Search

This paper analyzes the inherent relationship between convolutional neural networks and sparse representation and proposes an improved convolutional neural network model for image synthesis in response to problems with current methods. In the testing phase, the calculation of the sparse coefficients involves the solution of complex optimization problems, which greatly reduce the operating efficiency, inspired by the successful application of convolutional neural networks in the field of image reconstruction. Compared with the traditional image portrait synthesis method, this model not only has an end-to-end closed form but also does not need to solve complex optimization problems in the synthesis stage. The synthesis experiment on an image dataset shows that this method not only improves the synthesis effect but also improves the efficiency of the traditional method by one to two orders of magnitude, demonstrating its potential application value. Blocking processing is a common method for sparse domain image modeling. It improves the computational efficiency but also decreases the global structure of the image, which is difficult to compensate for through the aggregation and overlap of image blocks. In response to this problem, this paper proposes a low-rank image inpainting method based on a Gaussian mixture model. This method embeds the local statistical characteristics of image blocks into the kernel norm model and not only uses the Gaussian mixture model to maintain the local details of the image but also describes the global low-rank structure of the image through the kernel norm, thus restoring a class of image data with a potential low-rank structure and theoretically revealing the structured sparse nature of the Gaussian mixture model. This paper optimizes the strategy based on random hidden neuron nodes and proposes a dropout anti-overfitting strategy based on sparsity. The experiments show that this strategy can effectively improve the convergence speed while ensuring good performance and can effectively prevent overfitting.

Brain tumor detection and segmentation from Magnetic Resonance Imaging (MRI) images is being one of the emerging fields in the biomedicine. A formidable undertaking in brain tumor surgery, medical care, treatment programme and quantitative assessment of MRI images is to precisely diagnose its location and extent. Recently, the convolutional neural network (CNN) based detection and segmentation method on brain tumor MRI images is being one of the emerging fields in the medical imaging as an automatic clinic treatment and evaluation solution. In this article, we put forward a brand new quadruplet loss in CNN framework, which achieves higher accuracy in brain tumor detection and segmentation than other pairwise loss and triplet loss methods. By applying the proposed quadruplet loss to the original L2Net CNN architecture leads to a more compact descriptor named QuadrupletNet. From our experiments, QuadrupletNet shows higher performance than other state-of-the-art loss functions e.g., the Triplet loss, as indicated in experiments on Multimodal Brain Tumor Image Segmentation (BRATS 2018) datasets, and on our own collected MRI brain tumor datasets (named MBTD).

In large-scale applications of compressed sensing (CS), the time cost to reconstruct the original signal is too high. To accelerate the reconstruction and reduce the space cost of the measurement matrix, a novel parallel reconstruction approach based on a semi-tensor product (STP) is proposed. A low-dimensional random matrix where the dimensions are 1/4 (or 1/16, 1/64, 1/256, 1/1024 or even 1/4096) that of conventional CS is generated to sample the original data, and then a parallel reconstruction method is proposed to obtain the solution with the iteratively re-weighted least-squares (IRLS) algorithm. The peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and time cost of reconstruction were evaluated and compared with matrices of different dimensions, and comparisons were also conducted with other state-of-the-art methods. Numerical results show that the speed can be effectively improved (10×, or 100×, even 1000×) and the storage space of the matrix can also be remarkably reduced; that is, the matrix can be 1/4096 of conventional CS. Furthermore, the numerical results show that our formulation outperforms conventional CS in speed of reconstruction and in its comparable quality, which is important for real-time and physical implementation of applications.

Considering the movement of a sink node in this study to solve the data transmission problem of sensor nodes in a 3D network environment, we propose a Data Collection Algorithm of a 3D wireless sensor network (DCA_3D) that weighs node coverage rate and lifetime. DCA_3D establishes a data collection optimization model that weighs node coverage rate and lifetime with the constraints of the sink node's moving path selection, data flow, energy consumption, and link transmission. Subsequently, DCA_3D calculates the fitness value of the sink node's moving path by solving the data collection optimization model with the known sink node's moving path. Then it uses a modified artificial bee colony algorithm to solve the moving path selection problem of the sink node, and finally obtains the optimal scheme. In the scheme, sink node can find the optimal moving path, whereas the sensor node can find the optimal data communication path. The simulation results show that regardless of the moving path length of the sink node, the maximum data collection hops of the sink node and the number of static sensor nodes change, DCA_3D can find the optimal moving path of the sink node. It can improve the coverage rate of the sensor nodes, the network lifetime and average data transmission amount, and reduce the average energy consumption variance and the average packet loss rate. DCA_3D outperforms the state-of-arts such as RAND, GREED, EDG_3D, and ANT.

