Design Of Network Structure Research Articles

A Study on Intelligent Obscured Face Image Completion System Based on GAN Network

Image recognition and processing is an important part of modern information technology. However, in real life, the face will inevitably be blocked by hats, masks, or face photos are damaged, scratches, and images are added mosaic and other reasons, resulting in face image masking, increasing face information recognition and processing difficulties. Image completion is a kind of technology to restore the missing information of obscured image. By supplementing the information of missing area, it can reduce the difficulty of image recognition and simplify the processing of face information.So it has important research significance for the restoration of shaded image. The traditional image completion technology produces the face of poor natural degree. At present, most of the popular completion methods combine depth learning, and the repair methods based on generating anti-network GAN are the representatives of them.But the stability of the network is poor, so it is difficult to train pictures with large sheltered area or poor image quality.Some methods are not effective in the completion of arbitrary shaded areas. Therefore, this study focuses on the improvement of the stability of the generated adversarial network and the maintenance of the global and local semantic consistency between the completion image and the real image. The main work includes: (1) Sets a special network structure of global discriminator and local discriminator. When the input is arbitrarily shaded, the completion image can maintain the consistency of global and local semantics. (2) The least square loss function is used to improve the stability of the GAN network. So it performs great when the input image has large area masking or low visualization degree. (3) ADAM algorithm is used to accelerate the training of neural network. (4) Based on the completion model, a human-computer interaction software is designed and developed, which includes data set selection, image selection, image completion, result evaluation and so on. The system is applied to the removal of facial structures such as wrinkles and spots. Compared with some existing software, the system can effectively remove the selected facial structures without image distortion. Based on the design of special network structure, the system improves the supervision ability of global and local completion, the stability of GAN network and enhances its training ability in large area masking. The system is intelligent and has strong completion ability for arbitrary masking and large area masking. The research results of this paper effectively restore the information loss caused by image masking and prepare favorable conditions for subsequent image recognition and processing.

Optimization of Ant Colony for Next Generation Wireless Cognitive Networks

In this work, the spectrum handoff process carried out by secondary users according to priority classes in cognitive radio networks is proposed. The network structure design of analytical and simulation models is simulated with the RIVERBED software. Moreover, deciding on the most suitable spectrum handoff process is determined by using artificial intelligence techniques. Using multi-parameter decision-making processes is the objective of our work. Optimization of the spectrum handoff number with the help of an ant colony algorithm is the main goal of our work. By studying similar aspects of cognitive radio networks and ant colony algorithms, the number of handoffs is reduced within the most appropriate scenario.

Research and Design of P2P-SIP System Network Structure Based on JXTA

The network structure of P2P-SIP system based on JXTA is one of the core technologies in industrial intelligent Internet.JXTA is an open peer-to-peer platform, which provides basic application services for building P2P network. This paper chooses JXTA development platform to build P2P network, and then constructs SIP soft terminal on it. Firstly, it introduces the background knowledge, related research, P2P network technology, JXTA technology and SIP protocol. Then, the feasibility of p2p sip system is analyzed, including the combination of P2P and sip, network structure and framework structure. Then the routing mechanism of JXTA is deeply studied, including distributed index of shared resources, limited range traversal, dynamic view of sink node and loose and consistent distributed hash table (DHT). At the same time, this paper proposes an improvement strategy to overcome the shortcomings of JXTA routing mechanism such as limited range traversal and loose consistent DHT, that is, introducing the chord ring of consistent hash algorithm in the aggregation network layer. Based on the in-depth study of JXTA structured network and SIP protocol, a P2P SIP system is designed and implemented. The experimental results show that the system can maintain normal SIP communication without maintaining the central server.

A Multi-Objective Evolutionary Approach Based on Graph-in-Graph for Neural Architecture Search of Convolutional Neural Networks.

With the development of deep learning, the design of an appropriate network structure becomes fundamental. In recent years, the successful practice of Neural Architecture Search (NAS) has indicated that an automated design of the network structure can efficiently replace the design performed by human experts. Most NAS algorithms make the assumption that the overall structure of the network is linear and focus solely on accuracy to assess the performance of candidate networks. This paper introduces a novel NAS algorithm based on a multi-objective modeling of the network design problem to design accurate Convolutional Neural Networks (CNNs) with a small structure. The proposed algorithm makes use of a graph-based representation of the solutions which enables a high flexibility in the automatic design. Furthermore, the proposed algorithm includes novel ad-hoc crossover and mutation operators. We also propose a mechanism to accelerate the evaluation of the candidate solutions. Experimental results demonstrate that the proposed NAS approach can design accurate neural networks with limited size.

