Convergence Speed Of Network Research Articles

Lightweight Object Detection Algorithm for UAV Aerial Imagery

Addressing the challenges of low detection precision and excessive parameter volume presented by the high resolution, significant scale variations, and complex backgrounds in UAV aerial imagery, this paper introduces MFP-YOLO, a lightweight detection algorithm based on YOLOv5s. Initially, a multipath inverse residual module is designed, and an attention mechanism is incorporated to manage the issues associated with significant scale variations and abundant interference from complex backgrounds. Then, parallel deconvolutional spatial pyramid pooling is employed to extract scale-specific information, enhancing multi-scale target detection. Furthermore, the Focal-EIoU loss function is utilized to augment the model’s focus on high-quality samples, consequently improving training stability and detection accuracy. Finally, a lightweight decoupled head replaces the original model’s detection head, accelerating network convergence speed and enhancing detection precision. Experimental results demonstrate that MFP-YOLO improved the mAP50 on the VisDrone 2019 validation and test sets by 12.9% and 8.0%, respectively, compared to the original YOLOv5s. At the same time, the model’s parameter volume and weight size were reduced by 79.2% and 73.7%, respectively, indicating that MFP-YOLO outperforms other mainstream algorithms in UAV aerial imagery detection tasks.

Research on structural damage identification and localization based on artificial neural network

Abstract Structural health monitoring is a research hotspot in engineering, and structural damage identification is one of the key problems in structural health monitoring research. This paper proposes a study on structural damage identification and localization based on artificial neural networks, and for the problem that the learning convergence speed of the BP neural network is too slow, a genetic algorithm is used to optimize the update of weights and thresholds during training and learning. In the finite element simulation, the structure's primary and secondary damage are taken as input nodes, and the optimized GA-BP neural network is used for training and identification. For the localization recognition of the primary damage of the structure, the maximum recognition relative errors of both the BP neural network and the GA-BP neural network did not exceed 5%, but the latter's accuracy was 2.63% higher than that of the former. For the localization recognition of secondary damage, the GA-BP neural network can effectively recognize 90% of the samples. The artificial neural network-based structural damage recognition localization has high recognition efficiency and accuracy, which is conducive to improving the robustness of the structural damage recognition system and is of significant help to real-time structural health monitoring.

CAIR: Combining integrated attention with iterative optimization learning for sparse-view CT reconstruction

Sparse-view CT is an efficient way for low dose scanning but degrades image quality. Inspired by the successful use of non-local attention in natural image denoising and compression artifact removal, we proposed a network combining integrated attention and iterative optimization learning for sparse-view CT reconstruction (CAIR). Specifically, we first unrolled the proximal gradient descent into a deep network and added an enhanced initializer between the gradient term and the approximation term. It can enhance the information flow between different layers, fully preserve the image details, and improve the network convergence speed. Secondly, the integrated attention module was introduced into the reconstruction process as a regularization term. It adaptively fuses the local and non-local features of the image which are used to reconstruct the complex texture and repetitive details of the image, respectively. Note that we innovatively designed a one-shot iteration strategy to simplify the network structure and reduce the reconstruction time while maintaining image quality. Experiments showed that the proposed method is very robust and outperforms state-of-the-art methods in terms of both quantitative and qualitative, greatly improving the preservation of structures and the removal of artifacts.

Percussion-based concrete fiber content recognition using homologous heterogeneous data fusion and denoising deep learning network

With the development of materials science, fiber-reinforced concrete (FRC) has attracted much attention in building engineering areas and gradually alternated the traditional regular concrete in recent years. Since the fiber content significantly affects the performance of FRC material, it is of great importance to detect the fiber content in practical applications. Thus, this research proposes a novel percussion-based recognition method to evaluate the fiber content in FRC material, which contains a simple but effective time series data fusion approach and an adroitly integrated one-dimensional denoising autoencoder-enhanced convolutional neural network (1D-DAECNN). Particularly, compared with most current existing percussion-based recognition methods, which utilize single acoustic signals, the proposed method simultaneously adopts the homologous heterogeneous data obtained respectively from the microphone and embedded piezoceramic transducer after one-time percussion. Then, the homologous heterogeneous data are processed by the presented time series data fusion approach to achieve time domain waveform features fusion, which aims to improve the convergence speed of deep learning networks. After that, the 1D-DAECNN is constructed to identify the fiber content in FRC material. Via the front-end denoising autoencoder (DAE) and the back-end one-dimensional convolutional neural network (1D-CNN) in the integrated network, the noise suppression capability of the 1D-DAECNN can be improved, and the manual feature extraction can be avoided during the training process. Finally, the effectiveness of the proposed method is demonstrated via laboratory tests on five FRC specimens. The results indicate that, compared with the conventional percussion-based method, the convergence speed and recognition precision of the proposed method is significantly improved even under noise conditions.

