Effect of boron on falling of prunes (Prunus mume, Sieb, et Zucc) was studied by applying 50 g borateper tree into soil on December 15, 1993 (soil-B) and spraying leaves leves evenly twice with 1.5 g kg^-1 boratesolut...Effect of boron on falling of prunes (Prunus mume, Sieb, et Zucc) was studied by applying 50 g borateper tree into soil on December 15, 1993 (soil-B) and spraying leaves leves evenly twice with 1.5 g kg^-1 boratesolution on March 1 and 8, 1994 (spray-B) on the soil with 0.28 mg kg--1’ rapidly available B. Comparedwith no borate treatment (CK), B concentrations of leaves, short branches and flowers were higher and thepercentage of flower and fruit drop was lower in the treatments of soil-B and spray-B. B fertilizer increased Bconcentrations in flowers, leaves and short branches, promoted pollen germination, reduced the percentage offall of flowers and fruits of prunes, increased the percentage of fertile fruits, and thus increased yields of prunesby 46% and 34.3% in the treatments of soil-B and spray-B, respectively. It could be inferred preliminarilythat if B concentration of leaves was lower than 35 mg kg--1, the prunes should be fertilized with B. Themeasured leaves should be picked from branches (3-10 cm in length) germinating from the central sectionof a tree crown during the last ten days of May to the early days of June.展开更多
In the field of natural language processing(NLP),there have been various pre-training language models in recent years,with question answering systems gaining significant attention.However,as algorithms,data,and comput...In the field of natural language processing(NLP),there have been various pre-training language models in recent years,with question answering systems gaining significant attention.However,as algorithms,data,and computing power advance,the issue of increasingly larger models and a growing number of parameters has surfaced.Consequently,model training has become more costly and less efficient.To enhance the efficiency and accuracy of the training process while reducing themodel volume,this paper proposes a first-order pruningmodel PAL-BERT based on the ALBERT model according to the characteristics of question-answering(QA)system and language model.Firstly,a first-order network pruning method based on the ALBERT model is designed,and the PAL-BERT model is formed.Then,the parameter optimization strategy of the PAL-BERT model is formulated,and the Mish function was used as an activation function instead of ReLU to improve the performance.Finally,after comparison experiments with traditional deep learning models TextCNN and BiLSTM,it is confirmed that PALBERT is a pruning model compression method that can significantly reduce training time and optimize training efficiency.Compared with traditional models,PAL-BERT significantly improves the NLP task’s performance.展开更多
Considering the variations in imaging sizes of the unmanned aerial vehicles(UAV)at different aerial photography heights,as well as the influence of factors such as light and weather,which can result in missed detectio...Considering the variations in imaging sizes of the unmanned aerial vehicles(UAV)at different aerial photography heights,as well as the influence of factors such as light and weather,which can result in missed detection and false detection of the model,this paper presents a comprehensive detection model based on the improved lightweight You Only Look Once version 8s(YOLOv8s)algorithm used in natural light and infrared scenes(L_YOLO).The algorithm proposes a special feature pyramid network(SFPN)structure and substitutes most of the neck feature extraction module with the Special deformable convolution feature extraction module(SDCN).Moreover,the model undergoes pruning to eliminate redundant channels.Finally,the non-maximum suppression algorithm of intersection-union ratio based on minimum point distance(MPDIOU_NMS)algorithm has been integrated to eliminate redundant detection boxes,and a comprehensive validation has been conducted using the infrared aerial dataset and the Visdrone2019 dataset.The comprehensive experimental results demonstrate that when the number of parameters and floating-point operations is reduced by 30%and 20%,respectively,there is a 1.2%increase in mean average precision at a threshold of 0.5(mAP(0.5))and a 4.8%increase in mAP(0.5:0.95)on the infrared dataset.Finally,the mAP on the Visdrone2019 dataset has experienced an average increase of 12.4%.The accuracy and recall rates have seen respective increases of 9.2%and 3.6%.展开更多
To address the challenges of high complexity,poor real-time performance,and low detection rates for small target vehicles in existing vehicle object detection algorithms,this paper proposes a real-time lightweight arc...To address the challenges of high complexity,poor real-time performance,and low detection rates for small target vehicles in existing vehicle object detection algorithms,this paper proposes a real-time lightweight architecture based on You Only Look Once(YOLO)v5m.Firstly,a lightweight upsampling operator called Content-Aware Reassembly of Features(CARAFE)is introduced in the feature fusion layer of the network to maximize the extraction of deep-level features for small target vehicles,reducing the missed detection rate and false detection rate.Secondly,a new prediction layer for tiny targets is added,and the feature fusion network is redesigned to enhance the detection capability for small targets.Finally,this paper applies L1 regularization to train the improved network,followed by pruning and fine-tuning operations to remove redundant channels,reducing computational and parameter complexity and enhancing the detection efficiency of the network.Training is conducted on the VisDrone2019-DET dataset.The experimental results show that the proposed algorithmreduces parameters and computation by 63.8% and 65.8%,respectively.The average detection accuracy improves by 5.15%,and the detection speed reaches 47 images per second,satisfying real-time requirements.Compared with existing approaches,including YOLOv5m and classical vehicle detection algorithms,our method achieves higher accuracy and faster speed for real-time detection of small target vehicles in edge computing.展开更多
Network fault diagnosis methods play a vital role in maintaining network service quality and enhancing user experience as an integral component of intelligent network management.Considering the unique characteristics ...Network fault diagnosis methods play a vital role in maintaining network service quality and enhancing user experience as an integral component of intelligent network management.Considering the unique characteristics of edge networks,such as limited resources,complex network faults,and the need for high real-time performance,enhancing and optimizing existing network fault diagnosis methods is necessary.Therefore,this paper proposes the lightweight edge-side fault diagnosis approach based on a spiking neural network(LSNN).Firstly,we use the Izhikevich neurons model to replace the Leaky Integrate and Fire(LIF)neurons model in the LSNN model.Izhikevich neurons inherit the simplicity of LIF neurons but also possess richer behavioral characteristics and flexibility to handle diverse data inputs.Inspired by Fast Spiking Interneurons(FSIs)with a high-frequency firing pattern,we use the parameters of FSIs.Secondly,inspired by the connection mode based on spiking dynamics in the basal ganglia(BG)area of the brain,we propose the pruning approach based on the FSIs of the BG in LSNN to improve computational efficiency and reduce the demand for computing resources and energy consumption.Furthermore,we propose a multiple iterative Dynamic Spike Timing Dependent Plasticity(DSTDP)algorithm to enhance the accuracy of the LSNN model.Experiments on two server fault datasets demonstrate significant precision,recall,and F1 improvements across three diagnosis dimensions.Simultaneously,lightweight indicators such as Params and FLOPs significantly reduced,showcasing the LSNN’s advanced performance and model efficiency.To conclude,experiment results on a pair of datasets indicate that the LSNN model surpasses traditional models and achieves cutting-edge outcomes in network fault diagnosis tasks.展开更多