Computational design of porous stochastic fibre network structure

A novel methodology to design porous fibre network structure or material has been proposed. The methodology is computational in nature considering a MATLAB based algorithm, which was uniquely devised to achieve stochastic fibre networks with layered structure. The fibre network architecture that is actively controlled by the algorithm includes network porosity, fibre orientation distribution and fibre-segment aspect ratio. The algorithm generates coordinates of individual fibres which are then modelled into a planar mesh. Further, the fibre network structure is achieved by stacking layers of several such meshes considering an inter-mesh overlap volume. The designed porous structure has achieved close to 90% porosity. Its fibre orientation distribution was isotropic with fibres laying uniformly in all directions in the mesh planes. Overall, the developed porous fibre network structure has been computationally validated to possess the desired structural parameters.

3D brain glioma segmentation in MRI through integrating multiple densely connected 2D convolutional neural networks.

To overcome the computational burden of processing three-dimensional (3D) medical scans and the lack of spatial information in two-dimensional (2D) medical scans, a novel segmentation method was proposed that integrates the segmentation results of three densely connected 2D convolutional neural networks (2D-CNNs). In order to combine the low-level features and high-level features, we added densely connected blocks in the network structure design so that the low-level features will not be missed as the network layer increases during the learning process. Further, in order to resolve the problems of the blurred boundary of the glioma edema area, we superimposed and fused the T2-weighted fluid-attenuated inversion recovery (FLAIR) modal image and the T2-weighted (T2) modal image to enhance the edema section. For the loss function of network training, we improved the cross-entropy loss function to effectively avoid network over-fitting. On the Multimodal Brain Tumor Image Segmentation Challenge (BraTS) datasets, our method achieves dice similarity coefficient values of 0.84, 0.82, and 0.83 on the BraTS2018 training; 0.82, 0.85, and 0.83 on the BraTS2018 validation; and 0.81, 0.78, and 0.83 on the BraTS2013 testing in terms of whole tumors, tumor cores, and enhancing cores, respectively. Experimental results showed that the proposed method achieved promising accuracy and fast processing, demonstrating good potential for clinical medicine.

Near real-time single-beat myocardial infarction detection from single-lead electrocardiogram using Long Short-Term Memory Neural Network

This study proposes a novel Long Short-Term Memory Neural Network (LSTM) architecture for the diagnosis of myocardial infarctions from individual heartbeats of single-lead electrocardiograms (ECGs). The proposed model is trained using an unbiased patient split approach and validated using 10-fold cross-validation over 148 myocardial infarction and 52 Healthy Control patients from the Physikalisch-Technische Bundesanstalt diagnostic ECG Database to generate an inter-patient classifier. We further demonstrate why special care must be taken when generating the training and testing datasets by exploring the effects of various data-split techniques that could mask the occurrence of overfitting and produce misleadingly high testing metrics of the model's performance. A thorough assessment of these results is provided using several standard metrics for different data split methods to show their tendency to overfitting, data leakage, and bias introduced from previously seen heart beats during the training phase. The design achieves near real-time diagnosis of 40 ms while providing an accuracy of 89.56% (with a 95% Confidence Interval (CI) of ±2.79%), recall/sensitivity of 91.88% (±3.13% 95%CI), and a specificity of 80.81% (±9.62% 95%CI). The fast processing makes the model readily deployable on currently existing mobile devices and testing instruments. The achieved performance makes the proposed method a new research direction for attaining real-time and unbiased diagnosis. While, the modular architectural design of the LSTM network structure, which is amenable for the inclusion of other ECG leads, could serve as a platform for early detection of myocardial infarction and for the planning of early treatment(s).

A Review of Polymer Composites Based on Carbon Fillers for Thermal Management Applications: Design, Preparation, and Properties.

