Search

The key to network virtualization technology is virtual network mapping, which has been proven to be an NP-hard problem. At present, the methods to solve the problem of virtual network mapping still have the following defects. Most of the existing literature is limited to static virtual network (VN) mapping and static linear resource pricing, which rely on peak allocation and don't meet the user dynamic resource requirements. Therefore, this paper proposes a virtual network resource allocation model based on dynamic resource pricing named GSO-RBFDM. Firstly, group search optimization (GSO) is used to optimize the node mapping scheme during the network mapping process to reduce the cost of network mapping. Secondly, a dynamic nonlinear resource pricing model is established, and genetic algorithm (GA) is used to more accurately search a low-cost network mapping path instead of the traditional Dijkstra algorithm. Finally, virtual network dynamic modeling is performed according to the user dynamic resource requirements, and radial basis function (RBF) is used to predict resource requirements to realize the dynamic resource allocation to users. Simulation results show that, compared with traditional virtual network mapping algorithms, GSO-RBFDM can not only realize dynamic resource allocation, but also show good performance in terms of acceptance rate, network cost, link pressure and average network revenue.

Depression has become a serious disease that affects people’s mental health. How to detect it promptly and accurately is a difficult task. Electroencephalogram can reflect the spontaneous biological potential signals in the cerebral cortex and is widely used in the prediction and diagnosis of depression. With electroencephalogram, the key and most difficult challenge is to find the brain regions and frequencies associated with depression, especially mild depression. At present, the most commonly used method is the combination of feature selection and classification algorithm for detection. However, the classification accuracy needs to be further improved. The differential evolution is a population-based adaptive global optimization algorithm. Due to its fast convergence and strong robustness, this paper uses it to optimize the extracted features to achieve better result. Then the k-nearest neighbor classification algorithm is used to classify patients with mild depression and normal people. The experiment is performed on a data set of 10 subjects with mild depression and 10 normal subjects. The results show that the method can find the relatively optimal features and distinguish the two groups of subjects better. It effectively improves the classification accuracy and efficiency, and is superior to other feature optimization methods.

Electronic health records (EHRs) data plays an important role in the development of healthcare undertaking. There are many challenges in mining EHRs, such as temporality, irregularity, sparsity, bias, etc. Thus effective feature extraction and representation are key steps before any further applications. In this paper, we propose a multi-task deep representation learning method (MTDRL) with the objective of extracting the valuable clinical information from raw data and learning an effective and interpretable patient representation. Firstly, MTDRL utilizes Bidirectional Gated Recurrent Unit (BiGRU) as an encoder to learn the hidden state vectors which are used as inputs for the following networks. This encoding part can be seen as a shared network for all tasks. Secondly, patient’s in-hospital mortality prediction and sequence reconstruction are simultaneously conducted based on the encoding network. Specifically, an attention mechanism and a fully-connected layer are incorporated in prediction task and BiGRU is implemented in the other task to reconstruct the visit sequences. Finally, we apply MTDRL to real EHRs data and the experimental results demonstrate that MTDRL is capable of learning more effective patient representation and has a significant improvement in the performance of patient’s in-hospital mortality prediction. Meanwhile, the prediction results can be effectively interpreted with the attention mechanism and provide a clinically meaningful references.

Emotion plays an indispensable role in the process of human cognition and decision-making, but it is often neglected in human-computer interaction (HCI). Although many researchers have applied some decomposition algorithms to extract features from EEG, there are several shortcomings including the modal aliasing, a high computational cost and so on. In this paper, we introduce variational mode decomposition (VMD) and wavelet packet decomposition (WPD) algorithms and propose a multidimensional feature extraction method based on VMD and WPD for emotion recognition. Firstly, we apply VMD to decompose the EEG signal into a specific number of variational mode functions (VMF). Secondly, WPD is executed to generate an emotional frequency band. Then, we continue to extract the wavelet packet entropy (WPE), modified multi-scale sample entropy (MMSE), fractal dimension (FD) and first difference (1ST) of each emotional VMF to construct a new feature form. Finally, the random forest (RF) is utilized to classify the extracted emotional states. The experiment results demonstrate that the proposed method is more competitive and universal for emotion recognition, which provides a novel idea for the application of multidimensional feature extraction.