Diseases in tea trees can result in significant losses in both the quality and quantity of tea production.Regular monitoring can help to prevent the occurrence of large-scale diseases in tea plantations.However,existi...Diseases in tea trees can result in significant losses in both the quality and quantity of tea production.Regular monitoring can help to prevent the occurrence of large-scale diseases in tea plantations.However,existingmethods face challenges such as a high number of parameters and low recognition accuracy,which hinders their application in tea plantation monitoring equipment.This paper presents a lightweight I-MobileNetV2 model for identifying diseases in tea leaves,to address these challenges.The proposed method first embeds a Coordinate Attention(CA)module into the originalMobileNetV2 network,enabling the model to locate disease regions accurately.Secondly,a Multi-branch Parallel Convolution(MPC)module is employed to extract disease features across multiple scales,improving themodel’s adaptability to different disease scales.Finally,the AutoML for Model Compression(AMC)is used to compress themodel and reduce computational complexity.Experimental results indicate that our proposed algorithm attains an average accuracy of 96.12%on our self-built tea leaf disease dataset,surpassing the original MobileNetV2 by 1.91%.Furthermore,the number of model parameters have been reduced by 40%,making itmore suitable for practical application in tea plantation environments.展开更多
Prunusmumehas high ornamental value,and its maintenance and management should be more meticulous,with pruning being an important task.Pruning can make P.mume more robust,reduce the occurrence of diseases and pests,mai...Prunusmumehas high ornamental value,and its maintenance and management should be more meticulous,with pruning being an important task.Pruning can make P.mume more robust,reduce the occurrence of diseases and pests,maintain a good shape,and promote more flowering,further improving its ornamental value.The difficulty of pruning lies in flexibly adopting suitable pruning methods according to the time of the tree,which requires understanding the impact of pruning operations on the growth and flowering of P.mume,as well as some techniques in pruning operations.This paper introduces the botanical characteristics of P.mume,common pruning methods and achievable effects of P.mume,and suitable time for using various methods,and analyzes the possible consequences and reasons of some incorrect operations.Moreover,corresponding correct practices are provided,which can provide reference for standardized pruning of P.mume,thereby reducing or avoiding losses caused by improper operation.展开更多
It is of great significance to improve the efficiency of railway production and operation by realizing the fault knowledge association through the efficient data mining algorithm.However,high utility quantitative freq...It is of great significance to improve the efficiency of railway production and operation by realizing the fault knowledge association through the efficient data mining algorithm.However,high utility quantitative frequent pattern mining algorithms in the field of data mining still suffer from the problems of low time-memory performance and are not easy to scale up.In the context of such needs,we propose a related degree-based frequent pattern mining algorithm,named Related High Utility Quantitative Item set Mining(RHUQI-Miner),to enable the effective mining of railway fault data.The algorithm constructs the item-related degree structure of fault data and gives a pruning optimization strategy to find frequent patterns with higher related degrees,reducing redundancy and invalid frequent patterns.Subsequently,it uses the fixed pattern length strategy to modify the utility information of the item in the mining process so that the algorithm can control the length of the output frequent pattern according to the actual data situation and further improve the performance and practicability of the algorithm.The experimental results on the real fault dataset show that RHUQI-Miner can effectively reduce the time and memory consumption in the mining process,thus providing data support for differentiated and precise maintenance strategies.展开更多
The burgeoning robotics industry has catalyzed significant strides in the development and deployment of industrial and service robotic arms, positioning path planning as a pivotal facet for augmenting their operationa...The burgeoning robotics industry has catalyzed significant strides in the development and deployment of industrial and service robotic arms, positioning path planning as a pivotal facet for augmenting their operational safety and efficiency. Existing path planning algorithms, while capable of delineating feasible trajectories, often fall short of achieving optimality, particularly concerning path length, search duration, and success likelihood. This study introduces an enhanced Rapidly-Exploring Random Tree (RRT) algorithm, meticulously designed to rectify the issues of node redundancy and the compromised path quality endemic to conventional RRT approaches. Through the integration of an adaptive pruning mechanism and a dynamic elliptical search strategy within the Informed RRT* framework, our algorithm efficiently refines the search tree by discarding branches that surpass the cost of the optimal path, thereby refining the search space and significantly boosting efficiency. Extensive comparative analysis across both two-dimensional and three-dimensional simulation settings underscores the algorithm’s proficiency in markedly improving path precision and search velocity, signifying a breakthrough in the domain of robotic arm path planning.展开更多