Forecast of Advanced Human Capital Gap Based on PSO-BP Neural Network and Coordination Pathway: Example of Beijing–Tianjin–Hebei Region

The upgrading of human capital caused by education is significant to regional development. Reasonable predictions of the degree of advanced human capital in different regions are effective for formulating reasonable talent policies and accelerating regional coordinated development. The BP neural network is a widely used prediction technology. PSO-BP neural network has good global search ability, which can accelerate the convergence speed of traditional BP neural network, which is suitable for forecasting larger data. The study takes the provincial data of China from 2005 to 2019 as an example, using PSO-BP neural network algorithm to predict the advanced level of human capital through the influencing factors filtered by OLS regression. The results show that: (1) Innovation ability and urbanization can play a decisive role in advanced human capital filtered by OLS regression; (2) The results of predicting the development trend of advanced human capital in the Beijing–Tianjin–Hebei region in 2020–2025 through the PSO-BP neural network have showed that there is still a large gap between the senior human capital stock in Hebei-Beijing-Tianjin in terms of total and per capita in 2020–2025 compared with other regions in east of China; (3) Giving full attention to elaborate the positive role of economic quality and quantity development are suitable for narrowing the difference of advanced human capital in this region. Through the method of OLS-BP-neural network, this study explores the gap and influencing factors of the Beijing–Tianjin–Hebei region, excavates the reasons for the huge gradient difference in the development of this region, and extends the machine learning prediction method to the analysis of the advanced level of human capital and the research of narrowing the regional development gap.

A new method to retrieve rainfall intensity level from rain-contaminated X-band marine radar image

ABSTRACT In this study, a new methodcombining wave-number energy spectrum (WES) and Genetic algorithm-back propagation neural network (GA-BPNN) isproposed to retrieve the rainfall intensity level from rain-contaminated X-bandmarine radar image. Since the intensity of spatialrainfall can be reflected by the distribution of energy in the wavenumberfrequency domain, the obtained WES is divided into three wavenumber segments(low, medium and high wavenumber segments), and the ratio of the wavenumber ineach wavenumber segment to the total wavenumber is calculated separately as the characteristicparameters. Based on the excellent networkconvergence speed and data prediction accuracy of the GA-BPNN, these calculatedparameters are input into the constructed GA-BPNN for training to complete thetask of rainfall intensity level retrieval. The proposed method is tested usingdata collected at the ocean observation station of Haitan Island in PingtanCounty. Referring to the actual rainfallintensity synchronously recorded by the rain gauge, the retrieval accuracy ofthe proposed method is 97.4%, which is 4.3% higher than that of back propagation neural network (BPNN) not optimizedby Geneticalgorithm(GA). In addition, compared with theretrieval performance of the ratio of zero intensity to echo (RZE) method basedon the occlusion area of radar image, the retrievalaccuracy of the proposed method is improved by about 12.9%.

A spatial evaluation method for earthquake disaster using optimized BP neural network model