Although more and more researchers pay attention to the emotion classification, traditional emotion classification methods can not embrace changes in the global and local areas of the human brain after being stimulated. We propose an emotion classification method based on SVM combining brain functional connectivity. Firstly, the nonlinear phase-locked value (PLV) is used to calculate the multiband brain functional connectivity network, which is then converted into a binary brain network, and seven features of binary brain network are calculated. Secondly, support vector machines (SVM) are used to classify positive and negative emotions at the valence dimension and arousal dimension in the multiband. Experimental results on DEAP show that the best emotion classification accuracy of the proposed method is 86.67% in the arousal dimension, and 84.44% in the valence dimension. The results demonstrate that the classification accuracy of the arousal dimension is better than the valence dimension and the Beta2 frequency band is more suitable for emotion classification. Finally, several findings on brain functional connectivity network is discussed. The left and right areas of brain functional connectivity network are unbalanced in the low frequency band, and the feature values of clustering coefficient, average shortest path length, global efficiency, local efficiency, node degree are positively correlated with the arousal degree in the arousal dimension. Humans emotions are suppressed in the low frequency band, and the brain functional connectivity network after emotional stimulation is strengthened in the high frequency band. Our findings on emotion classification are valuable and consistent with the study of neural mechanisms.

Electronic health records are digital records of patients' medical history, diagnosis, medication, treatment plans. EHRs not only contain the patients' medical and treatment history, but also systematically collect patients' clinical data. Therefore, it is very valuable to improve the patient's health care management by mining the information in the EHRs. However, due to the irregularities and sparsity of EHRs, EHRs mining is very challenging. In this paper, the laboratory data, physiological indicators and diagnosis time during the patients' hospitalization period are extracted from the MIMIC-III database. Then, the extracted features are used to generate the patients' representation vector. Finally, we propose a prediction model based on BiLSTM and attention mechanism, which is called Bi-Attention. The BiLSTM is adopted to learn the forward and backward timing information in the patient's representation vectors and to predict the patient's disease by utilizing the specific clinical information in the timed medical record with the attention mechanism. The experimental results show that compared with other methods, the proposed model can effectively improve the prediction performance.

This paper proposes an adaptive processing method and a dynamic storage method of pulsed wave signal based on wireless body area network. Firstly, the ZigBee star network model is adopted as the basic network architecture of the system. Secondly, an adaptive Gaussian filter model is designed to eliminate the influence of environmental and man-made factors. Finally, the dynamic storage algorithm is used to map the pulsed wave signal to a new stored sample, and the regression model is used to restore the new sample to the original signal. Simulation experiments show that: 1) The adaptive Gaussian model minimizes the signal noise and better maintains the characteristics of the pulsed wave signal. 2) On the premise of using ARIMA regression model of neural network, dynamic storage algorithm improves the regression accuracy compared with fixed storage algorithm.

A novel color watermarking algorithm based on digital signature is proposed. First, the color host image is split into red, green and blue components. Discrete Wavelet Transform level-3 is performed on blue component to obtain various sub-bands. Next, the cD2 sub-band is decomposed by Singular value decomposition, and the watermark information is embedded into the singular values. Then, the digital signature of the orthogonal matrices is generated, and the digital signature is embedded into subbands cH3 and cV3. Last, a set of common attacks are imposed on the watermarked image to test the performance of the new algorithm. The experimental results show that the new algorithm obtains outstanding imperceptibility and robustness compared to the previous algorithms. Meanwhile, the security of the new algorithm catches a higher level by combing digital signature.