Slope stability prediction plays a significant role in landslide disaster prevention and mitigation.This paper’s reduced error pruning(REP)tree and random tree(RT)models are developed for slope stability evaluation a...Slope stability prediction plays a significant role in landslide disaster prevention and mitigation.This paper’s reduced error pruning(REP)tree and random tree(RT)models are developed for slope stability evaluation and meeting the high precision and rapidity requirements in slope engineering.The data set of this study includes five parameters,namely slope height,slope angle,cohesion,internal friction angle,and peak ground acceleration.The available data is split into two categories:training(75%)and test(25%)sets.The output of the RT and REP tree models is evaluated using performance measures including accuracy(Acc),Matthews correlation coefficient(Mcc),precision(Prec),recall(Rec),and F-score.The applications of the aforementionedmethods for predicting slope stability are compared to one another and recently established soft computing models in the literature.The analysis of the Acc together with Mcc,and F-score for the slope stability in the test set demonstrates that the RT achieved a better prediction performance with(Acc=97.1429%,Mcc=0.935,F-score for stable class=0.979 and for unstable case F-score=0.935)succeeded by the REP tree model with(Acc=95.4286%,Mcc=0.896,F-score stable class=0.967 and for unstable class F-score=0.923)for the slope stability dataset The analysis of performance measures for the slope stability dataset reveals that the RT model attains comparatively better and reliable results and thus should be encouraged in further research.展开更多
Nowadays,the rapid development of edge computing has driven an increasing number of deep learning applications deployed at the edge of the network,such as pedestrian and vehicle detection,to provide efficient intellig...Nowadays,the rapid development of edge computing has driven an increasing number of deep learning applications deployed at the edge of the network,such as pedestrian and vehicle detection,to provide efficient intelligent services to mobile users.However,as the accuracy requirements continue to increase,the components of deep learning models for pedestrian and vehicle detection,such as YOLOv4,become more sophisticated and the computing resources required for model training are increasing dramatically,which in turn leads to significant challenges in achieving effective deployment on resource-constrained edge devices while ensuring the high accuracy performance.For addressing this challenge,a cloud-edge collaboration-based pedestrian and vehicle detection framework is proposed in this paper,which enables sufficient training of models by utilizing the abundant computing resources in the cloud,and then deploying the well-trained models on edge devices,thus reducing the computing resource requirements for model training on edge devices.Furthermore,to reduce the size of the model deployed on edge devices,an automatic pruning method combines the convolution layer and BN layer is proposed to compress the pedestrian and vehicle detection model size.Experimental results show that the framework proposed in this paper is able to deploy the pruned model on a real edge device,Jetson TX2,with 6.72 times higher FPS.Meanwhile,the channel pruning reduces the volume and the number of parameters to 96.77%for the model,and the computing amount is reduced to 81.37%.展开更多
Secondary lignocellulosic biomass has proved to be useful as an energy source through its oxidation by means of combustion processes.In accordance with the above,in this paper,we wanted to study the ash from urban pru...Secondary lignocellulosic biomass has proved to be useful as an energy source through its oxidation by means of combustion processes.In accordance with the above,in this paper,we wanted to study the ash from urban pruning residues that are generated in cities in the Neotropics.Species such as Licania tomentosa,Azadirachta indica,Ficus benjamina,Terminalia catappa,Leucaena leucocephala,Prosopis juliflora and Pithecellobium dulce were selected because they have been previously studied and showed potential for thermal energy generation.These materials were calcined in an oxidizing atmosphere and characterized by X-ray diffraction and fluorescence,scanning electron microscopy with microchemistry,BET surface area,thermal gravimetric analysis,and differential scanning calorimetry.The pH and apparent density were also established.The results show high basicity materials(average pH 10),a behavior associated with the presence of chemical elements such as calcium,potassium,magnesium,chlorine,phosphorus,and sulfur.Structurally,these materials have a very significant amorphous fraction(between 49%and 74.5%),the dominant crystalline phases are calcite,arcanite,sylvite,and hydroxyapatite.These ashes have low surface area and do not exceed 13 m^(2)/g.Two characteristic morphological aspects were observed in these ashes:a morphology of rounded grains where silicon content is highlighted,and lamellar morphologies where the presence of chlorine is highlighted.Thermally,these ashes show four significant mass loss events(400℃,430℃,680℃,and 920℃),causing mass losses that vary between 25%and 40%.Through this study,it was possible to establish that,from a chemical point of view,these ashes are less dangerous in comparison with those of a mineral coal that was used as a reference.However,they require additional treatments for their disposal due to their high basicity.Because of their composition,these ashes have the potential to be used in the ceramic and cement industries,and in the manufacture of fertilizers.展开更多
Automatic speech recognition(ASR)systems have emerged as indispensable tools across a wide spectrum of applications,ranging from transcription services to voice-activated assistants.To enhance the performance of these...Automatic speech recognition(ASR)systems have emerged as indispensable tools across a wide spectrum of applications,ranging from transcription services to voice-activated assistants.To enhance the performance of these systems,it is important to deploy efficient models capable of adapting to diverse deployment conditions.In recent years,on-demand pruning methods have obtained significant attention within the ASR domain due to their adaptability in various deployment scenarios.However,these methods often confront substantial trade-offs,particularly in terms of unstable accuracy when reducing the model size.To address challenges,this study introduces two crucial empirical findings.Firstly,it proposes the incorporation of an online distillation mechanism during on-demand pruning training,which holds the promise of maintaining more consistent accuracy levels.Secondly,it proposes the utilization of the Mogrifier long short-term memory(LSTM)language model(LM),an advanced iteration of the conventional LSTM LM,as an effective alternative for pruning targets within the ASR framework.Through rigorous experimentation on the ASR system,employing the Mogrifier LSTM LM and training it using the suggested joint on-demand pruning and online distillation method,this study provides compelling evidence.The results exhibit that the proposed methods significantly outperform a benchmark model trained solely with on-demand pruning methods.Impressively,the proposed strategic configuration successfully reduces the parameter count by approximately 39%,all the while minimizing trade-offs.展开更多
A probabilistic multi-dimensional selective ensemble learning method and its application in the prediction of users' online purchase behavior are studied in this work.Firstly, the classifier is integrated based on...A probabilistic multi-dimensional selective ensemble learning method and its application in the prediction of users' online purchase behavior are studied in this work.Firstly, the classifier is integrated based on the dimension of predicted probability, and the pruning algorithm based on greedy forward search is obtained by combining the two indicators of accuracy and complementarity.Then the pruning algorithm is integrated into the Stacking ensemble method to establish a user online shopping behavior prediction model based on the probabilistic multi-dimensional selective ensemble method.Finally, the research method is compared with the prediction results of individual learners in ensemble learning and the Stacking ensemble method without pruning.The experimental results show that the proposed method can reduce the scale of integration, improve the prediction accuracy of the model, and predict the user's online purchase behavior.展开更多
