In this paper, we study the distributed synchronization and pinning of stochastic coupled neural networks via randomly occurring control. Two Bernoulli variables are used to describe occurrences adaptive control updating law according certain probabilities. Both for each vertex in a network depend on state information vertex's neighborhood. By constructing appropriate Lyapunov functions employing analysis techniques, prove that complex can be achieved mean square. Additionally, is compared...

Locally linear embedding (LLE) is one of the most well-known manifold learning methods. As representative extension LLE, orthogonal neighborhood preserving projection (ONPP) has attracted widespread attention in field dimensionality reduction. In this paper, a unified sparse framework proposed by introducing sparsity or L1-norm learning, which further extends LLE-based methods to cases. Theoretical connections between ONPP and are discovered. The optimal embeddings derived from can be...

Robustness to noises, outliers, and corruptions is an important issue in linear dimensionality reduction. Since the sample-specific outliers exist, class-special structure or local geometric destroyed, thus, many existing methods, including popular manifold learning- based fail achieve good performance recognition tasks. In this paper, we focus on unsupervised robust reduction corrupted data by introducing low-rank representation (LRR). Thus, a technique termed embedding (LRE) proposed which...

Low-rank representation (LRR) has been successfully applied in exploring the subspace structures of data. However, previous LRR-based semi-supervised clustering methods, label information is not used to guide affinity matrix construction so that cannot deliver strong discriminant information. Moreover, these methods guarantee an overall optimum since and are often independent steps. In this paper, we propose a robust method based on non-negative LRR (NNLRR) address problems. By combining...

Linear regression (LR) and some of its variants have been widely used for classification problems. Most these methods assume that during the learning phase, training samples can be exactly transformed into a strict binary label matrix, which has too little freedom to fit labels adequately. To address this problem, in paper, we propose novel regularized relaxation LR method, following notable characteristics. First, proposed method relaxes matrix slack variable by introducing nonnegative LR,...

Compact hash code learning has been widely applied to fast similarity search owing its significantly reduced storage and highly efficient query speed. However, it is still a challenging task learn discriminative binary codes for perfectly preserving the full pairwise similarities embedded in high-dimensional real-valued features, such that promising performance can be guaranteed. To overcome this difficulty, paper, we propose novel scalable supervised asymmetric hashing (SSAH) method, which...

Feature extraction plays a significant role in pattern recognition. Recently, many representation-based feature methods have been proposed and achieved successes applications. As an excellent unsupervised method, latent low-rank representation (LatLRR) has shown its power extracting salient features. However, LatLRR the following three disadvantages: 1) dimension of features obtained using cannot be reduced, which is not preferred extraction; 2) two matrices are separately learned so that...

This technical note deals with the design of mode-dependent H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> filters for a class discrete-time switched systems nonlinearities. In this systems, when system mode changes, filter designed specific subsystem also switches accordingly. The main contribution is on use information sojourn probability-the probability staying in each subsystem-to build new kind model additional available. Sojourn...

This paper proposes a novel method, called robust latent subspace learning (RLSL), for image classification. We formulate an RLSL problem as joint optimization over both the SL and classification model parameter predication, which simultaneously minimizes: 1) regression loss between learned data representation objective outputs 2) reconstruction error original inputs. The can be used bridge that is expected to seamlessly connect origin visual features their class labels hence improve overall...

Ridge regression (RR) and its extended versions are widely used as an effective feature extraction method in pattern recognition. However, the RR-based methods sensitive to variations of data can learn only limited number projections for To address these problems, we propose a new called robust discriminant (RDR) extraction. In order enhance robustness, L2,1-norm is basic metric proposed RDR. The designed objective function form be solved by iterative algorithm containing eigenfunction,...

Fine-grained attribute recognition is critical for fashion understanding, yet missing in existing professional and comprehensive datasets. In this paper, we present a large scale dataset with manual annotation high quality. To end, complex knowledge disassembled into mutually exclusive concepts form hierarchical structure to describe the cognitive process. Such well-structured reflected by terms of its clear definition precise annotation. The problems which are common process annotation,...

In this paper, stochastic synchronization is studied for complex networks with delayed coupling and mixed impulses. Mixed impulses are composed of desynchronizing synchronizing The term involves transmission delay self-feedback delay. By using the average impulsive interval approach comparison principle, several conditions derived to guarantee that exponential achieved in mean square. closely related strengths, frequency impulse occurrence, structure networks. Numerical simulations presented...