Subliminal stimulation can trigger emotional responses and this paper analyzes their relationship based on EEG signals. According to the nonlinear characteristics of EEG signals and based on the multi-scale sample entropy, we propose a method to recognize unconscious emotion. We extract the multi-scale sample entropy of EEG signals when angry or happy face pictures are subliminally presented as eigenvalues. Then, we compute the distribution and regularity of eigenvalues. Experimental results demonstrate that the multi-scale sample entropy can be used to distinguish different emotions when different emotional pictures are subliminally presented, namely, the multi-scale sample entropy is larger when angry face pictures are subliminally presented than happy face pictures. Furthermore, the p-values of extracted features are calculated by the method of Kolmogorov-Smirnov test (KS test). The result shows that the EEG signals have significant difference $(\mathrm{p}\lt 0.05)$ for the two different emotional states and the method is effective for unconscious emotion recognition.

When the social force model is used to simulate the crowd evacuation, how to reduce the amount of calculation of the original model is an important problem which can not only affect the speed of solving the model but also control the program running time. This paper investigates the impact that the number of each team’s members and the scale of crowd have on the evacuation time and the program running time in rooms with different exit width. Therefore, it proposes the grouping social force model, in which the crowd is to divide into multiple groups and each group is considered as a whole moving unit to decrease the computation time of the social force model. The simulation results show that evacuation time and programs run time will decrease with declining of the groups’ number, and the amplitude of attenuation is dynamic changing. Meanwhile, the results suggest that the proper number of each team’s members should be set to a different width exit in a regular rectangle room, and for large scale crowd, the effect of the grouping strategy is more effective.

Energy efficiency is a key issue for wireless sensor nodes, especially for wireless body area networks (WBANs) that operate near the human body or in the human body. Aiming at the problem that WBAN system still has too fast energy consumption, we propose a ZigBee star network model with multiple sensing nodes as end nodes, and design an adaptive transmission power correction control algorithm with adjustment factors to select the appropriate transmit power to reduce the energy consumption. Experiments show that the proposed power control algorithm reduces the overall energy consumption by reasonably controlling the transmission power in the ZigBee star network model.

With the emergence of network virtualization, the infrastructure can be effectively integrated to overcome the "ossification" of the Internet. The biggest challenge in network virtualization is the problem of virtual network embedding. Unfortunately, most of the researches on virtual network embedding only focus on static algorithms, which allocate fixed or invariable resources to virtual networks until the end of their lifetime. However, the demand of virtual requests for resources are dynamically changing and fluctuating in reality. Therefore, the traditional static schemes not only greatly reduce the utilization of substrate resources, but also decrease the revenue of the service providers. In this paper, we aim to satisfy the dynamic requirement of resources for virtual networks. We propose a dynamic embedding algorithm (RBF-VNE) which is based on RBF neural network to learn and predict the dynamic changes of resources, and then we dynamically adjust and allocate resources according to the predicted results. Simulation results show that our approach can well integrate substrate resources to solve above problems and performs well than static embedding algorithms.

Network virtualization is a dynamic context leading to imbalance for substrate network (SN) resources which can affect the virtual network (VN) request further. In this paper, we propose the GRC-mopso-VNE algorithm to solve the VNE problem taking link resources into consideration in the stage of node mapping. A novel metric is developed, i.e., global resource capacity (GRC), to evaluate the mapping potential of each node. In the mapping process, we consider the vector GRC to accomplish node mapping and adopt the shortest path routing at link mapping. When virtual node request is rejected using this method, we take the multi-objective particle swarm optimization (MOPSO) algorithm into account at node mapping striving for the minimization of resource consumption and load balancing. Through the simulation, the proposed algorithm demonstrates better performance in terms of the VN acceptance ratio, long-term average cost and revenue comparing to G-SP, D-ViNE, R-ViNE and the GRC-VNE that only considers the GRC.