The oil palm leaf miner, Coelaenomenodera lameensis, is currently the most destructive pest of oil palm in Ghana and other African oil palm growing countries, causing significant losses in fresh fruit bunch yield. Pro...The oil palm leaf miner, Coelaenomenodera lameensis, is currently the most destructive pest of oil palm in Ghana and other African oil palm growing countries, causing significant losses in fresh fruit bunch yield. Progressive pruning is an oil palm pruning method in which pruning is done at the same time as fresh fruit bunch harvesting. This study evaluated the impact of progressive pruning on leaf miner population in oil palm and how these two factors (leaf miner and progressive pruning) affect the yield of oil palm at the Benso Oil Palm Plantation Public listed company (BOPP. Plc). Five distinct blocks in the plantation were selected for observations on fronds at various ranks (33, 25, or 17) based on the degree of defoliation by counting the number of pests on leaflets at different phases of insect development. Fronds from selected plots were sampled in a Completely Randomized Design (CRD). The size of plots used for the study ranged between 19 to 45 hectares. A minimum of 78 fronds were evenly cut from each block for pest count depending on the block size. Secondary data on annual yields of fresh fruit bunches before and after the introduction of progressive pruning were also obtained from BOPP. Plc records from 2011-2020. The results from the analyzed data on leaf miner index before and after the introduction of progressive pruning showed that progressive pruning has, to a high extent (64% to 36%), reduced leaf miner populations in the plantation. Paired t-test on fresh fruit bunch yield has also revealed a significant (p < 0.001) increase in annual fresh fruit bunch yield due to progressive pruning. A regression analysis, however, revealed a lower rate of yield loss (3.05 to 2.70 tonnes) to leaf miner infestation after the introduction of progressive pruning. The study recommends progressive pruning as a key cultural practice for improving crop yields in leaf miner prone plantations.展开更多
According to the production experience,the author summarizes the cultivation techniques of Qiuyue pear from orchard construction,shaping and pruning,fruit management,underground management,coping with natural disaster...According to the production experience,the author summarizes the cultivation techniques of Qiuyue pear from orchard construction,shaping and pruning,fruit management,underground management,coping with natural disasters,and pest control,in order to provide a reference for producers.展开更多
As computers have become faster at performing computations over the decades, algorithms to play games have also become more efficient. This research paper seeks to see how the performance of the Minimax search evolves...As computers have become faster at performing computations over the decades, algorithms to play games have also become more efficient. This research paper seeks to see how the performance of the Minimax search evolves on increasing Connect-4 grid sizes. The objective of this study is to evaluate the effectiveness of the Minimax search algorithm in making optimal moves under different circumstances and to understand how well the algorithm scales. To answer this question we tested and analyzed the algorithm several times on different grid sizes with a time limit to see its performance as the complexity increases, we also looked for the average search depth for each grid size. The obtained results show that despite larger grid sizes, the Minimax search algorithm stays relatively consistent in terms of performance.展开更多
Convolutional neural networks continually evolve to enhance accuracy in addressing various problems,leading to an increase in computational cost and model size.This paper introduces a novel approach for pruning face r...Convolutional neural networks continually evolve to enhance accuracy in addressing various problems,leading to an increase in computational cost and model size.This paper introduces a novel approach for pruning face recognition models based on convolutional neural networks.The proposed method identifies and removes inefficient filters based on the information volume in feature maps.In each layer,some feature maps lack useful information,and there exists a correlation between certain feature maps.Filters associated with these two types of feature maps impose additional computational costs on the model.By eliminating filters related to these categories of feature maps,the reduction of both computational cost and model size can be achieved.The approach employs a combination of correlation analysis and the summation of matrix elements within each feature map to detect and eliminate inefficient filters.The method was applied to two face recognition models utilizing the VGG16 and ResNet50V2 backbone architectures.In the proposed approach,the number of filters removed in each layer varies,and the removal process is independent of the adjacent layers.The convolutional layers of both backbone models were initialized with pre-trained weights from ImageNet.For training,the CASIA-WebFace dataset was utilized,and the Labeled Faces in the Wild(LFW)dataset was employed for benchmarking purposes.In the VGG16-based face recognition model,a 0.74%accuracy improvement was achieved while reducing the number of convolution parameters by 26.85%and decreasing Floating-point operations per second(FLOPs)by 47.96%.For the face recognition model based on the ResNet50V2 architecture,the ArcFace method was implemented.The removal of inactive filters in this model led to a slight decrease in accuracy by 0.11%.However,it resulted in enhanced training speed,a reduction of 59.38%in convolution parameters,and a 57.29%decrease in FLOPs.展开更多
文摘Effect of boron on falling of prunes (Prunus mume, Sieb, et Zucc) was studied by applying 50 g borateper tree into soil on December 15, 1993 (soil-B) and spraying leaves leves evenly twice with 1.5 g kg^-1 boratesolution on March 1 and 8, 1994 (spray-B) on the soil with 0.28 mg kg--1’ rapidly available B. Comparedwith no borate treatment (CK), B concentrations of leaves, short branches and flowers were higher and thepercentage of flower and fruit drop was lower in the treatments of soil-B and spray-B. B fertilizer increased Bconcentrations in flowers, leaves and short branches, promoted pollen germination, reduced the percentage offall of flowers and fruits of prunes, increased the percentage of fertile fruits, and thus increased yields of prunesby 46% and 34.3% in the treatments of soil-B and spray-B, respectively. It could be inferred preliminarilythat if B concentration of leaves was lower than 35 mg kg--1, the prunes should be fertilized with B. Themeasured leaves should be picked from branches (3-10 cm in length) germinating from the central sectionof a tree crown during the last ten days of May to the early days of June.
