Pub Date : 2023-05-26DOI: 10.1109/CCAI57533.2023.10201253
Yuxiang Liu, Yinghua Zhou, Xiaodan Liu
In recent years, seismic data denoising has attracted more and more scholars' attention and research, and the suppression of random noise is the key to improving the signal-to-noise ratio of seismic data. Aiming at the problem that traditional denoising methods are difficult to effectively remove a large amount of random noise and retain effective signals, we propose a neural network model based on dual channel residual attention network (DCRANet). Specifically, the model consists of a residual attention block (RAB), a dilated convolution sparse block (DCSB) and a feature enhancement block (FEB). The residual blocks in RAB can avoid some problems such as gradient vanishing and gradient exploding when the network is too deep, and the use of attention mechanism can guide the network to effectively extract complex noise information. The DCSB recovers the useful details from complex noise information by expanding the receptive field, fully acquiring important structural information and edge features of seismic data. The FEB integrates the noise features extracted by RAB and DCSB, it uses convolutional layers to extract the noise information of seismic data, and finally reconstructs clean seismic data image by the residual learning strategy. Compared with NL-Bayes, BM3D, DnCNN, CBDNet and DudeNet, DCRANet effectively suppresses random noise while retaining more local details and obtains a higher average peak signal-to-noise ratio (PSNR) and average structural similarity (SSIM).
{"title":"Research on Seismic Data Denoising Based on Dual Channel Residual Attention Network","authors":"Yuxiang Liu, Yinghua Zhou, Xiaodan Liu","doi":"10.1109/CCAI57533.2023.10201253","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201253","url":null,"abstract":"In recent years, seismic data denoising has attracted more and more scholars' attention and research, and the suppression of random noise is the key to improving the signal-to-noise ratio of seismic data. Aiming at the problem that traditional denoising methods are difficult to effectively remove a large amount of random noise and retain effective signals, we propose a neural network model based on dual channel residual attention network (DCRANet). Specifically, the model consists of a residual attention block (RAB), a dilated convolution sparse block (DCSB) and a feature enhancement block (FEB). The residual blocks in RAB can avoid some problems such as gradient vanishing and gradient exploding when the network is too deep, and the use of attention mechanism can guide the network to effectively extract complex noise information. The DCSB recovers the useful details from complex noise information by expanding the receptive field, fully acquiring important structural information and edge features of seismic data. The FEB integrates the noise features extracted by RAB and DCSB, it uses convolutional layers to extract the noise information of seismic data, and finally reconstructs clean seismic data image by the residual learning strategy. Compared with NL-Bayes, BM3D, DnCNN, CBDNet and DudeNet, DCRANet effectively suppresses random noise while retaining more local details and obtains a higher average peak signal-to-noise ratio (PSNR) and average structural similarity (SSIM).","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125528970","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-26DOI: 10.1109/CCAI57533.2023.10201299
Xiaotian Yu
The traditional setting for pure exploration of multi-armed bandits is to identify an optimal arm in a decision set, which contains a finite number of stochastic slot machines. The finite-arm setting restricts classic bandit algorithms, because the decision set for optimal selection can be continuous and infinite in many practical applications, e.g., determining the optimal parameter in communication networks. In this paper, to generalize bandits into wider real scenarios, we focus on the problem of pure exploration of Continuum-Armed Bandits (CAB), where the decision set is a compact and continuous set. Compared to the traditional setting of pure exploration, identifying the optimal arm in CAB raises new challenges, of which the most notorious one is the infinite number of arms. By fully taking advantage of the structure information of payoffs, we successfully solve the challenges. In particular, we derive an upper bound of sample complexity for pure exploration of CAB with concave structures via gradient methodology. More importantly, we develop a warm-restart algorithm to solve the problem where a quadratic growth condition is further satisfied, and derive an improved upper bound of sample complexity. Finally, we conduct experiments with real-world oracles to demonstrate the superiority of our warm-restart algorithm.