Network virtualization plays an important role in the development of the network because of its dynamics and flexibility on network infrastructure configuration. Virtual network mapping is the main method to realize the network virtualization. Most of the current virtual network mappings always allocate network resources in an exclusive and excessive way. For example, the entire bandwidth amount of the virtual network will be allocated according to its peak demand of the traffic. However, the truth is that the actual traffic needs of virtual network are changing constantly throughout its lifetime, and the distribution of static mapping will inevitably lead to the underutilization of the assigned resource, high user cost as well as low carrier revenue. In order to solve the above problems, we need to predict the changes of virtual network's demands accurately and adjust the allocation of resources dynamically. In this paper, we propose a virtual network mapping method based on support vector machine (SVM) to dynamically allocate and adjust the network resources. In addition, to improve the accuracy of regression forecasting, the relatively better prediction parameters are selected in the proposed method. Experimental results show that the proposed embedding method can make full use of resources, improve the acceptance rate of the virtual networks, and increase the revenue of the operators significantly.

Video delivery networks (VDNs) using cloud storage have recently started to emerge. Different to the conventional VDNs, Cloud Video Delivery Networks (CVDNs) take advantage of the cloud computing technology which helps to reduce the storage and bandwidth cost. However, the existing CVDNs model cannot guarantee the Quality of Service (QoS) adequately because of the dynamic demands of end users such as the delay and the priority. In this paper, we put forward a path-splitting-supported and priority-oriented CVDNs method to increase the acceptance ratio of requests and provide the QoS of end users. In the proposed method, we divide the end users into ordinary users and paying users while the paying users having higher priorities. Then, we formulate the priority-oriented cloud video delivery problem as a mixed integer programming (MIP) problem while considering the routing constraints. The experimental results show that the acceptance rate of path-splitting-supported method is 20% higher than the method of path-unsplitting. In addition, our method can also make CVDNs providers get higher earnings.

In the process of adopting network virtualization method to solve the network ossification problem, reconfiguration can effectively improve the acceptance ratio of virtual network requests and the resource utilization of substrate network. Energy consumption as an important evaluation factors, and combined with a reconfiguration algorithm, this paper propose energy-aware reconfiguration algorithm. The target physical nodes are chosen on the constraint of the minimum energy consumption, after then, the virtual node will be migrated to the target physical node, and then the virtual links that need be migrated will be remapped to the physical links by shortest path algorithm. This method can effectively improve the virtual network request acceptance ratio and reduce the energy consumption of the substrate network after reconfiguration. The results of simulation verify the effectiveness of this algorithm.

Virtual network embedding problem is a multi-objective NP-hard problem. Various mapping objectives are often mutually restricted with each other. The existing methods are mostly converting them into a single-objective optimization, which make it difficult to balance the mapping objectives. Based on the concept of multi-objective optimization, a new virtual network embedding model is proposed in this paper. In order to balance mutual constraints of the objectives, two objective functions are defined and optimized simultaneously. Multi-objective group search optimizer is used to achieve the Pareto-optimal set, in which, the ranger searching strategy is improved to optimize the performance of the algorithm. Experimental results demonstrate that the proposed method owns good performance in terms of the revenue and acceptance ratio.

Currently, energy cost has become an important burden of infrastructure providers in network virtual environment. Therefore, consolidating nodes and links to save energy is an effective method when virtual networks are embedded on substrate network. The traditional embedding method usually considers cost minimization or revenue maximization as its goal. This paper studies the problem of energy consumption, cost and revenue simultaneously and a virtual network embedding model saving energy is proposed. Then, a heuristic algorithm base on group search optimizer is developed. The simulation experiments show that the proposal can effectively reduce energy consumption when a large quantity of virtual networks arrives.

Nowadays, mining approximate frequent itemsets from noisy data has attracted much attention in real applications. However, there is not widely accepted algorithm at present to solve the problem under noisy databases, which dues to two key issues. Firstly, the anti-monotonicity property does not hold which is used to prune candidate itemsets efficiently. And secondly, the computation of support counting turns out to be NP-hard. In this paper, we propose a novel model which is based on rough set theory and capable to recover the noise-tolerant frequent itemsets from "reduced itemsets". The novel model applies depth-first growing method to generate candidate itemsets and exerts effective pruning strategies, which narrows the searching space and mines indeed meaningful noise-tolerant frequent itemsets efficiently.