基金Supported by Sichuan Science and Technology Program(2021YFQ0003,2023YFSY0026,2023YFH0004).
文摘In the field of natural language processing(NLP),there have been various pre-training language models in recent years,with question answering systems gaining significant attention.However,as algorithms,data,and computing power advance,the issue of increasingly larger models and a growing number of parameters has surfaced.Consequently,model training has become more costly and less efficient.To enhance the efficiency and accuracy of the training process while reducing themodel volume,this paper proposes a first-order pruningmodel PAL-BERT based on the ALBERT model according to the characteristics of question-answering(QA)system and language model.Firstly,a first-order network pruning method based on the ALBERT model is designed,and the PAL-BERT model is formed.Then,the parameter optimization strategy of the PAL-BERT model is formulated,and the Mish function was used as an activation function instead of ReLU to improve the performance.Finally,after comparison experiments with traditional deep learning models TextCNN and BiLSTM,it is confirmed that PALBERT is a pruning model compression method that can significantly reduce training time and optimize training efficiency.Compared with traditional models,PAL-BERT significantly improves the NLP task’s performance.
文摘Considering the variations in imaging sizes of the unmanned aerial vehicles(UAV)at different aerial photography heights,as well as the influence of factors such as light and weather,which can result in missed detection and false detection of the model,this paper presents a comprehensive detection model based on the improved lightweight You Only Look Once version 8s(YOLOv8s)algorithm used in natural light and infrared scenes(L_YOLO).The algorithm proposes a special feature pyramid network(SFPN)structure and substitutes most of the neck feature extraction module with the Special deformable convolution feature extraction module(SDCN).Moreover,the model undergoes pruning to eliminate redundant channels.Finally,the non-maximum suppression algorithm of intersection-union ratio based on minimum point distance(MPDIOU_NMS)algorithm has been integrated to eliminate redundant detection boxes,and a comprehensive validation has been conducted using the infrared aerial dataset and the Visdrone2019 dataset.The comprehensive experimental results demonstrate that when the number of parameters and floating-point operations is reduced by 30%and 20%,respectively,there is a 1.2%increase in mean average precision at a threshold of 0.5(mAP(0.5))and a 4.8%increase in mAP(0.5:0.95)on the infrared dataset.Finally,the mAP on the Visdrone2019 dataset has experienced an average increase of 12.4%.The accuracy and recall rates have seen respective increases of 9.2%and 3.6%.
基金funded by the General Project of Key Research and Develop-ment Plan of Shaanxi Province(No.2022NY-087).
文摘To address the challenges of high complexity,poor real-time performance,and low detection rates for small target vehicles in existing vehicle object detection algorithms,this paper proposes a real-time lightweight architecture based on You Only Look Once(YOLO)v5m.Firstly,a lightweight upsampling operator called Content-Aware Reassembly of Features(CARAFE)is introduced in the feature fusion layer of the network to maximize the extraction of deep-level features for small target vehicles,reducing the missed detection rate and false detection rate.Secondly,a new prediction layer for tiny targets is added,and the feature fusion network is redesigned to enhance the detection capability for small targets.Finally,this paper applies L1 regularization to train the improved network,followed by pruning and fine-tuning operations to remove redundant channels,reducing computational and parameter complexity and enhancing the detection efficiency of the network.Training is conducted on the VisDrone2019-DET dataset.The experimental results show that the proposed algorithmreduces parameters and computation by 63.8% and 65.8%,respectively.The average detection accuracy improves by 5.15%,and the detection speed reaches 47 images per second,satisfying real-time requirements.Compared with existing approaches,including YOLOv5m and classical vehicle detection algorithms,our method achieves higher accuracy and faster speed for real-time detection of small target vehicles in edge computing.
基金supported by National Key R&D Program of China(2019YFB2103202).