With the development of microelectronic devices having miniaturized and integrated electronic components, an efficient thermal management system with lightweight materials, which have outstanding thermal conductivity and processability, is becoming increasingly important. Recently, the use of polymer-based thermal management systems has attracted much interest due to the intrinsic excellent properties of the polymer, such as the high flexibility, low cost, electrical insulation, and excellent processability. However, most polymers possess low thermal conductivity, which limits the thermal management applications of them. To address the low thermal conduction of the polymer materials, many kinds of thermally conductive fillers have been studied, and the carbon-based polymer composite is regarded as one of the most promising materials for the thermal management of the electric and electronic devices. In addition, the next generation electronic devices require composite materials with various additional functions such as flexibility, low density, electrical insulation, and oriented heat conduction, as well as ultrahigh thermal conductivity. In this review, we introduce the latest papers on thermally conductive polymer composites based on carbon fillers with sophisticated structures to meet the above requirements. The topic of this review paper consists of the following four contents. First, we introduce the design of a continuous three-dimensional network structure of carbon fillers to reduce the thermal resistance between the filler–matrix interface and individual filler particles. Second, we discuss various methods of suppressing the electrical conductivity of carbon fillers in order to manufacture the polymer composites that meet both the electrical insulation and thermal conductivity. Third, we describe a strategy for the vertical alignment of carbon fillers to improve the through-plane thermal conductivity of the polymer composite. Finally, we briefly mention the durability of the thermal conductivity performance of the carbon-based composites. This review presents key technologies for a thermal management system of next-generation electronic devices.

The future of three-dimensional skin graft: a mini-review

Over the last few years, the development of three-dimensional (3D) skin bioprinting has increased rapidly. The concept of bioprinting involves a computer in designing using conventional 3D printing. 3D bioprinting can produce tissue structures such as skin and have the opportunity to substitute autografts, which are still the gold standard today. Network construction in bioprinting requires stages in the network structure design, selection of cell and biomaterial raw materials, and the molding process. The use of keratinocytes and fibroblasts is used as the main raw material in the manufacture of the dermis and epidermis layers. The biomimetic technique will combine these raw materials with other biomaterials (hydrogel, collagen, fibrin, gelatin, alginate, chitosan, and chitin) and growth factors (hydrocortisone, thrombin, and FGF-2) so that the physiological condition of the skin can be created. Currently, 3D bioprinting research continues to develop in response to the various shortcomings and challenges faced. The purpose of this review is to discuss the current use of 3D bioprinting technology, the principles of this technology and the prospects for 3D skin bioprinting in the future.

Research on Optimal Design of Short-Life Cycle Product Logistics Supply Chain Based on Multicriteria Decision Model

In order to study the special production, distribution, transportation, and sales of short-lived and short-lived multimedia products, this paper studies the cost and benefit optimization of a three-level supply chain network consisting of a supplier-producer-distribution center. Simultaneously, to effectively describe the actual inventory level of the retailer’s multimedia products, this paper introduces a new nonlinear function to define the real-time inventory of multimedia products to establish a biobjective nonlinear mixed-integer programming model to estimate the inventory. The research results show that the model established in this paper can not only provide the overall cost and optimal decision-making plan for the short-lived dairy supply chain but also the design method is better than the standard constraint method. Relevant research results have important potential guiding significance for research and design of supply chain network structure, measurement of multimedia product inventory levels, and improvement of customer satisfaction. The optimized design of the logistics supply chain of multimedia products with short life cycle under the background of new retail is beneficial.

Genetic Algorithm-based Optimization of Deep Neural Network Ensemble

Personal identification is the task of authenticating a person using individual biological features. Deep neural networks (DNNs) have demonstrated an impressive performance in this field. Since no general algorithm is available for the design of network structures and the parameters adopted in DNNs for every application problem, DNNs should be determined according to the programmers’ experiments and know-how. For a new application task, it is very time-consuming for non-experts to design network structure, hyperparameters and an ensemble of base models adequately and effectively. In this paper, we present a genetic algorithm (GA)-based approach to construct network structures, tune their hyperparameters, and generate base models for the ensemble algorithm. The ensemble is constructed from base models with different network structures according to the voting ensemble algorithm. Our original personal identification dataset is employed as the numerical example to illustrate the performance of the proposed method. The results show that the prediction accuracy of the ensemble model is better than that of the base models and that the prediction of walking behavior toward the Kinect at 90 degrees and 225 degrees is more difficult than other walking behaviors.

Design of P2P-SIP System Network Structure Based on JXTA

Session Initiation Protocol (SIP) is one of the most important signaling connection protocols in the field of VoIP (Voice over Internet Protocol). However, the existing SIP network has some problems such as manual configuration, difficult maintenance and poor scalability. With the development of P2P (Peer-to-Peer) technology, the combination of SIP and P2P has become a trend. JXTA is an open peer-to-peer platform, which provides basic application services for building P2P network. The topic chooses JXTA development platform to build P2P network, and then constructs SIP soft terminal on it. Based on the in-depth study of JXTA structured network and SIP protocol, a P2P SIP system is designed and implemented. The system can maintain normal SIP communication without maintaining the central server. In this paper, the detailed design and implementation process of P2P-SIP system are given, including node join and leave, user join, user search and session establishment. This paper focuses on the implementation process of user search, that is, the collaborative process of discovery service, resolution service, aggregation service and endpoint routing service. Finally, the feasibility of the system is verified by functional test.

