Search

Selecting appropriate measurement to characterize the connection strength is of great significance to construct epileptic brain network for the purpose of brain disease course recognition. In this paper, Kullback–Leibler (KL) divergence is introduced as an effective measure to describe the similarity between EEG channels, and the network based on KL divergence is used to identify the significant differences of special channels and detect seizures. First, the KL divergence is calculated based on the empirical cumulative distribution function, and the inverse divergence connecting matrix is constructed according to the symmetrized KL divergence. Then the node strength, weighted clustering coefficient, weighted characteristic path length and small-world property of morphometric similarity brain network are calculated based on the connecting matrix. Finally, Wilcoxon signed-rank significance test is adopted to analyze the differences between the different periods of epilepsy for the channel from different perspectives. The experiment is conducted on the original EEG signal and EEG signals under different frequency bands, and different complex network measures are calculated respectively. At the significance level of 0.05, it is found that the channels with significant differences during seizures are FP1-F7, T7-P7, F4-C4, FP2-F8 and P8-O2 for CHB-MIT dataset. In addition, the dynamic changes of node strength of each channel, mean weighted clustering coefficient and mean weighted characteristic path length over time are visualized to explore the significant changes in the long course of epileptic evolution, and the significant changes are found before the onset of the seizure. [ABSTRACT FROM AUTHOR]

In this paper, a type of delayed Clifford-valued memristor-based neural networks (CLVMNNs) with impulsive disturbances is established, and the global exponential stability is investigated by using generalized norm. Firstly, the n-dimensional Clifford-valued systems are decomposed into 2 m n -dimensional real-valued systems to address the non-commutativity problem of the multiplication of Clifford numbers. Secondly, the generalized ∞-norm and 1-norm are introduced to induce the global exponential stability for CLVMNNs, and two special Lyapunov functionals are established to prove the stability. Thirdly, the strict assumption of the boundedness of activation function in previous research is loosened, and some less conservative conditions of stability are obtained based on the constructed Lyapunov functionals. Finally, the theoretical results are verified by two numerical simulations, and an image encryption scheme is proposed to show the application in real world situation based on the delayed CLVMNNs. • Impulsive disturbances and memristor are taken into account for Clifford-valued NNs. • The generalized { ξ , 1 } -norm and { ξ , ∞ } -norm are introduced to investigate the global exponential stability of CLVMNNs. • The global exponential stability of CLVMNNs is discussed and some new criteria are derived. [ABSTRACT FROM AUTHOR]

In this paper, a type of fractional-order quaternion-valued neural networks (FOQVNNs) with leakage and time-varying delays is established to simulate real-world situations, and the global Mittag-Leffler stability of the system is investigated by using the non-decomposition method. First, to avoid decomposing the system into two complex-valued systems or four real-valued systems, a new sign function for quaternion numbers is introduced based on the ones for real and complex numbers. And two novel lemmas for quaternion-valued sign function and Caputo fractional derivative are established in quaternion domain, which are used to investigate the stability of FOQVNNs. Second, a concise and flexible quaternion-valued state feedback controller is directly designed and a novel 1-norm Lyapunov function composed of the absolute values of real and imaginary parts is established. Then, based on the designed quaternion-valued state feedback controller and the proposed lemmas, some sufficient conditions are given to ensure the global Mittag-Leffler stability of the system. Finally, a numerical simulation is given to verify the theoretical results. [ABSTRACT FROM AUTHOR]

As multi-gene networks transmit signals and products by synchronous cooperation, investigating the synchronization of gene regulatory networks may help us to explore the biological rhythm and internal mechanisms at molecular and cellular levels. We aim to induce a type of fractional-order gene regulatory networks to synchronize at finite-time point by designing feedback controls. Firstly, a unique equilibrium point of the network is proved by applying the principle of contraction mapping. Secondly, some sufficient conditions for finite-time synchronization of fractional-order gene regulatory networks with time delay are explored based on two kinds of different control techniques and fractional Lyapunov function approach, and the corresponding setting time is estimated. Finally, some numerical examples are given to demonstrate the effectiveness of the theoretical results. [ABSTRACT FROM AUTHOR]