文摘Network fault diagnosis methods play a vital role in maintaining network service quality and enhancing user experience as an integral component of intelligent network management.Considering the unique characteristics of edge networks,such as limited resources,complex network faults,and the need for high real-time performance,enhancing and optimizing existing network fault diagnosis methods is necessary.Therefore,this paper proposes the lightweight edge-side fault diagnosis approach based on a spiking neural network(LSNN).Firstly,we use the Izhikevich neurons model to replace the Leaky Integrate and Fire(LIF)neurons model in the LSNN model.Izhikevich neurons inherit the simplicity of LIF neurons but also possess richer behavioral characteristics and flexibility to handle diverse data inputs.Inspired by Fast Spiking Interneurons(FSIs)with a high-frequency firing pattern,we use the parameters of FSIs.Secondly,inspired by the connection mode based on spiking dynamics in the basal ganglia(BG)area of the brain,we propose the pruning approach based on the FSIs of the BG in LSNN to improve computational efficiency and reduce the demand for computing resources and energy consumption.Furthermore,we propose a multiple iterative Dynamic Spike Timing Dependent Plasticity(DSTDP)algorithm to enhance the accuracy of the LSNN model.Experiments on two server fault datasets demonstrate significant precision,recall,and F1 improvements across three diagnosis dimensions.Simultaneously,lightweight indicators such as Params and FLOPs significantly reduced,showcasing the LSNN’s advanced performance and model efficiency.To conclude,experiment results on a pair of datasets indicate that the LSNN model surpasses traditional models and achieves cutting-edge outcomes in network fault diagnosis tasks.
基金supported by National Key Research and Development Program(No.2016YFD0201305-07)Guizhou Provincial Basic Research Program(Natural Science)(No.ZK[2023]060)Open Fund Project in Semiconductor Power Device Reliability Engineering Center of Ministry of Education(No.ERCMEKFJJ2019-06).
文摘Diseases in tea trees can result in significant losses in both the quality and quantity of tea production.Regular monitoring can help to prevent the occurrence of large-scale diseases in tea plantations.However,existingmethods face challenges such as a high number of parameters and low recognition accuracy,which hinders their application in tea plantation monitoring equipment.This paper presents a lightweight I-MobileNetV2 model for identifying diseases in tea leaves,to address these challenges.The proposed method first embeds a Coordinate Attention(CA)module into the originalMobileNetV2 network,enabling the model to locate disease regions accurately.Secondly,a Multi-branch Parallel Convolution(MPC)module is employed to extract disease features across multiple scales,improving themodel’s adaptability to different disease scales.Finally,the AutoML for Model Compression(AMC)is used to compress themodel and reduce computational complexity.Experimental results indicate that our proposed algorithm attains an average accuracy of 96.12%on our self-built tea leaf disease dataset,surpassing the original MobileNetV2 by 1.91%.Furthermore,the number of model parameters have been reduced by 40%,making itmore suitable for practical application in tea plantation environments.
文摘Prunusmumehas high ornamental value,and its maintenance and management should be more meticulous,with pruning being an important task.Pruning can make P.mume more robust,reduce the occurrence of diseases and pests,maintain a good shape,and promote more flowering,further improving its ornamental value.The difficulty of pruning lies in flexibly adopting suitable pruning methods according to the time of the tree,which requires understanding the impact of pruning operations on the growth and flowering of P.mume,as well as some techniques in pruning operations.This paper introduces the botanical characteristics of P.mume,common pruning methods and achievable effects of P.mume,and suitable time for using various methods,and analyzes the possible consequences and reasons of some incorrect operations.Moreover,corresponding correct practices are provided,which can provide reference for standardized pruning of P.mume,thereby reducing or avoiding losses caused by improper operation.
基金supported by the Research on Key Technologies and Typical Applications of Big Data in Railway Production and Operation(P2023S006)the Fundamental Research Funds for the Central Universities(2022JBZY023).
文摘It is of great significance to improve the efficiency of railway production and operation by realizing the fault knowledge association through the efficient data mining algorithm.However,high utility quantitative frequent pattern mining algorithms in the field of data mining still suffer from the problems of low time-memory performance and are not easy to scale up.In the context of such needs,we propose a related degree-based frequent pattern mining algorithm,named Related High Utility Quantitative Item set Mining(RHUQI-Miner),to enable the effective mining of railway fault data.The algorithm constructs the item-related degree structure of fault data and gives a pruning optimization strategy to find frequent patterns with higher related degrees,reducing redundancy and invalid frequent patterns.Subsequently,it uses the fixed pattern length strategy to modify the utility information of the item in the mining process so that the algorithm can control the length of the output frequent pattern according to the actual data situation and further improve the performance and practicability of the algorithm.The experimental results on the real fault dataset show that RHUQI-Miner can effectively reduce the time and memory consumption in the mining process,thus providing data support for differentiated and precise maintenance strategies.
文摘The burgeoning robotics industry has catalyzed significant strides in the development and deployment of industrial and service robotic arms, positioning path planning as a pivotal facet for augmenting their operational safety and efficiency. Existing path planning algorithms, while capable of delineating feasible trajectories, often fall short of achieving optimality, particularly concerning path length, search duration, and success likelihood. This study introduces an enhanced Rapidly-Exploring Random Tree (RRT) algorithm, meticulously designed to rectify the issues of node redundancy and the compromised path quality endemic to conventional RRT approaches. Through the integration of an adaptive pruning mechanism and a dynamic elliptical search strategy within the Informed RRT* framework, our algorithm efficiently refines the search tree by discarding branches that surpass the cost of the optimal path, thereby refining the search space and significantly boosting efficiency. Extensive comparative analysis across both two-dimensional and three-dimensional simulation settings underscores the algorithm’s proficiency in markedly improving path precision and search velocity, signifying a breakthrough in the domain of robotic arm path planning.