To meet the needs of innovative sensor network applications, sensor nodes have long evolved from underpowered single microcontroller designs into complex architectures that accommodate multiple processors and Field Programmable Gate Arrays (FPGAs). We address the problem of conceiving and implementing programs for such sensor node architectures. We rely on a universal, layered hardware/software interface that provides seamless interconnection between tasks running on a micro-controller and tasks running on a FPGA. Resource sharing is handled transparently through a shared-bus communication architecture. We demonstrate our methodology, through a heterogeneous sensor node simulator called SUNSHINE [1] for an application running the sensor nodes. We validate SUNSHINE by demonstrating the applications on a multiprocessor sensor node's testbed, which consists of a FPGA, a microcontroller and a radio front-end.

Nowadays, frequent item set mining is the major task in association rule mining. With the observation that the support plays an important role in mining frequent item sets, in this paper, we review the previous efficient algorithms and study the effect of different order of support in the performance of frequent item sets mining algorithms and propose our improved schedule. In our new algorithm, items are sorted in descending order according to the frequencies in transaction cache while item sets use ascending order of support during support count. Compared with other algorithms, the results of experiments show that the new algorithm gains better performance on on well-known benchmark data sets.

Wireless distributed computing (WDC) enables the radio nodes with reduced computing abilities to cooperate in processing complex computational tasks for minimizing the overall processing latency (makespan). However, the uncertainty of the dynamic mobile wireless environment, which is not an issue for the traditional distributed computing, poses a challenge for WDC. In this paper, a variation-aware approach for WDC is proposed to determine the task allocation by considering the heterogeneous computing capability of the radio nodes as well as the impact of the radio environment. To this end, the transmission latency is characterized as a random variable that depends on the channel fading and the transport protocol. A variation-aware task graph analysis is proposed for the estimation of the makespan’s distribution. The evolutionary algorithms are employed for the allocation mapping. We use simulation results to affirm makespan estimation improvement of the proposed approach compared with the traditional deterministic approach, and give insights on the dominating factors for the improvement.

With the development of Internet, e-learning becomes a new modern teaching mode. Educational evaluation is an important part of network teaching system and an important means to improve the teaching quality. PSK-means algorithm is used in teaching quality evaluation system of computer experiments. Teachers and students are classified and analyzed according to their evaluation scores. Based on the evaluation results, network teaching and learning of computer experiments will be improved by teachers and students.

More and more clustering approaches are used in educational data mining for too many data are generated by web-based educational systems. To deal with the problem of premature convergence of the traditional K-means algorithm and computational costs, an improved K-means clustering algorithm based cooperative PSO frame is proposed in this paper. Cooperative PSO has been proved effective for large scale and complex problems via a divide-and-conquer strategy by simulating coevolutionary techniques in nature. Therefore, K-means clustering algorithm based cooperative PSO frame is effective in educational data clustering.

Many practical applications are dynamic over time, which require optimization algorithms not only to converge to optimum as soon as possible but also to track the changing optimum. In this paper, a Cooperative Dual-swarm PSO (CDPSO) is proposed to deal with dynamic optimization problems. CDPSO adopts dual-swarm structure to keep swarm diversity and track the changing optimum. Fractional Global Best Formation technique is employed to construct artificial global bests which are potential to be better. Also an adaptive mutation operator is designed to maintain particle diversity. The experiments demonstrate that the proposed algorithm is effective and stable in dynamic environment.

Character Modeling is becoming more and more difficult in animation industry today. Lots of designers are usually involved to cooperatively accomplish a character by computer networks or the Internet. This paper presents a Multi-Objective evaluation based Cooperative Character Modeling System(MOCMS), which can evolve various character models to generate creative ones based on multi-objective evaluations. The objectives are designed to embody different personalities, including qualitative and quantitative aspects. The former are given by different cooperative designers while the latter are calculated automatically by computers. This can incorporate qualitative and quantitative evaluation in a formal manner. Case study demonstrates that the proposed method can evolve character models according to the designers' intentions and shorten design time.