Extracting features from EEG signals through time, frequency and spatial-domain gives rise to the problem of neglecting the property of nonlinear manifold structure of the data, and traditional feature extractions based on Euclidean space are unfavorable for the full utilization of the manifold feature. In this paper, a new EEG classification method is proposed by representing EEG signals on the Grassmann manifold. Firstly, a low rank representation method based on matrix recovery is introduced into the Grassmann manifold to explore the subspace information and re-express the EEG signals. Then a deep neural network is proposed to reduce the dimension of the recovered EEG data, and the low rank representation makes the EEG data separable by using T-SNE visualization; Finally, the dimension reduced features are input to the traditional classifier for classification. The proposed method is performed on four datasets for seizure detection, disease recognition and mental arithmetic recognition respectively. The corresponding accuracy reaches 99.99%, 99.61%, 100%, and 99.72% which shows that the proposed method outperforms other state of art methods on EEG classification. • An EEG signal representation is proposed using Grassmann manifold. • Low rank representation is introduced to the manifold for the signal recovery. • The feature representation of four datasets are used for different purposes. [ABSTRACT FROM AUTHOR]

In this paper, a Leslie-type predator–prey system with simplified Holling type IV functional response and strong Allee effect on prey is proposed. The dissipativity of the system and the existence of all possible equilibria are investigated. The investigation emphasizes the exploring of bifurcation. It is shown that the system exists several non-hyperbolic positive equilibria, such as a weak focus of multiplicities one and two, (degenerate) saddle–nodes and Bogdanov–Takens singularities (cusp case) of codimensions 2 and 3. At these equilibria, it is proved that the system undergoes various kinds of bifurcations, such as saddle–node bifurcation, Hopf bifurcation, degenerate Hopf bifurcation and Bogdanov–Takens bifurcation of codimensions 2 and 3. With the parameters selected properly, there exhibits a limit cycle, a homoclinic loop, two limit cycles, a semistable limit cycle, or the simultaneous occurrence of a homoclinic loop and a limit cycle in the system. Moreover, it is also proved that the system has a cusp of codimension at least 4. Hence, there may exist three limit cycles generated from Hopf bifurcation of codimension 3. Numerical simulations are done to support the theoretical results. [ABSTRACT FROM AUTHOR]

Abstract: ZnO was doped into coprecipitated Ni-Al2O3 catalyst to promote the activity and the stability in methane decomposition to hydrogen and carbon nanofibers. The promoting effects were examined with XRD, TPR, XPS and TEM, using an in situ thermal balance reactor and a tubular fixed-bed reactor. The results showed that there was a strong interaction between Ni-Al2O3 and the doped ZnO, which may result in the formation of ZnAl2O4 spinel-like structure. The doping of ZnO could improve both the activity and the stability of nickel particles in methane decomposition. Such promotion effects became more pronounced with the increase of ZnO content. The evolution of the morphologies of the carbon produced and of the catalyst particles with the reaction temperature suggested that the doping of ZnO may delay the appearance of the quasi-liquid state of the catalyst particles to the range of higher temperatures and may weaken the interfacial wetting effect between catalyst particles and the growing carbon layers to delay the encapsulation process of the catalyst particles. [Copyright &y& Elsevier]