Rapid spatial evaluation of seismic disaster after earthquake occurrence is required in disaster emergency rescue management, because of its importance in decreasing casualties and property losses. Among many categories of seismic disaster, evaluation of earthquake-affected population is of great significance to clarify the severity of earthquake disaster. For simple classic regression model, it is difficult to describe the strong nonlinear relationship between multiple influencing factors and earthquake disasters. In present study, an optimized BP neural network model considering spatial characteristic of influencing factors is proposed to evaluate the population distribution affected by earthquake. The correlation between earthquake-affected population and influencing factors is analysed using data of 2013 Ms7.0 Lushan earthquake. Ten influencing factors including elevation, slope angle, population density, per capita GDP, distance to fault, distance to river, NDVI, PGA, PGV, and distance to the epicentre, were classified into environmental and seismic factors. Correlation analysis revealed that per capita GDP and PGA factor had a stronger correlation with the earthquake-affected population. The earthquake-affected population was evaluated using a BP neural network by optimizing training samples considering spatial characteristics of per capita GDP and PGA factors. Different numbers of sample points, instead of a random distribution of sample points, were generated in areas with different value intervals of the influencing factors. The optimized samples improved the convergence speed and generalization capability of neuron network compared to random samples. The trained network was applied to the 2017 Ms7.0 Jiuzhaigou earthquake to verify its prediction accuracy. The MAE of the estimated earthquake-affected populations of different counties under Jiuzhaigou earthquake were 1.276 people/km2 using network model from optimized samples, smaller than the results of network model from random samples and linear regression model. The results indicate that BP neural network, which considers correlation characteristics of factors, has capability to evaluate spatial earthquake disaster.

Novel Spatial–Spectral Channel Attention Neural Network for Land Cover Change Detection with Remote Sensed Images

Land cover change detection (LCCD) with remote-sensed images plays an important role in observing Earth’s surface changes. In recent years, the use of a spatial-spectral channel attention mechanism in information processing has gained interest. In this study, aiming to improve the performance of LCCD with remote-sensed images, a novel spatial-spectral channel attention neural network (SSCAN) is proposed. In the proposed SSCAN, the spatial channel attention module and convolution block attention module are employed to process pre- and post-event images, respectively. In contrast to the scheme of traditional methods, the motivation of the proposed operation lies in amplifying the change magnitude among the changed areas and minimizing the change magnitude among the unchanged areas. Moreover, a simple but effective batch-size dynamic adjustment strategy is promoted to train the proposed SSCAN, thus guaranteeing convergence to the global optima of the objective function. Results from comparative experiments of seven cognate and state-of-the-art methods effectively demonstrate the superiority of the proposed network in accelerating the network convergence speed, reinforcing the learning efficiency, and improving the performance of LCCD. For example, the proposed SSCAN can achieve an improvement of approximately 0.17–23.84% in OA on Dataset-A.

A novel deep deterministic policy gradient model applied to intelligent transportation system security problems in 5G and 6G network scenarios

Traffic congestion has been an actual problem in large cities, causing personal inconvenience and environmental pollution. To solve this problem, new applications for Intelligent Transportation System (ITS) have been created, to monitor actual traffic conditions. Therefore, fast, reliable and safe systems are desirable for creating a real intelligent transportation environment. Deep learning algorithms have been proposed for a better understanding of traffic behavior from a security-related perspective. Thus, we aim to maximize the safety problems using a deep learning algorithm, where a novel policy gradient model is presented for detecting vehicular misuse. The proposed model uses a triple network replay algorithm, maximizing the network convergence speed. Three networks are selected to optimize the policy network variables. Finally, the replay algorithm is partitioned with the aim of obtaining a faster model. Simulations on a real urban map are performed in a scenario with the integration of 5G or 6G networks. An architectural model for the integration of a Vehicular Ad-hoc Network (VANET) and cellular networks is determined in software-defined networking (SDN). The results show that the accuracy prediction of the proposed system presents better performance compared to related studies, where the proposed model increases its convergence speed and cumulative reward. Thus, the ITS improvement by the proposed deep learning algorithm increases the prediction accuracy, and reduces the transmission delay, treating the traffic path according to the congestion.

Dynamically adaptive adjustment loss function biased towards few‐class learning

Convolution neural networks have been widely used in the field of computer vision, which effectively solve practical problems. However, the loss function with fixed parameters will affect the training efficiency and even lead to poor prediction accuracy. In particular, when there is a class imbalance in the data, the final result tends to favor the large-class. In detection and recognition problems, the large-class will dominate due to its quantitative advantage, and the features of few-class can be not fully learned. In order to learn few-class, batch nuclear-norm maximization is introduced to the deep neural networks, and the mechanism of the adaptive composite loss function is established to increase the diversity of the network and thus improve the accuracy of prediction. The proposed loss function is added to the crowd counting, and verified on ShanghaiTech and UCF_CC_50 datasets. Experimental results show that the proposed loss function improves the prediction accuracy and convergence speed of deep neural networks.