Many applications can be formalized as dynamic problems and their main goals are to track the moving optimum. Many algorithms that address efficiently static problems, fail when applied to dynamic problems due to their inability to adapt and respond to changes in the environment. A different topology multi-swarm PSO is proposed to deal with problems in dynamic environment. The proposed algorithm integrates two different topology subswarms and the two subswarms exchange their best particle at checkpoints. One rapidly convergence to the optima and the other is responsible for local search and diversity reservation. The experiments demonstrate the proposed algorithm is effective and stable in dynamic environment.

Most existing evolutionary algorithms are only effective for optimization problems with no more than one hundred decision variables. However, many optimization problems involve more than several hundreds decision variables, even one thousand. To deal with these problems, a fast cooperative coevolutionary particle swarm optimizer (FCPSO) is proposed, which is constructed based on cooperative coevolutionary framework and particle swarm optimizer with simple mutation operator. Simulation experiments on several benchmark functions from one hundred to one thousand decision variables are presented. FCPSO can solve these unimodal and multimodal benchmark functions while the function evaluations needed only linearly increase with decision variables. Therefore, FCPSO is a competitive optimizer for large scale and complex problems.

In this paper we suppose the formation and evolvement mechanism of city system is the site selection of special entities in the city. We construct model to simulate the mechanism based on multi-agent system, the experiment results show that the distribution of a virtual city system can be generated by the mechanism. Further, we prove that the generated distribution of city system is corresponding to that of a real city system by validating with the Zipf law. The work can help to build the virtual city model.

A cooperative coevolutioanry PSO is proposed in this paper, named CCPSO. In CCPSO, the simple mutation operator is incorporated and the number of subswarm is defined according to the relationship of decision variables. Collaborators are selected from each subswarm and used to construct individuals so as to evaluate the subswarm. Simulation experiments of CCPSO based on uniform distribution mutation and Gaussian distribution mutation on several benchmark functions are presented. It is shown that CCPSO can be used to optimize unimodal and multimodal benchmark functions and it's convergence rate and precision is competitive.

Component-based product design usually appears as a multi-objective optimization problem (MOP). Traditional methods solving MOPs are robust and have proven their effectiveness in handling many classes of optimization problems. However, such techniques can encounter difficulties such as getting trapped in local minima, increasing computational complexity, and not being applicable to certain classes of objective functions. Multi-Objective Evolutionary Algorithms (MOEAs) can overcome these disadvantages and have shown great potentials to solve MOPs. In this paper, an h-MOEA is proposed by employing effective strategies from evolutionary computation, which is suitable for solving the MOP in design optimization and can generate more diverse solutions in an accepted time span. Then, the effectiveness and correctness of h-MOEA is verified using several popular benchmark functions. Also, a prototype is developed and used in component-based product design optimization. Finally, the optimization results of a product design case are shown in detail.

Multi-objective collaborative design synthesis and its evaluation is usually viewed as a multi-objective optimization problem (MOP). In this paper, it is formulated firstly and then, a hybrid Multi-objective Evolutionary Algorithm (h-MOEA) is proposed by introducing ideas from evolutionary computation, which is suitable for solving the MOP in design synthesis and its evaluation. Furthermore, a design synthesis and its evaluation method supporting multi-objective collaborative design, composed of many iterative steps, is developed and the h-MOEA is encapsulated as a black-box optimization tool to generate design solutions, namely the Pareto optimal set. And also, a Web-based experimental prototype is developed to verify the proposed method. Finally, a case study has been done to evaluate the effectiveness of proposed methods.

Lots of geographically distributed designers are usually involved to accomplish a complex design task and creativity and amenity of artefacts are more and more emphasized, so researches on cooperative creative conceptual design system are valuable both theoretically and practically. This paper firstly identifies some major requirements of cooperative creative conceptual design system. Then, Creative Genetic Algorithm based on Simple Genetic Algorithm is proposed to serve as creative mechanisms and methods. The design and development of prototype TripleCDS is described and some typical design images are shown, so the creative mechanisms and methods and prototype are validated.