基金supported by the National Key Research and Development Plan of China under Grant No.2021YFB2600703.
文摘Slope stability prediction plays a significant role in landslide disaster prevention and mitigation.This paper’s reduced error pruning(REP)tree and random tree(RT)models are developed for slope stability evaluation and meeting the high precision and rapidity requirements in slope engineering.The data set of this study includes five parameters,namely slope height,slope angle,cohesion,internal friction angle,and peak ground acceleration.The available data is split into two categories:training(75%)and test(25%)sets.The output of the RT and REP tree models is evaluated using performance measures including accuracy(Acc),Matthews correlation coefficient(Mcc),precision(Prec),recall(Rec),and F-score.The applications of the aforementionedmethods for predicting slope stability are compared to one another and recently established soft computing models in the literature.The analysis of the Acc together with Mcc,and F-score for the slope stability in the test set demonstrates that the RT achieved a better prediction performance with(Acc=97.1429%,Mcc=0.935,F-score for stable class=0.979 and for unstable case F-score=0.935)succeeded by the REP tree model with(Acc=95.4286%,Mcc=0.896,F-score stable class=0.967 and for unstable class F-score=0.923)for the slope stability dataset The analysis of performance measures for the slope stability dataset reveals that the RT model attains comparatively better and reliable results and thus should be encouraged in further research.
基金supported by Key-Area Research and Development Program of Guangdong Province(2021B0101420002)the Major Key Project of PCL(PCL2021A09)+3 种基金National Natural Science Foundation of China(62072187)Guangdong Major Project of Basic and Applied Basic Research(2019B030302002)Guangdong Marine Economic Development Special Fund Project(GDNRC[2022]17)Guangzhou Development Zone Science and Technology(2021GH10,2020GH10).
文摘Nowadays,the rapid development of edge computing has driven an increasing number of deep learning applications deployed at the edge of the network,such as pedestrian and vehicle detection,to provide efficient intelligent services to mobile users.However,as the accuracy requirements continue to increase,the components of deep learning models for pedestrian and vehicle detection,such as YOLOv4,become more sophisticated and the computing resources required for model training are increasing dramatically,which in turn leads to significant challenges in achieving effective deployment on resource-constrained edge devices while ensuring the high accuracy performance.For addressing this challenge,a cloud-edge collaboration-based pedestrian and vehicle detection framework is proposed in this paper,which enables sufficient training of models by utilizing the abundant computing resources in the cloud,and then deploying the well-trained models on edge devices,thus reducing the computing resource requirements for model training on edge devices.Furthermore,to reduce the size of the model deployed on edge devices,an automatic pruning method combines the convolution layer and BN layer is proposed to compress the pedestrian and vehicle detection model size.Experimental results show that the framework proposed in this paper is able to deploy the pruned model on a real edge device,Jetson TX2,with 6.72 times higher FPS.Meanwhile,the channel pruning reduces the volume and the number of parameters to 96.77%for the model,and the computing amount is reduced to 81.37%.
基金Ministry of Science,Technology and Innovation of Colombia through the“Fondo Francisco Joséde Caldas”National Financing Fund for Science,Technology and Innovation for the financing provided for the development of the project (Project 120885272102,Call 852 of 2019).
文摘Secondary lignocellulosic biomass has proved to be useful as an energy source through its oxidation by means of combustion processes.In accordance with the above,in this paper,we wanted to study the ash from urban pruning residues that are generated in cities in the Neotropics.Species such as Licania tomentosa,Azadirachta indica,Ficus benjamina,Terminalia catappa,Leucaena leucocephala,Prosopis juliflora and Pithecellobium dulce were selected because they have been previously studied and showed potential for thermal energy generation.These materials were calcined in an oxidizing atmosphere and characterized by X-ray diffraction and fluorescence,scanning electron microscopy with microchemistry,BET surface area,thermal gravimetric analysis,and differential scanning calorimetry.The pH and apparent density were also established.The results show high basicity materials(average pH 10),a behavior associated with the presence of chemical elements such as calcium,potassium,magnesium,chlorine,phosphorus,and sulfur.Structurally,these materials have a very significant amorphous fraction(between 49%and 74.5%),the dominant crystalline phases are calcite,arcanite,sylvite,and hydroxyapatite.These ashes have low surface area and do not exceed 13 m^(2)/g.Two characteristic morphological aspects were observed in these ashes:a morphology of rounded grains where silicon content is highlighted,and lamellar morphologies where the presence of chlorine is highlighted.Thermally,these ashes show four significant mass loss events(400℃,430℃,680℃,and 920℃),causing mass losses that vary between 25%and 40%.Through this study,it was possible to establish that,from a chemical point of view,these ashes are less dangerous in comparison with those of a mineral coal that was used as a reference.However,they require additional treatments for their disposal due to their high basicity.Because of their composition,these ashes have the potential to be used in the ceramic and cement industries,and in the manufacture of fertilizers.
基金supported by Institute of Information&communications Technology Planning&Evaluation(IITP)grant funded by the Korea government(MSIT)(No.2022-0-00377,Development of Intelligent Analysis and Classification Based Contents Class Categorization Technique to Prevent Imprudent Harmful Media Distribution).