The dynamical behaviors of a Leslie type predator–prey system are explored when the functional response is increasing for both predator and prey. Qualitative and quantitative analysis methods based on stability theory, bifurcation theory and numerical simulation are adopted. It is showed that the system is dissipative and permanent, and its solutions are bounded. Global stability of the unique positive equilibrium is investigated by constructing Dulac function and applying Poincaré–Bendixson theorem. The bifurcation behaviors are further explored and the number of limit cycles is determined. By calculating the first Lyapunov number and the first two focus values, it is proved that the positive equilibrium is not a center but a weak focus of multiplicity at most two, so the system undergoes Hopf bifurcation and Bautin bifurcation. The normal form of Bautin bifurcation is also obtained by introducing the complex system. Moreover, numerical simulations are run to demonstrate the validity of theoretical results. • An improved predator–prey model is constructed by introducing a functional response increasing in both predator and prey. • Global stability is investigated by constructing Dulac function and applying Poincaré–Bendixson theorem. • Hopf bifurcation and Bautin bifurcation are demonstrated by calculating the first Liapunov number and the first two focal values. • It is proved that the unique positive equilibrium cannot be a center and is a weak focus of multiplicity at most two. [ABSTRACT FROM AUTHOR]

Sleep staging is essential in assessing sleep quality and diagnosing sleep-related disorders, but the lack of labeled data impedes the development of automatic sleep staging models. Generally, institutions rely on semi-supervised approaches to enhance the utilization of their own unlabeled data. However, the task knowledge obtained from a limited amount of labeled data is often insufficient to guide the learning based on large amounts of unlabeled data, which may even lead to catastrophic forgetting and further degrade the performance of most existing methods. In this paper, we propose a novel strategy of building secure collaboration among multiple institutions, to achieve the implicit augmentation of labeled data and expansion of task knowledge for each participating institution by acquiring external knowledge from others. We adopt the Federated Learning (FL) to facilitate secure collaboration and propose a federated semi-supervised sleep staging method based on knowledge sharing, which enables the automatic scoring of sleep stages using only single-channel EEG data. The task knowledge in our method is contained in relationships, which exist naturally among sleep stages and can be extracted from both local labeled and unlabeled data. Furthermore, the knowledge sharing among participating institutions can be achieved by aligning the local relationships to the aggregated global relationships. Additionally, we employ prototype-contrastive learning to enhance the clarity of relationships extracted from labeled data, and propose pseudo-labeling optimization to generate reliable pseudo-labels for subsequent relationship extraction from unlabeled data. Our method is shown to be effective and outperforms compared methods in extensive experiments conducted on two publicly available datasets. • A federated semi-supervised method is used to address the issue of insufficient labeled data. • Prototype-contrastive learning is employed to establish more distinct relationships among sleep stages. • A pseudo-labeling optimization is introduced to generate reliable pseudo-labels. [ABSTRACT FROM AUTHOR]

Many successful models of saliency have been proposed to detect salient regions for 2D images. Because stereopsis, with its distinctive depth information, influences human viewing, it is necessary for stereoscopic saliency detection to consider depth information as an additional cue. In this paper, we propose a 3D stereoscopic saliency model based on both contrast and depth-guided-background prior. First, a depth-guided-background prior is specifically detected from a disparity map apart from the conventional prior, assuming boundary super-pixels as background. Then, saliency based on disparity with the help of the proposed prior is proposed to prioritize the contrasts among super-pixels. In addition, a scheme to combine the contrast of disparity and the contrast of color is presented. Finally, 2D spatial dissimilarity features are further employed to refine the saliency map. Experimental results on the PSU stereo saliency benchmark dataset (SSB) show that the proposed method performs better than existing saliency models. [ABSTRACT FROM AUTHOR]

Abstract: The region growing pulse coupled neural network (PCNN) algorithm is an efficient method for multi-value image segmentation. However, as a kind of PCNN models, choosing appropriate parameters are usually difficult. This paper brings forward a new approach which improves the region growing PCNN model by modifying the linking channel function and decreases the complexity of adjusting parameters. The region growing PCNN is not effective when processing the edge pixels between different regions because the edge pixels and central pixels are dealt with unfairly. In order to overcome this disadvantage, the proposed method processes the edge pixels by setting the edge pixels and central pixels to receive same linking input if they are in similar condition. Computer simulations prove it can process the edge pixels efficiently and obtain clear boundaries between different regions. [Copyright &y& Elsevier]