Design of a SiOC network structure with oxidation stability and application to hydrogen separation membranes at high temperatures

SiOC membranes were fabricated by a sol-gel method using three types of silica precursors: vinyltrimethoxysilane (VTMS), triethoxysilane (TRIES), and 1,1,3,3-tetramethyldisiloxane (TMDSO). The timing of the addition of TMDSO was evaluated via single-gas permeation for its effect on formation of the SiOC structure and on membrane performance. The effect that VTMS compounding exerted on membrane performance and on oxidation stability was also evaluated. The network size of SiOC membranes was precisely tailored by controlling the timing of the addition of TMDSO during the sol preparation process. There was no significant change in the network pore size either before or after thermally induced hydrosilylation (Si–C=C + Si–H → Si–CH2–CH2–Si) and the cross-linking of Si–CH3 groups (Si–CH3 + Si–CH3 → Si–CH2–Si + CH4). These SiOC membranes were highly selective of H2 over larger molecules such as CF4 and SF6. The effect of VTMS compounding on network pore size was small, but the oxidation stability of the SiOC structure at 500 °C was dramatically enhanced.

Structure and performance preparation on alginate-based fibrous aerogel with double network

To sodium alginate and acrylamide (AM) as raw material, using ammonium persulfate (APS) as initiator, N,N-methylene bisacrylamide (MBA) as crosslinking agent, SA/PAM-based fibrous aerogels are prepared by extrusion and freeze drying techniques based on dual network structure design. The results show that the best preparation process is one step molding, acrylamide 50 wt%, 1.2 mm spinneret pore size, 9 mm/min spinning rate, 3 wt% CaCl2 coagulation bath, 5 min coagulation time. The aerogels present a microscopic multi-layered pore structure with macroporous intervening mesopores and micropores, and rendered as macroscopic fibrous. The average pore size of macropores is 2093.70 nm, the specific surface area is 230.00 m2/g, the porosity is 72.10%. The average pore size of mesopores and micropores is 40.44 nm, the specific surface area is 18 m2/g and the thermal conductivity of fiber aerogels/epoxy composites is 0.1321 w/m·k.

Analysis of small-scale topology and macroscale mechanical properties of shape memory chiral-lattice metamaterials

Chiral-lattice metamaterials with unique mechanical, optical, electrical and magnetic properties have been widely used in many engineering applications, such as chiral recognition and separation, stealth materials and devices, as well as biological sensors. Various design strategies have been developed to obtain chiral-lattice metamaterials with different mechanical properties. The fast grown design strategies that can provide a variety of Poisson’s ratios and elastic moduli have brought new opportunities to emerging biomedical applications. In this study, hexagonal and tetragonal chiral-lattice metamaterials with positive and negative Poisson’s ratios were designed. The design strategy was based on a network structure with a periodic lattice topology, which was formed by three outer rings connected to the central ring via ligaments. The wide applications of this type of metamaterial were investigated via experiments and finite-element analysis. Demonstrative examples and the shape-memory effect (SME) suggest that the chiral-lattice structure has great potential in developing programmable metamaterials. The findings of this study provide essential guidance for the design of chiral network structures with the desired mechanical properties.

A Survey on Continuous Object Tracking and Boundary Detection Schemes in IoT Assisted Wireless Sensor Networks

With the new age of data innovation, the Internet of Things (IoT) proliferation has drawn enormous thought and has applied to help applications in different fields i.e. natural assurance, military observing, and industrial applications. WSNs are the essential segment of IoT for monitoring as well as tracking. The most preeminent applications provide confinement and identification of continuous objects i.e. wildfire, toxic gas, bio synthetics concoctions, and so forth. In the case of continuous objects such as fire and toxic gases are detected to identify the boundary of damage and alert teams for rescue efforts. It is also helpful in identifying safe paths for rescue. We have investigated various existing surveys that carried out different concepts associated with continuous object tracking and find out the deficit of boundary detection of object. In order to replete the present cleft of analysis, we have inspected various current state-of-the-art works on boundary detection of a continuous object that has yet not been added to the current writing. This paper presents an extensive overview of different continuous object tracking schemes which involve energy efficiency, boundary detection, communication, data aggregation, and network structural design in literature with the aid of featuring taxonomy. We summarized, compared, and classified these schemes along with their analysis and performance. Moreover, for further evaluation mechanism, strengths and weaknesses of these schemes are presented. Finally, various state-of-the-art open research challenges are identified. Moreover, there is a need to overcome these challenges through novel and reliable arrangements by the researchers.