With the gradual improvement of people’s living standards, the production and drinking of all kinds of food is increasing. People’s disease rate has increased compared with before, which leads to the increasing number of medical image processing. Traditional technology cannot meet most of the needs of medicine. At present, convolutional neural network (CNN) algorithm using chaotic recursive diagonal model has great advantages in medical image processing and has become an indispensable part of most hospitals. This paper briefly introduces the use of medical science and technology in recent years. The hybrid algorithm of CNN in chaotic recursive diagonal model is mainly used for technical research, and the application of this technology in medical image processing is analysed. The CNN algorithm is optimized by using chaotic recursive diagonal model. The results show that the chaotic recursive diagonal model can improve the structure of traditional neural network and improve the efficiency and accuracy of the original CNN algorithm. Then, the application research and comparison of medical image processing are performed according to CNN algorithm and optimized CNN algorithm. The experimental results show that the CNN algorithm optimized by chaotic recursive diagonal model can help medical image automatic processing and patient condition analysis.

In the distributed file sharing system, a large number of users share bandwidth, upload resources and store them in a decentralized manner, thus offering both an abundant supply of high-quality resources and high-speed download. However, some users only enjoy the convenient service without uploading or sharing, which is called free riding. Free-riding may discourage other honest users. When free-riding users mount to a certain number, the platform may fail to work. The current available incentive mechanisms, such as reciprocal incentive mechanisms and reputation-based incentive mechanisms, which suffer simple incentive models, inability to achieve incentive circulation and dependence on a third-party trusted agency, are unable to completely solve the free-riding problem.In this paper we build a blockchain-based distributed file sharing platform and design a nested incentive scheme for this platform. The proposed nested incentive mechanism achieves the circulation of incentives in the platform and does not rely on any trusted third parties for incentive distribution, thus providing a better solution to free-riding. Our distributed file sharing platform prototype is built on the current mainstream blockchain. Nested incentive scheme experiments on this platform verify the effectiveness and superiority of our incentive scheme in solving the free-riding problem compared to other schemes.

In order to improve the revenue of attacking mining pools and miners under block withholding attack, we propose the miner revenue optimization algorithm (MROA) based on Pareto artificial bee colony in blockchain network. MROA establishes the revenue optimization model of each attacking mining pool and revenue optimization model of entire attacking mining pools under block withholding attack with the mathematical formulas such as attacking mining pool selection, effective computing power, mining cost and revenue. Then, MROA solves the model by using the modified artificial bee colony algorithm based on the Pareto method. Namely, the employed bee operations include evaluation value calculation, selection probability calculation, crossover operation, mutation operation and Pareto dominance method, and can update each food source. The onlooker bee operations include confirmation probability calculation, crowding degree calculation, neighborhood crossover operation, neighborhood mutation operation and Pareto dominance method, and can find the optimal food source in multidimensional space with smaller distribution density. The scout bee operations delete the local optimal food source that cannot produce new food sources to ensure the diversity of solutions. The simulation results show that no matter how the number of attacking mining pools and the number of miners change, MROA can find a reasonable miner work plan for each attacking mining pool, which increases minimum revenue, average revenue and the evaluation value of optimal solution, and reduces the spacing value and variance of revenue solution set. MROA outperforms the state of the arts such as ABC, NSGA2 and MOPSO.

Time series have broad usage in the wireless Internet of Things. This article proposes a nonlinear time series prediction algorithm based on the Small-World Scale-Free Network after the AIC-Optimized Subtractive Clustering Algorithm (AIC-DSCA-SSNET, AD-SSNET) to predict the nonlinear and unstable time series, which improves the prediction accuracy. The AD-SSNET is introduced as a reservoir based on the echo state network to improve the predictive capability of nonlinear time series, and combined with artificial intelligence method to construct the prediction model training samples. First, the optimal clustering scheme of randomly distributed neurons in the network is adaptively obtained by the AIC-DSCA, then the AD-SSNET is constructed according to the intra-cluster priority connection algorithm. Finally, the reservoir synaptic matrix is calculated according to the synaptic information. Experimental results show that the proposed nonlinear time series prediction algorithm extends the feasible range of spectral radii of the reservoir, improves the prediction accuracy of nonlinear time series, and has great significance to time series analysis in the era of wireless Internet of Things.

In order to improve the revenues of attack mining pools and miners under block withholding attack, we propose the mining revenue optimization algorithm (MROA) of miners in PoW-based blockchain network. MROA establishes the revenue optimization model of each attack mining pool and revenue optimization model of entire mining attack pools under block withholding attack with the mathematical formulas such as attack mining pool selection, effective computing power, mining cost and revenue. Then MROA solves the model by using the modified artificial bee colony algorithm based on Pareto. Namely, employed bee operations include evaluation value calculation, selection probability calculation, crossover operation, mutation operation and Pareto domination calculation, and can update each food source. The onlooker bee operations include confirmation probability calculation, crowding degree calculation, neighborhood crossover operation, neighborhood mutation operation and Pareto domination calculation, and can find the optimal food source in multidimensional space with smaller distribution density. Scout bee operations delete the local optimal food source which cannot produce new food sources to ensure the diversity of solutions. The simulation results show that no matter how the number of attack mining pools and the number of miners change, MROA can find a reasonable miner work plan for each attack mining pool, which improves minimum revenue, average revenue and the evaluation value of optimal solution, and reduces the spacing value and variance of revenue solution set. MROA outperforms the state-of-arts such as ABC, NSGA2 and MOPSO.