Reliability Optimization for Channel Resource Allocation in Multihop Wireless Network: A Multigranularity Deep Reinforcement Learning Approach

This article investigates the high-reliable data transmission for multihop and multichannel wireless sensor networks (WSNs), which jointly optimizes the channel allocation and channel access mechanisms. We propose a novel wireless paradigm empowered by mobile-edge computing (MEC) and deep reinforcement learning (DRL) to improve the data process ability of WSNs and formulate the joint resource allocation problem for reliability maximization as a partially observable Markov decision process (POMDP). Meanwhile, we introduce the distributed decision-making (DDM) framework to decouple channel optimization into two subproblems: 1) channel allocation and 2) channel access. Correspondingly, we present an asynchronous channel allocation algorithm for multiagent scenario and enable the neighbor cooperation to tackle the nonstationary problem, which can significantly improve the network convergence speed. Besides, we present a collision-free channel access algorithm including three submodules that can simultaneously eliminate vanishing nodes, hidden terminal, and exposed terminal problems in large-scale WSNs. Simulation results demonstrate that the proposed algorithm significantly improves network performance in terms of convergence, throughput, collision, and packet delivery ratio (PDR).

Improving the prediction of DNA-protein binding by integrating multi-scale dense convolutional network with fault-tolerant coding

Accurate prediction of DNA-protein binding (DPB) is of great biological significance for studying the regulatory mechanism of gene expression. In recent years, with the rapid development of deep learning techniques, advanced deep neural networks have been introduced into the field and shown to significantly improve the prediction performance of DPB. However, these methods are primarily based on the DNA sequences measured by the ChIP-seq technology, failing to consider the possible partial variations of the motif sequences and errors of the sequencing technology itself. To address this, we propose a novel computational method, termed MSDenseNet, which combines a new fault-tolerant coding (FTC) scheme with the dense connectional deep neural networks. Three important factors can be attributed to the success of MSDenseNet: First, MSDenseNet utilizes a powerful feature representation approach, which transforms the raw DNA sequence into fusion coding using the fault-tolerant feature sequence; Second, in terms of network structure, MSDenseNet uses a multi-scale convolution within the dense layer and the multi-scale convolution preceding the dense block. This is shown to be able to significantly improve the network performance and accelerate the network convergence speed, and third, building upon the advanced deep neural network, MSDenseNet is capable of effectively mining the hidden complex relationship between the internal attributes of fusion sequence features to enhance the prediction of DPB. Benchmarking experiments on 690 ChIP-seq datasets show that MSDenseNet achieves an average AUC of 0.933 and outperforms the state-of-the-art method. The source code of MSDenseNet is available at https://github.com/csbio-njust-edu/msdensenet. The results show that MSDenseNet can effectively predict DPB. We anticipate that MSDenseNet will be exploited as a powerful tool to facilitate a more exhaustive understanding of DNA-binding proteins and help toward their functional characterization.

An Improved Yolov5 for Multi-Rotor UAV Detection

Multi-rotor drones have a wide range of applications in practical scenarios; however, the use of multi-rotor drones for illegal acts is also on the rise, in order to improve the recognition accuracy of multi-rotor drones. A new multi-rotor drone detection algorithm is proposed. Firstly, the Yolov5 backbone is replaced with Efficientlite, thus reducing the number of parameters in the model. Secondly, adaptively spatial feature fusion is injected into the head of the baseline model to facilitate the fusion of feature maps with different spatial resolutions, in order to balance the accuracy loss caused by the lightweight of the model backbone. Finally, a constraint of angle is introduced into the original regression loss function to avoid the mismatch between the prediction frame and the real frame orientation during the training process in order to improve the speed of network convergence. Experiments show that the improved Yolov5s exhibits better detection performance, which provides a superior method for detecting multi-rotor UAVs in real-world scenarios.

Research on Lung Tumor Cell Segmentation Method Based on Improved UNet Algorithm

In order to improve the completeness of the computer-aided diagnosis system for the segmentation of large-sized lung tumors and the segmentation accuracy of small-sized lung tumors, a dual-attention 3D-UNet lung tumor segmentation network model was constructed. The upsampling operation in the traditional 3D-UNet network is replaced with the DUpsampling structure. By minimizing the loss between the pixels of the feature map and the compressed label image, a more expressive feature map is obtained, thereby improving the network convergence speed. On this basis, the spatial attention module and the channel attention module are integrated so that similar features in single channel and multichannel are related to each other, and the global correlation of feature maps is increased to improve the accuracy of segmentation results. The experimental results show that compared with methods such as 3D-UNet, the model effectively improves the accuracy of lung tumor cell segmentation, and the MIoU score on the public dataset LIDC-IDRI reaches 89.4%. The segmentation method will be closer to the facts and in line with the safety and health of human life.