{"title":"Pure Exploration of Continuum-Armed Bandits under Concavity and Quadratic Growth Conditions","authors":"Xiaotian Yu","doi":"10.1109/CCAI57533.2023.10201299","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201299","url":null,"abstract":"The traditional setting for pure exploration of multi-armed bandits is to identify an optimal arm in a decision set, which contains a finite number of stochastic slot machines. The finite-arm setting restricts classic bandit algorithms, because the decision set for optimal selection can be continuous and infinite in many practical applications, e.g., determining the optimal parameter in communication networks. In this paper, to generalize bandits into wider real scenarios, we focus on the problem of pure exploration of Continuum-Armed Bandits (CAB), where the decision set is a compact and continuous set. Compared to the traditional setting of pure exploration, identifying the optimal arm in CAB raises new challenges, of which the most notorious one is the infinite number of arms. By fully taking advantage of the structure information of payoffs, we successfully solve the challenges. In particular, we derive an upper bound of sample complexity for pure exploration of CAB with concave structures via gradient methodology. More importantly, we develop a warm-restart algorithm to solve the problem where a quadratic growth condition is further satisfied, and derive an improved upper bound of sample complexity. Finally, we conduct experiments with real-world oracles to demonstrate the superiority of our warm-restart algorithm.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129629632","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-26DOI: 10.1109/CCAI57533.2023.10201293
Shuai Li, Linze Wang, Shengzhe Xu, D. Gao
Needle insertion is a minimally invasive treatment technique. During the surgery, the insertion path needs to be planned in advance to manipulate the flexible needle to avoid nerves and organs. In order to predict the insertion trajectory, a force-visual perception prediction model based on the BP neural network is established. Through the force analysis of the flexible needle, the displacement L of the needle holder and the reaction force Fr, insertion force F and torque M on the needle holder are used as the force visual perception model. Input to predict the trajectory of the needle tip during insertion. Through experiments on three different types of flexible needles, data are collected to train the model. The lowest mean absolute error (MAE) of the model is 0.7490, the correlation coefficient R is between 0.99962 and 0.99996, and the accuracy is high. The force visual perception model provides a feasible prediction of the needle tip trajectory. The results show that the displacement of the needle tip in the X and Y directions predicted by the model is basically consistent with the experimental results, and the insertion trajectory can be predicted more accurately.
{"title":"Research on Force Visual Perception of Bevel-Tip Needle Based on BP Neural Network","authors":"Shuai Li, Linze Wang, Shengzhe Xu, D. Gao","doi":"10.1109/CCAI57533.2023.10201293","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201293","url":null,"abstract":"Needle insertion is a minimally invasive treatment technique. During the surgery, the insertion path needs to be planned in advance to manipulate the flexible needle to avoid nerves and organs. In order to predict the insertion trajectory, a force-visual perception prediction model based on the BP neural network is established. Through the force analysis of the flexible needle, the displacement L of the needle holder and the reaction force Fr, insertion force F and torque M on the needle holder are used as the force visual perception model. Input to predict the trajectory of the needle tip during insertion. Through experiments on three different types of flexible needles, data are collected to train the model. The lowest mean absolute error (MAE) of the model is 0.7490, the correlation coefficient R is between 0.99962 and 0.99996, and the accuracy is high. The force visual perception model provides a feasible prediction of the needle tip trajectory. The results show that the displacement of the needle tip in the X and Y directions predicted by the model is basically consistent with the experimental results, and the insertion trajectory can be predicted more accurately.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133553159","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-26DOI: 10.1109/CCAI57533.2023.10201328
Deheng Li, Qingyun Meng, Yi Liu, Wanyi Zhu
The emergence of the fifth-generation (5G) technology will accelerate the digitization of the economy and society, but the public network-based 5G core network (5GC) cannot meet the usage needs of vertical industries, while the cloud-native lightweight 5GC can be customized according to the usage scenarios, with the characteristics of low cost, customizability, simple deployment and operation and maintenance, which is conducive to the 5G technology in vertical industries. This paper proposes a solution for lightweight 5GC solution on cloud native technology and verifies its feasibility.