In this paper, we propose a deep multimodal feature learning (DMFL) network for RGB-D salient object detection. The color and depth features are firstly extracted from low level to high level feature using CNN. Then the features at the high layer are shared and concatenated to construct joint feature representation of multi-modalities. The fused features are embedded to a high dimension metric space to express the salient and non-salient parts. And also a new objective function, consisting of cross-entropy and metric loss, is proposed to optimize the model. Both pixel and attribute level discriminative features are learned for semantical grouping to detect the salient objects. Experimental results show that the proposed model achieves promising performance and has about 1% to 2% improvement to conventional methods. [ABSTRACT FROM AUTHOR]

• We propose the CAN architecture to learn discriminative feature representations for saliency detection in RGB D images, by modeling multi modal and multi scale context dependencies within the context aware fusion and context dependent deconvolution. It is demonstrated that the proposed end to end CAN can a chieve favourable performance compared with state of the art methods. • A context aware fusion unit based on the LSTM architecture (MCFLSTM) is developed to learn complementary contexts from two modalities. The positive effect of this fusion approach is demonstrated experimentally. • A hierarchical LSTM structure called HSCLSTM is proposed to progressively refine saliency cues by modelling the context dependencies among different scales. Its effectiveness is also verified by experimental results. Convolutional neural networks (CNNs) have shown unprecedented success in object representation and detection. Nevertheless, CNNs lack the capability to model context dependencies among objects, which are crucial for salient object detection. As the long short-term memory (LSTM) is advantageous in propagating information, in this paper, we propose two variant LSTM units for the exploration of contextual dependencies. By incorporating these units, we present a context-aware network (CAN) to detect salient objects in RGB-D images. The proposed model consists of three components: feature extraction, context fusion of multiple modalities and context-dependent deconvolution. The first component is responsible for extracting hierarchical features in color and depth images using CNNs, respectively. The second component fuses high-level features by a variant LSTM to model multi-modal spatial dependencies in contexts. The third component, embedded with another variant LSTM, models local hierarchical context dependencies of the fused features at multi-scales. Experimental results on two public benchmark datasets show that the proposed CAN can achieve state-of-the-art performance for RGB-D stereoscopic salient object detection. [ABSTRACT FROM AUTHOR]

• We develop a coupled CNN (CoCNN) architecture and present a systematic way to fuse multimodal and multiscale features for stereoscopic salient object detection. • We propose a cascaded span network (CSN) that leverages a constrained span unit to fuse disparity and color cues and serves as a simple yet effective module for multimodal information fusion. This fusion method is able to mitigate the ambiguity of features from different modalities and enables the exploration of feature complementarity and correlations across different layers. • The experimental results demonstrate that the proposed model outperforms other approaches. We verify the effectiveness of the proposed CSN through extensive experiments. Many convolutional neural network (CNN)-based approaches for stereoscopic salient object detection involve fusing either low-level or high-level features from the color and disparity channels. The former method generally produces incomplete objects, whereas the latter tends to blur object boundaries. In this paper, a coupled CNN (CoCNN) is proposed to fuse color and disparity features from low to high layers in a unified deep model. It consists of three parts: two parallel multilinear span networks, a cascaded span network and a conditional random field module. We first apply the multilinear span network to compute multiscale saliency predictions based on RGB and disparity individually. Each prediction, learned under separate supervision, utilizes the multilevel features extracted by the multilinear span network. Second, a proposed cascaded span network, under deep supervision, is designed as a coupling unit to fuse the two feature streams at each scale and integrate all fused features in a supervised manner to construct a saliency map. Finally, we formulate a constraint in the form of a conditional random field model to refine the saliency map based on the a priori assumption that objects with similar saliency values have similar colors and disparities. Experiments conducted on two commonly used datasets demonstrate that the proposed method outperforms previous state-of-the-art methods. [ABSTRACT FROM AUTHOR]