooperative actuator fault accommodation strategy for a team of linear time-invariant (LTI) multi-agent systems with a switching topology and directed communication network graph is studied in this paper. The faults could occur in more than one agent simultaneously and the fault severity is not needed to be precisely estimated. The proposed fault accommodation strategy is performed in two levels: the agent level fault recovery (ALFR) and the team level fault recovery (TLFR). Whenever a fault is detected, the recovery strategy is to locally recover faulty agents based on inaccurate estimates of the faults in the first step and to reconfigure the weights of the information flow graph in the second step. The proposed strategy is based on the sub-optimal solution of a bi-linear matrix inequality (BMI) and guarantees the consensus achievement of the team. The stability properties of the proposed controller and the recovery strategy are investigated based on Lyapunov analysis. The effectiveness of our proposed consensus algorithm is illustrated by performing numerical simulations for a team of ten agents and the performance of our strategy is compared with the centralized and decentralized fault recovery approaches in the literature.

Community search that finds query-dependent communities has been studied on various kinds of graphs. As one instance of community search, intimate-core group (community) search over a weighted graph is to find a connected <inline-formula><tex-math notation="LaTeX">$k$</tex-math></inline-formula> -core containing all query nodes with the smallest group weight. However, existing state-of-the-art methods start from the maximal <inline-formula><tex-math notation="LaTeX">$k$</tex-math></inline-formula> -core to refine an answer, which is practically inefficient for large networks. In this paper, we develop an efficient framework, called <u>l</u> ocal <u>e</u> xploration <u>k</u> -core <u>s</u> earch (LEKS), to find intimate-core groups in graphs. We propose a small-weighted spanning tree to connect query nodes, and then expand the tree level by level to a connected <inline-formula><tex-math notation="LaTeX">$k$</tex-math></inline-formula> -core, which is finally refined as an intimate-core group. In addition, to support the intimate group search over large weighted graphs, we develop a weighted-core index (WC-index) and two new WC-index-based algorithms for expansion and refinement phases in LEKS. Specifically, we propose a WC-index-based expansion to efficiently find a candidate graph of intimate-core group, leveraging on a two-level expansion of <inline-formula><tex-math notation="LaTeX">$k$</tex-math></inline-formula> -breadth and 1-depth. We propose two graph removal strategies: coarse-grained refinement is designed for large graphs to delete a batch of nodes in a few iterations; fine-grained refinement is designed for small graphs to remove nodes carefully and achieve high-quality answers. Extensive experiments on real-life networks with ground-truth communities validate the effectiveness and efficiency of our proposed methods.

Incomplete multi-view clustering via kernelized graph learning

In this paper, we focus on the problem of image retrieval in trademark database and concentrate on the method for efficient trademark image retrieval based on new shape feature without complicated image segmentation. The proposed trademark image retrieval scheme comprises two stages: first, edge detection based on wavelet transform is performed on the trademark image, second, novel wavelet -based shape features are introduced to reflect the edge's characteristic. The new features contain four directionality information of the trademark edge, namely 0, pi/4 , 3pi/4 and pi. We used the shape features to indexing the image for the trademark image retrieval. In addition, one of the merit of the scheme is that the two stage are both used wavelet transform, this can quicken the retrieval efficiency. A wide range of trademark images for the experiments have performed, Experimental results show the method is promising.

Barycentric coordinate-based distributed localization for wireless sensor networks subject to random lossy links

This paper considers the fundamental power allocation problem in cell-free massive mutiple-input and multiple-output (MIMO) systems which aims at maximizing the total energy efficiency (EE) under a sum power constraint at each access point (AP) and a quality-of-service (QoS) constraint at each user. Existing solutions for this optimization problem are based on solving a sequence of second-order cone programs (SOCPs), whose computational complexity scales dramatically with the network size. Therefore, they are not implementable for practical large-scale cell-free massive MIMO systems. To tackle this issue, we propose an iterative power control algorithm based on the frame work of an accelerated projected gradient (APG) method. In particular, each iteration of the proposed method is done by simple closed-form expressions, where a penalty method is applied to bring constraints into the objective in the form of penalty functions. Finally, the convergence of the proposed algorithm is analytically proved and numerically compared to the known solution based on SOCP. Simulations results demonstrate that our proposed power control algorithm can achieve the same EE as the existing SOCPs-based method, but more importantly, its run time is much lower (one to two orders of magnitude reduction in run time, compared to the SOCPs-based approaches).