文摘Automatic speech recognition(ASR)systems have emerged as indispensable tools across a wide spectrum of applications,ranging from transcription services to voice-activated assistants.To enhance the performance of these systems,it is important to deploy efficient models capable of adapting to diverse deployment conditions.In recent years,on-demand pruning methods have obtained significant attention within the ASR domain due to their adaptability in various deployment scenarios.However,these methods often confront substantial trade-offs,particularly in terms of unstable accuracy when reducing the model size.To address challenges,this study introduces two crucial empirical findings.Firstly,it proposes the incorporation of an online distillation mechanism during on-demand pruning training,which holds the promise of maintaining more consistent accuracy levels.Secondly,it proposes the utilization of the Mogrifier long short-term memory(LSTM)language model(LM),an advanced iteration of the conventional LSTM LM,as an effective alternative for pruning targets within the ASR framework.Through rigorous experimentation on the ASR system,employing the Mogrifier LSTM LM and training it using the suggested joint on-demand pruning and online distillation method,this study provides compelling evidence.The results exhibit that the proposed methods significantly outperform a benchmark model trained solely with on-demand pruning methods.Impressively,the proposed strategic configuration successfully reduces the parameter count by approximately 39%,all the while minimizing trade-offs.
基金Supported by the Scientific Research Foundation of Liaoning Provincial Department of Education (No.LJKZ0139)。
文摘A probabilistic multi-dimensional selective ensemble learning method and its application in the prediction of users' online purchase behavior are studied in this work.Firstly, the classifier is integrated based on the dimension of predicted probability, and the pruning algorithm based on greedy forward search is obtained by combining the two indicators of accuracy and complementarity.Then the pruning algorithm is integrated into the Stacking ensemble method to establish a user online shopping behavior prediction model based on the probabilistic multi-dimensional selective ensemble method.Finally, the research method is compared with the prediction results of individual learners in ensemble learning and the Stacking ensemble method without pruning.The experimental results show that the proposed method can reduce the scale of integration, improve the prediction accuracy of the model, and predict the user's online purchase behavior.
文摘The oil palm leaf miner, Coelaenomenodera lameensis, is currently the most destructive pest of oil palm in Ghana and other African oil palm growing countries, causing significant losses in fresh fruit bunch yield. Progressive pruning is an oil palm pruning method in which pruning is done at the same time as fresh fruit bunch harvesting. This study evaluated the impact of progressive pruning on leaf miner population in oil palm and how these two factors (leaf miner and progressive pruning) affect the yield of oil palm at the Benso Oil Palm Plantation Public listed company (BOPP. Plc). Five distinct blocks in the plantation were selected for observations on fronds at various ranks (33, 25, or 17) based on the degree of defoliation by counting the number of pests on leaflets at different phases of insect development. Fronds from selected plots were sampled in a Completely Randomized Design (CRD). The size of plots used for the study ranged between 19 to 45 hectares. A minimum of 78 fronds were evenly cut from each block for pest count depending on the block size. Secondary data on annual yields of fresh fruit bunches before and after the introduction of progressive pruning were also obtained from BOPP. Plc records from 2011-2020. The results from the analyzed data on leaf miner index before and after the introduction of progressive pruning showed that progressive pruning has, to a high extent (64% to 36%), reduced leaf miner populations in the plantation. Paired t-test on fresh fruit bunch yield has also revealed a significant (p < 0.001) increase in annual fresh fruit bunch yield due to progressive pruning. A regression analysis, however, revealed a lower rate of yield loss (3.05 to 2.70 tonnes) to leaf miner infestation after the introduction of progressive pruning. The study recommends progressive pruning as a key cultural practice for improving crop yields in leaf miner prone plantations.
文摘According to the production experience,the author summarizes the cultivation techniques of Qiuyue pear from orchard construction,shaping and pruning,fruit management,underground management,coping with natural disasters,and pest control,in order to provide a reference for producers.
文摘As computers have become faster at performing computations over the decades, algorithms to play games have also become more efficient. This research paper seeks to see how the performance of the Minimax search evolves on increasing Connect-4 grid sizes. The objective of this study is to evaluate the effectiveness of the Minimax search algorithm in making optimal moves under different circumstances and to understand how well the algorithm scales. To answer this question we tested and analyzed the algorithm several times on different grid sizes with a time limit to see its performance as the complexity increases, we also looked for the average search depth for each grid size. The obtained results show that despite larger grid sizes, the Minimax search algorithm stays relatively consistent in terms of performance.
文摘Convolutional neural networks continually evolve to enhance accuracy in addressing various problems,leading to an increase in computational cost and model size.This paper introduces a novel approach for pruning face recognition models based on convolutional neural networks.The proposed method identifies and removes inefficient filters based on the information volume in feature maps.In each layer,some feature maps lack useful information,and there exists a correlation between certain feature maps.Filters associated with these two types of feature maps impose additional computational costs on the model.By eliminating filters related to these categories of feature maps,the reduction of both computational cost and model size can be achieved.The approach employs a combination of correlation analysis and the summation of matrix elements within each feature map to detect and eliminate inefficient filters.The method was applied to two face recognition models utilizing the VGG16 and ResNet50V2 backbone architectures.In the proposed approach,the number of filters removed in each layer varies,and the removal process is independent of the adjacent layers.The convolutional layers of both backbone models were initialized with pre-trained weights from ImageNet.For training,the CASIA-WebFace dataset was utilized,and the Labeled Faces in the Wild(LFW)dataset was employed for benchmarking purposes.In the VGG16-based face recognition model,a 0.74%accuracy improvement was achieved while reducing the number of convolution parameters by 26.85%and decreasing Floating-point operations per second(FLOPs)by 47.96%.For the face recognition model based on the ResNet50V2 architecture,the ArcFace method was implemented.The removal of inactive filters in this model led to a slight decrease in accuracy by 0.11%.However,it resulted in enhanced training speed,a reduction of 59.38%in convolution parameters,and a 57.29%decrease in FLOPs.