Mechanical Fault Diagnosis Methods Based on Convolutional Neural Network: a Review

Deep learning is good at abstract features from massive data and has good generalization ability, which has attracted more and more researchers’ attention. The Convolutional Neural Network (CNN) is a classic structure of deep learning and which is being widely and successfully used in the fields of computer vision, target detection, natural language processing, and speech recognition. Based on a detailed analysis of the current status and needs of mechanical system fault diagnosis, this paper introduces the structure of CNN and summarizes the application of CNN in the field of mechanical faults from the aspects of input data type, network structure design, and migration learning. The problems of deep feature extraction and visualization are also discussed, and finally, the difficulties in mechanical fault diagnosis are analyzed and several problems to be solved in the field of mechanical fault diagnosis based on CNN prospect.

Target Recognition of SAR Image Based on CN-GAN and CNN in Complex Environment

In recent years, with the rapid development of deep learning, the research of radar image automatic target recognition (ATR) has made great progress. However, because of the complex environments and special imaging principles, Synthetic Aperture Radar (SAR) image still have the problems of sample scarcity and strong speckle noise, which affects the target recognition performance. To solve the above problems, we proposed a target recognition method of SAR image based on Constrained Naive Generative Adversarial Networks (CN-GAN) and Convolutional Neural Network (CNN). Combining Least Squares Generative Adversarial Networks (LSGAN) and Image-to-Image Translation (Pix2Pix), CN-GAN can overcome these problems of low Signal-to-Clutter-Noise Ratio (SCNR), model instability and the excessive freedom degree of the output, which are produced by conventional naive GAN. Besides, we adopted a shallow network structure design in CNN, which can effectively improve the generalization ability of the model and avoid the problem of model overfitting. The experimental results in this paper demonstrate that CN-GAN has achieved the data generation and data enhancement, the SCNR of generated data is higher than the origin data set and data sets gained by other forms of GANs, the recognition performance based on the extended data set is better than the origin data set, and the recognition rate of data set enhanced by CN-GAN is higher than that of other common data enhancement methods.

Forecast of the Employment Situation of College Graduates Based on the LSTM Neural Network.

Scientific and reasonable forecast model of graduates' employment data can efficaciously embody the complex characteristics of graduates' employment data and embody the nonlinear dynamic interaction of influencing elements of graduates' employment situation. It has a strong and steady characteristic learning capability, thus selecting the main influence data that influence the change of graduates' employment data. In this paper, according to the situation embodied by students' employment, a data mining analysis model is set up by using the statistical method based on the model of cluster analysis technology to forecast the employment situation of graduates. In this paper, a forecast technique of graduates' employment situation based on the long short-term memory (LSTM) recurrent neural network is conceived, including network structure design, network training, and forecast process implementation algorithm. In addition, aiming at minimizing the forecasting error, an LSTM forecasting model parameter optimization algorithm based on multilayer grid search is conceived. It also verifies the applicability and correctness of the LSTM forecasting model and its parameter optimization algorithm in the analysis of graduates' employment situation.

Mixed Attention Densely Residual Network for Single Image Super-Resolution

Recent applications of convolutional neural networks (CNNs) in single image super-resolution (SISR) have achieved unprecedented performance. However, existing CNN-based SISR network structure design consider mostly only channel or spatial information, and cannot make full use of both channel and spatial information to improve SISR performance further. The present work addresses this problem by proposing a mixed attention densely residual network architecture that can make full and simultaneous use of both channel and spatial information. Specifically, we propose a residual in dense network structure composed of dense connections between multiple dense residual groups to form a very deep network. This structure allows each dense residual group to apply a local residual skip connection and enables the cascading of multiple residual blocks to reuse previous features. A mixed attention module is inserted into each dense residual group, to enable the algorithm to fuse channel attention with laplacian spatial attention effectively, and thereby more adaptively focus on valuable feature learning. The qualitative and quantitative results of extensive experiments have demonstrate that the proposed method has a comparable performance with other state-of-the-art methods.