{"title":"Research on Lightweight 5G Core Network on Cloud Native Technology","authors":"Deheng Li, Qingyun Meng, Yi Liu, Wanyi Zhu","doi":"10.1109/CCAI57533.2023.10201328","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201328","url":null,"abstract":"The emergence of the fifth-generation (5G) technology will accelerate the digitization of the economy and society, but the public network-based 5G core network (5GC) cannot meet the usage needs of vertical industries, while the cloud-native lightweight 5GC can be customized according to the usage scenarios, with the characteristics of low cost, customizability, simple deployment and operation and maintenance, which is conducive to the 5G technology in vertical industries. This paper proposes a solution for lightweight 5GC solution on cloud native technology and verifies its feasibility.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114425955","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Modern exascale supercomputers require more efficient I/O service than traditional single-shared filesystems can provide to support applications with varying I/O loads. Although current supercomputers can offer multiple storage resources for meeting different job I/O requirements, mainstream job schedulers need the ability to allocate hardware based on job I/O characteristics automatically. Job schedulers must first predict the I/O characteristics of the high-performance computing job to enable this ability. However, the traditional I/O feature prediction method uses I/O performance metrics collected after the job starts. The I/O channels are generally built for the job at the beginning, meaning the job schedulers must predict I/O characteristics before the job starts. This paper proposes an I/O characteristics prediction method for supercomputer jobs using only job description information, which can be collected before the jobs start and contain text and numerical data. We solved the problem of integrating different data types and selected the most suitable model through model selectors during the data training process. The method achieves 80.2% accuracy and 88.6% precision through validation using more than 40 days of job records on the new generation Tianhe supercomputer.
{"title":"IOScout: an I/O Characteristics Prediction Method for the Supercomputer Jobs","authors":"Yuqi Li, Li-Quan Xiao, Jinghua Feng, Jian Zhang, Gang Zheng, Yuan Yuan","doi":"10.1109/CCAI57533.2023.10201270","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201270","url":null,"abstract":"Modern exascale supercomputers require more efficient I/O service than traditional single-shared filesystems can provide to support applications with varying I/O loads. Although current supercomputers can offer multiple storage resources for meeting different job I/O requirements, mainstream job schedulers need the ability to allocate hardware based on job I/O characteristics automatically. Job schedulers must first predict the I/O characteristics of the high-performance computing job to enable this ability. However, the traditional I/O feature prediction method uses I/O performance metrics collected after the job starts. The I/O channels are generally built for the job at the beginning, meaning the job schedulers must predict I/O characteristics before the job starts. This paper proposes an I/O characteristics prediction method for supercomputer jobs using only job description information, which can be collected before the jobs start and contain text and numerical data. We solved the problem of integrating different data types and selected the most suitable model through model selectors during the data training process. The method achieves 80.2% accuracy and 88.6% precision through validation using more than 40 days of job records on the new generation Tianhe supercomputer.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125359085","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-26DOI: 10.1109/CCAI57533.2023.10201244
Md. Abrar Hamim, Jeba Tahseen, Kazi Md. Istiyak Hossain, N. Akter, Umme Fatema Tuj Asha
Finding rotten fruits and vegetables has been important, especially in the agricultural industry. Computer vision has significant applications in the automation of damaged, freshness detection of fruits and vegetables. In recent decades, the farming sector has discovered computer machine vision and image processing technology to be more and more beneficial, particularly for implementations in quality control by identifying rotten and freshness. Farmers cannot contribute effectively between fresh and rotten fruits, vegetables because this is mainly done by people. People tire out after performing the same task for several days, whereas robots do not. By identifying weaknesses in agricultural product, the study suggested a technique for minimizing human effort and worktime. Vegetables and fruits with defects might affect healthy fruits if they are not identified in time. As an outcome, we put up a methodology to stop rottenness from spreading. The suggested model detects between fresh and decaying fruits and vegetables depending on the input fruit and vegetable photos. In this work, we used six different types of fruits and vegetables like carrot, potato, calabash, cucumber, eggplant, and cauliflower, as well as fruits likes mango, banana, star fruit, jackfruit, guava, and papaya. This study discusses multiple image processing methods for rottenness categorization of fruits and vegetables. A Convolutional Neural Network (CNN), KNN, and SVM are used to gather the features from the data fruit and vegetable photos. On Google and Kaggle datasets, the efficiency of the suggested model is evaluated, and CNN model shows the greatest accuracy which is 95 percent.