Joint time scheduling and transaction fee selection in blockchain-based RF-powered backscatter cognitive radio network

In this paper, we develop a new framework called blockchain-and backscatter-aided Internet of Things (IoT) system. In the framework, IoT devices as secondary transmitters transmit their sensing data to a secondary gateway by using an RF-powered backscatter cognitive radio technology. The data collected at the gateway is then sent to a blockchain network for further verification, storage and processing. As such, the framework enables an IoT system to simultaneously optimize the spectrum usage and maximize the energy efficiency. Moreover, the framework ensures that the data collected from the IoT devices is verified, stored and processed in a decentralized but in a trusted manner. To achieve the goal, we formulate a stochastic optimization problem for the gateway under dynamics of the primary channel, uncertainty of the IoT devices, and unpredictability of the blockchain environment. In the problem, the gateway jointly decides (i) time scheduling, i.e., the energy harvesting time, backscatter time, and transmission time, among the IoT devices, (ii) the blockchain network, and (iii) a transaction fee rate to maximize the network throughput while minimizing the cost. To solve the stochastic optimization problem, we then propose to employ, evaluate, and assess a Deep Reinforcement Learning (DRL) algorithm with Dueling Double Deep Q-Networks (D3QN) to derive the optimal policy for the gateway. Simulation results clearly show that the proposed solution outperforms the conventional baseline approaches such as the conventional Q-Learning algorithm and non-learning algorithms in terms of network throughput and convergence speed. Furthermore, the proposed solution guarantees that the data is stored in the blockchain network at a reasonable cost.

Internet of Things Task Migration Algorithm under Edge Computing in the Design of English Translation Theory and Teaching Practice Courses

This study aims to improve the quality of English teaching in contemporary colleges and universities, so as to cultivate more English translation talents. Taking corpus, English translation, and teaching practice as the main research horizons, this study analyzes the application of master of translation and interpreting (MTI) course design, English translation, and task migration by combining the task migration algorithm under Internet of Things (IoT) with the self-built corpus, so as to realize the cultivation of translation talents and design of teaching practice. A translation corpus is constructed, a two-way interactive online course is designed, and the experimental results of the complete local migration algorithm (CLM algorithm), random migration algorithm (RM algorithm), and greedy heuristic migration algorithm (GHM algorithm) used in English teaching practice courses are analyzed and compared. The experimental simulation results reveal that the GHM algorithm proposed in this study shows good system stability, its duration is 60% better than that of the CLM algorithm, and its system throughput is increased by 50% compared with the CLM algorithm. In addition, the maximum delay time has little effect on the system throughput. When the system time slot length is fixed at 20 ms, the user migration rate of the genetic algorithm is the highest under the different total numbers of users. In addition, in view of the wide application of neural network in English translation teaching, this study establishes an English translation evaluation model based on the combination of particle swarm optimization (PSO) algorithm and neural network and compares it with the traditional neural network model in simulation experiments. The results show that the addition of PSO algorithm can effectively improve the convergence speed of artificial neural network (ANN), reduce the training time of the model, and improve the accuracy of the ANN network. Using the PSO algorithm to train the neural network, the optimal solutions of different particle swarms can be obtained, and the error is small. The PSO-ANNs model can promote the quality of English translation teaching and improve the English translation ability of the students. Therefore, applying the task transfer algorithm and PSO algorithm to the practice of English translation teaching has greatly improved the efficiency of the English classroom. To sum up, this study provides new ideas for the curriculum design of contemporary college English teaching and has reference value for the cultivation of college English translation talents.