Large and dense mobile ad hoc networks often meet scalability problems, the hierarchical structures are needed to achieve performance of network such as cluster control structure. Clustering in mobile ad hoc networks is an organization method dividing the nodes in groups, which are managed by the nodes called cluster-heads. As far as we know, the difficulty of clustering algorithm lies in determining the number and positions of cluster-heads. In this article, the subtractive clustering algorithm based on the Akaike information criterion is proposed. First, Akaike information criterion is introduced to formulate the optimal number of the cluster-heads. Then, subtractive clustering algorithm is used in mobile ad hoc networks to get several feasible clustering schemes. Finally, the candidate schemes are evaluated by the index of minimum of the largest within-cluster distance variance to determine the optimal scheme. The results of simulation show that the performance of the proposed algorithm is superior to widely referenced clustering approach in terms of average cluster-head lifetime.

Mobile sensor networks (MSN) in the city environment have dynamic topology and large amounts of data. In order to overcome the problem of data uploading and exchange, data uploading and exchange algorithm (DUEA) is proposed. In DUEA, the three-tier network architecture composed of vehicle nodes, relay nodes and sink nodes is considered. The link estimator, routing engine, and forwarding engine are modified based on CTP (collection tree protocol). In the link estimator, beacon packet is simplified. In the routing engine, the numbers of elements in the send queue of neighbor nodes, and link quality are weighed. The new calculation formulae of ETX value and selection method of father node are proposed. In the forwarding engine, different snooping methods are used for two types of data named general data and warning data. Simulation results show that compared with CTP, DUEA increases data throughput and transmission rate and reduces energy consumption of packet transmission when data is uploaded. DUEA also reduces average transmission delay when data is exchanged. Physical test results show that DUEA achieves desired effect in the testbed. DUEA is effective and feasible and outperforms CTP in the city traffic environment.

In order to overcome the energy hole problem in some wireless sensor networks (WSNs), lifetime optimization algorithm with mobile sink nodes for wireless sensor networks (LOA_MSN) is proposed. In LOA_MSN, hybrid positioning algorithm of satellite positioning and RSSI positioning is proposed to save energy. Based on location information, movement path constraints, flow constraint, energy consumption constraint, link transmission constraint, and other constraints are analyzed. Network optimization model is established and decomposed into movement path selection model and lifetime optimization model with known grid movement paths. Finally, the two models are solved by distributed method. Sink nodes gather data of sensor nodes along the calculated paths. Sensor nodes select father nodes and transmit data to corresponding sink node according to local information. Simulation results show that LOA_MSN makes full use of node energy to prolong network lifetime. LOA_MSN with multiple sink nodes also reduces node energy consumption and data gathering latency. Under certain conditions, it outperforms MCP, subgradient algorithm, EASR, and GRAND.

Ad Hoc network is recently a hot spot in wireless network researching domain. The design of routing algorithm, which is as the core layer technology in Ad Hoc networks, has drawn great concern. After analyzing the drawbacks of the existing algorithms, this paper puts forward a congestion control routing algorithm based on the node degree. This paper explains the principle of the node degree algorithm; we select nodes with low degree to forward data packets in the process of routing searching. Results of simulation show: number of relay nodes is reduced significantly in the process of routing setup.

The Coverage Model is constructed to meet the requirement that every point of the whole area is in the coverage of repeaters. The main algorithm is using small circles to cover the big circle with the rule that the distance between two circles' centers is not bigger than the radius of a repeater's coverage. To simplify the matter, we replace circles with particular regular hexagons, whose centers are locations to put repeaters. Using this model, we can work out the number of repeaters' locations.

To prolong network lifetime, nearest-neighbor clustering power control algorithm is proposed. In the algorithm, k-means algorithm is used to divide nodes into k different clusters. Nearest-neighbor distances algorithm is used to calculate the optimal transmission power of nodes which are in the same cluster. Then sink node broadcasts to inform other nodes that they use optimal transmission power for communicating with neighbor nodes which are in the same cluster and use default maximum transmission power for communicating with neighbor nodes which are in the different clusters. Simulation results show that by selecting the appropriate parameters, NNCPCA can prolong network lifetime, reduce network latency time and enable cost-effective energy consumption. Under certain conditions, it is fit for multi-cluster wireless sensor networks.