{"title":"Bangladeshi Fresh-Rotten Fruit & Vegetable Detection Using Deep Learning Deployment in Effective Application","authors":"Md. Abrar Hamim, Jeba Tahseen, Kazi Md. Istiyak Hossain, N. Akter, Umme Fatema Tuj Asha","doi":"10.1109/CCAI57533.2023.10201244","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201244","url":null,"abstract":"Finding rotten fruits and vegetables has been important, especially in the agricultural industry. Computer vision has significant applications in the automation of damaged, freshness detection of fruits and vegetables. In recent decades, the farming sector has discovered computer machine vision and image processing technology to be more and more beneficial, particularly for implementations in quality control by identifying rotten and freshness. Farmers cannot contribute effectively between fresh and rotten fruits, vegetables because this is mainly done by people. People tire out after performing the same task for several days, whereas robots do not. By identifying weaknesses in agricultural product, the study suggested a technique for minimizing human effort and worktime. Vegetables and fruits with defects might affect healthy fruits if they are not identified in time. As an outcome, we put up a methodology to stop rottenness from spreading. The suggested model detects between fresh and decaying fruits and vegetables depending on the input fruit and vegetable photos. In this work, we used six different types of fruits and vegetables like carrot, potato, calabash, cucumber, eggplant, and cauliflower, as well as fruits likes mango, banana, star fruit, jackfruit, guava, and papaya. This study discusses multiple image processing methods for rottenness categorization of fruits and vegetables. A Convolutional Neural Network (CNN), KNN, and SVM are used to gather the features from the data fruit and vegetable photos. On Google and Kaggle datasets, the efficiency of the suggested model is evaluated, and CNN model shows the greatest accuracy which is 95 percent.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124508003","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-26DOI: 10.1109/CCAI57533.2023.10201280
Wenxiao Xu, Zhijie Chen, Jie Jin, Jiangjie Huang, Yuhong Sheng
At present, online teaching resources are mainly produced by some software like Rain classroom, or recording video in professional rooms. Both methods have their own advantages and disadvantages, and put into school level, there are still a large part of schools can not provide students with sufficient and excellent online teaching resources. In order to realize an easier way to digitize teaching resources, this paper uses Raspberry Pi as the core processor, camera module and Python-OpenCV library as the basis, and combines Yolov4-tiny algorithm to realize the digitization of teaching resources, which provides a more economical and convenient way to utilize teaching resources. The digital method could promote the circulation of teaching resources and raise the frequency of information exchange inside and outside the school.
{"title":"Research on the Digital Method of Teaching Resources Based on Raspberry Pi","authors":"Wenxiao Xu, Zhijie Chen, Jie Jin, Jiangjie Huang, Yuhong Sheng","doi":"10.1109/CCAI57533.2023.10201280","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201280","url":null,"abstract":"At present, online teaching resources are mainly produced by some software like Rain classroom, or recording video in professional rooms. Both methods have their own advantages and disadvantages, and put into school level, there are still a large part of schools can not provide students with sufficient and excellent online teaching resources. In order to realize an easier way to digitize teaching resources, this paper uses Raspberry Pi as the core processor, camera module and Python-OpenCV library as the basis, and combines Yolov4-tiny algorithm to realize the digitization of teaching resources, which provides a more economical and convenient way to utilize teaching resources. The digital method could promote the circulation of teaching resources and raise the frequency of information exchange inside and outside the school.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125430065","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Aiming at the problems of complex background, high difficulty of small target detection and high miss detection rate in target detection of satellite remote sensing images, this paper proposes a multiscale target detection model based on YOLOv5 network with attention mechanism. —The feature extraction capability of the backbone network is enhanced by fusing efficient channel attention modules in the backbone network, and the detection head is decoupled and parallel convolution is used to perform classification and regression tasks separately to alleviate the conflict between classification and regression tasks. After experimental validation, the algorithm achieves 74.2% mAP and 64 FPS detection speed on Dior remote sensing dataset. experimental results show that the improved detection algorithm can effectively improve the detection capability of YOLOv5 for small and medium targets in remote sensing images and meet the real-time performance of detection.