Urban forestry detection by deep learning method with GaoFen-2 remote sensing images

Urban forestry is regarded as the green infrastructure of cities, the analysis of which is of great significance for research on statistical analysis, such as greening, carbon sink, etc. on the basis of remote sensing detection. However, current urban forestry detection algorithms are focused on typical machine learning, whose results are limited by the manually selected features. Therefore, an innovative detection algorithm for analyzing urban forestry is proposed. First, the simple linear iterative clustering algorithm is adopted to segment the dataset, and the edge information of the object is preserved as much as possible. Second, a new network structure is constructed on the basis of Mask-RCNN to refine the detection results: (1) dilated convolution is adopted to replace the normal convolution of C1 in backbone (ResNet101) to increase the size of receptive field; (2) convolutional block attention module is introduced in C2 to give more weight to the shallow features such as edges and colors; and (3) atrous spatial pyramid pooling (ASPP) is added to branch of C3, enhancing the ability of multiscale fusion. With the effect of ASPP on detection results at different positions of backbone compared, the results show that the more backward the ASPP is, the more obvious the trend of negative correlation between network convergence speed and detection accuracy (C3: 89.2%, C4: 84.3%, and C5: 81.6%) is. The accuracy of random forest and support vector machines is also compared, with the former of 69.4% and the latter of 71.2%. In addition, the method of transfer learning to unmanned aerial vehicle data is used in this research, with the accuracy of 81.6%. The result shows that the improved segmentation model features better ability of learning shallow features than the original, able to accurately extract urban forestry from remote sensing images, which provides the reference for the further research on urban forestry detection, mapping, as well as the multisource remote sensing data fusion.

Enhanced polarization demosaicking network via a precise angle of polarization loss calculation method.

Many existing polarization networks reconstruct polarization information based on calculating the angle of polarization (AoP) loss. Yet, the conventional loss calculation method, which is based on a linear difference approach, compromises the reconstruction accuracy and causes additional training time when combined with learning-based methods. In this Letter, we present a new, to the best of our knowledge, method to calculate the AoP loss and apply it in an enhanced color polarization demosaicking network with a "multi-branch" structure, i.e., ePDNet. Experiments are performed to demonstrate the efficacy and superiority of the method, which improves the network convergence speed by three times as well as the output image quality. The new method may find important applications in the field of polarimetric imaging.

A high‐performance insulators location scheme based on YOLOv4 deep learning network with GDIoU loss function

This paper proposes a Gaussian Distance Intersection over Union (GDIoU) loss function-based YOLOv4 deep learning network to solve the problem of slow speed and low accuracy insulator location in power facilities health inspection. In the scheme, A GDIoU loss function is designed to accelerate the convergence speed of the YOLOv4 deep learning network; at the same time, the GDIoU loss is added as one part of the network propagation loss, and the insulator's location accuracy is accordingly improved. Moreover, a re-location scheme for tilt insulators correction is proposed to enhance the location accuracy of the insulators in different spatial angle states. Large amounts of field insulator images were gathered as training and testing samples to evaluate the performance of the proposed scheme. The experimental results have demonstrated that the GDIoU-based YOLOv4 deep learning network combined with the tilt correction scheme can improve the insulator location speed by three times compared with the peer schemes, and the average precision is increased by 7.37% compared with the naive YOLOv4 network. The performance of the proposed scheme meets the requirement of online insulator location adequately.

Particle Swarm Optimization Based Fire Risk Valuation Model: Shopping-Mall

In view of the shortcomings of the existing small-scale shopping mall fire prediction models, the effectiveness and scalability of the prediction results, a BP neural network prediction model is constructed to improve the prediction accuracy by considering a variety of fire risk factors. On this basis, the convergence speed of the BP neural network is accelerated with the help of the particle swarm optimization (PSO) algorithm. Then, a mixed multi-factor shopping mall fire risk grade prediction model based on a PSO based back-propagation (PSO-BP) neural network model is proposed. The constructed prediction model can simultaneously consider climate factors (daily maximum temperature, daily average temperature, 24-h precipitation, continuous drought days, sunshine hours, daily average relative humidity, and daily average wind speed), landform factors (altitude, slope, slope direction, soil water content), combustible factors (vegetation type, combustible water content, ground cover load), and human factors (density of population, distance from human activity area). Based on the actual data and field measurement data collected by the sensor network of the shopping mall (Lahore, Pakistan), the validity of the proposed model was verified by a group of experiments. The results show that the model based on the training data set and the test samples can effectively predict the fire risk level; the computational complexity of the model is significantly lower than that of the BP neural network model alone.