{"title":"Remote Sensing Image Object Detection Based on Improved YOLOv5","authors":"Shenglan Zhou, Rongrong Guo, Jianhua Zhang, Weilong Chen, Yujia Peng, Yushen Tong, Yuebao Dai","doi":"10.1109/CCAI57533.2023.10201315","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201315","url":null,"abstract":"Aiming at the problems of complex background, high difficulty of small target detection and high miss detection rate in target detection of satellite remote sensing images, this paper proposes a multiscale target detection model based on YOLOv5 network with attention mechanism. —The feature extraction capability of the backbone network is enhanced by fusing efficient channel attention modules in the backbone network, and the detection head is decoupled and parallel convolution is used to perform classification and regression tasks separately to alleviate the conflict between classification and regression tasks. After experimental validation, the algorithm achieves 74.2% mAP and 64 FPS detection speed on Dior remote sensing dataset. experimental results show that the improved detection algorithm can effectively improve the detection capability of YOLOv5 for small and medium targets in remote sensing images and meet the real-time performance of detection.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129206911","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-26DOI: 10.1109/CCAI57533.2023.10201272
W. Luo, Jinyu Xue
There exist some issues such as occlusions, variable human body poses, complex backgrounds in the human pose images, so there are still challenges in the task of human body pose estimation. By adding a new attention mechanism module and reweighting the last feature maps by the original HRNet, We propose an improved HRNet model. The ability of the model is enhanced to learn spatial and semantic information. The experiments on the COCO dataset and MPII dataset show that our model could detect some key points that are missed or detected incorrectly by the original network, and the accuracy is also increased.
{"title":"Human Pose Estimation Based on Improved HRNet Model","authors":"W. Luo, Jinyu Xue","doi":"10.1109/CCAI57533.2023.10201272","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201272","url":null,"abstract":"There exist some issues such as occlusions, variable human body poses, complex backgrounds in the human pose images, so there are still challenges in the task of human body pose estimation. By adding a new attention mechanism module and reweighting the last feature maps by the original HRNet, We propose an improved HRNet model. The ability of the model is enhanced to learn spatial and semantic information. The experiments on the COCO dataset and MPII dataset show that our model could detect some key points that are missed or detected incorrectly by the original network, and the accuracy is also increased.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"404 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122858561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-26DOI: 10.1109/CCAI57533.2023.10201290
Junliang Wang, Baohong Lin
The increasing demand for high-performance storage and machine learning services in data center networks has led to the adoption of RDMA (Remote Direct Memory Access) as a replacement for the traditional TCP protocol stack. To ensure the reliability of RDMA in real-world deployments, it is crucial to perform a comprehensive reliability evaluation before deploying it in a production environment. However, current reliability evaluations of RDMA in data center networks are often limited to small-scale experiments and models, making it difficult to validate the reliability of RDMA in large-scale deployments. To address this issue, we propose a reliability evaluation model for RDMA in large-scale data center networks. The model calculates the reliability of RDMA transmission flows in complex large-scale topologies. Our experiments demonstrate that the model accurately predicts the reliability of RDMA, providing quick and convergent evaluation results on a large scale.
{"title":"RDMA Reliability Evaluation Model for Large-Scale Data Center Networks","authors":"Junliang Wang, Baohong Lin","doi":"10.1109/CCAI57533.2023.10201290","DOIUrl":"https://doi.org/10.1109/CCAI57533.2023.10201290","url":null,"abstract":"The increasing demand for high-performance storage and machine learning services in data center networks has led to the adoption of RDMA (Remote Direct Memory Access) as a replacement for the traditional TCP protocol stack. To ensure the reliability of RDMA in real-world deployments, it is crucial to perform a comprehensive reliability evaluation before deploying it in a production environment. However, current reliability evaluations of RDMA in data center networks are often limited to small-scale experiments and models, making it difficult to validate the reliability of RDMA in large-scale deployments. To address this issue, we propose a reliability evaluation model for RDMA in large-scale data center networks. The model calculates the reliability of RDMA transmission flows in complex large-scale topologies. Our experiments demonstrate that the model accurately predicts the reliability of RDMA, providing quick and convergent evaluation results on a large scale.","PeriodicalId":285760,"journal":{"name":"2023 IEEE 3rd International Conference on Computer Communication and Artificial Intelligence (CCAI)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128033156","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：