Pub Date : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674161
Li Mujin
As an idea for handling uncertainty, Dempster-Shafer theory has attracted much of people's attention. An open issue is that Dempster's combination rules will obtain counter-intuitive results when used directly to handle high conflict information. In this paper, an exponential function is defined to modify the data model for eliminating the effect of conflict. Moreover, this method makes up for the deficiency of the classical combination rule. For a multiple-sensor data fusion system, it can avoid highly conflicting situations and have a better judge of surroundings. At the same time, some numerical instances and experiments on the Iris dataset are given to demonstrate the efficiency of the proposed method.
{"title":"Conflict Management in Evidence Theory: An Exponential Mode","authors":"Li Mujin","doi":"10.1109/ICCWAMTIP53232.2021.9674161","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674161","url":null,"abstract":"As an idea for handling uncertainty, Dempster-Shafer theory has attracted much of people's attention. An open issue is that Dempster's combination rules will obtain counter-intuitive results when used directly to handle high conflict information. In this paper, an exponential function is defined to modify the data model for eliminating the effect of conflict. Moreover, this method makes up for the deficiency of the classical combination rule. For a multiple-sensor data fusion system, it can avoid highly conflicting situations and have a better judge of surroundings. At the same time, some numerical instances and experiments on the Iris dataset are given to demonstrate the efficiency of the proposed method.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133969597","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674167
Xi He, Jianping Li, Tiankai Li, He Liu
With the development of Internet technology, more and more Chinese platforms are creating massive Chinese texts. At present, obtaining data samples for training is no longer a problem, and more and more researchers are beginning to devote themselves to obtaining great value from mining text information. Chinese text classification is mainly used for user sentiment analysis, personalized recommendation, topic tracking, and public opinion monitoring. However, Chinese texts naturally have many difficulties, such as many ambiguities, difficult word segmentation, fewer words, and sparse features. Traditional machine learning has a poor realization effect on Chinese texts. Deep neural networks are gradually becoming a new trend in Chinese text classification.
{"title":"Chinese Short Text Classification Based On Deep Learning","authors":"Xi He, Jianping Li, Tiankai Li, He Liu","doi":"10.1109/ICCWAMTIP53232.2021.9674167","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674167","url":null,"abstract":"With the development of Internet technology, more and more Chinese platforms are creating massive Chinese texts. At present, obtaining data samples for training is no longer a problem, and more and more researchers are beginning to devote themselves to obtaining great value from mining text information. Chinese text classification is mainly used for user sentiment analysis, personalized recommendation, topic tracking, and public opinion monitoring. However, Chinese texts naturally have many difficulties, such as many ambiguities, difficult word segmentation, fewer words, and sparse features. Traditional machine learning has a poor realization effect on Chinese texts. Deep neural networks are gradually becoming a new trend in Chinese text classification.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114158622","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674085
Xie Wenqiang, Chen Huaixin, Wang Zhixi
In order to improve the detection accuracy of gypsophila in the display screen, a defect detection model based on human visual perception is proposed. The model uses human visual perception information as the key point of detection. First, the HSV color space is used to obtain the color information in the original image, and it is fused with the mean-constrained RGB gray-scale image to make the grayscale image contain local color information; Taking the grayscale image as the optimization benchmark, adaptively obtain the single-channel image constraint coefficients containing global color information. The single-channel gray map constrained by the transform coefficient is used for defect detection, which improves the accuracy of defect detection. The experimental results show that the average defect detection accuracy and recall rate of the algorithm in this paper are more than 95%. Compared with the traditional detection method, the accuracy rate is improved by more than 50%. The detection method in this paper meets the needs of industrial production.
{"title":"Method for Detecting Gypsophila Defect of Display Screen Based on Human Visual Perception","authors":"Xie Wenqiang, Chen Huaixin, Wang Zhixi","doi":"10.1109/ICCWAMTIP53232.2021.9674085","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674085","url":null,"abstract":"In order to improve the detection accuracy of gypsophila in the display screen, a defect detection model based on human visual perception is proposed. The model uses human visual perception information as the key point of detection. First, the HSV color space is used to obtain the color information in the original image, and it is fused with the mean-constrained RGB gray-scale image to make the grayscale image contain local color information; Taking the grayscale image as the optimization benchmark, adaptively obtain the single-channel image constraint coefficients containing global color information. The single-channel gray map constrained by the transform coefficient is used for defect detection, which improves the accuracy of defect detection. The experimental results show that the average defect detection accuracy and recall rate of the algorithm in this paper are more than 95%. Compared with the traditional detection method, the accuracy rate is improved by more than 50%. The detection method in this paper meets the needs of industrial production.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"118 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116376134","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674055
Jing Xie, Xiang Yin, Xiyi Zhang, Juan Chen, Q. Wen
In the conventional federated learning, the local models of multiple clients are trained independently by their privacy data, and the center server generates the shared global model by aggregating local models. However, the global model often fails to adapt to each client due to statistical heterogeneities, such as non-IID data. To address the problem, we propose the Subclass Personalized Federated Learning (SPFL) algorithm for non-IID data. In SPFL, the server uses the Softmax Normalized Gradient Similarity (SNGS) to weight the relationship between clients, and sends the personalized global model to each client. The stage strategy of ResNet is also applied to improve the performance of our algorithm. The experimental results show that the SPFL algorithm used on non-IID data outperforms the vanilla FedAvg, Per-FedAvg, FedUpdate, and pFedMe algorithms, improving the accuracy by 1.81∼18.46% on four datasets (CIFAR10, CIFAR100, MNIST, EMNIST), while still maintaining the state-of-the-art performance on IID data.
{"title":"Personalized Federated Learning with Gradient Similarity","authors":"Jing Xie, Xiang Yin, Xiyi Zhang, Juan Chen, Q. Wen","doi":"10.1109/ICCWAMTIP53232.2021.9674055","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674055","url":null,"abstract":"In the conventional federated learning, the local models of multiple clients are trained independently by their privacy data, and the center server generates the shared global model by aggregating local models. However, the global model often fails to adapt to each client due to statistical heterogeneities, such as non-IID data. To address the problem, we propose the Subclass Personalized Federated Learning (SPFL) algorithm for non-IID data. In SPFL, the server uses the Softmax Normalized Gradient Similarity (SNGS) to weight the relationship between clients, and sends the personalized global model to each client. The stage strategy of ResNet is also applied to improve the performance of our algorithm. The experimental results show that the SPFL algorithm used on non-IID data outperforms the vanilla FedAvg, Per-FedAvg, FedUpdate, and pFedMe algorithms, improving the accuracy by 1.81∼18.46% on four datasets (CIFAR10, CIFAR100, MNIST, EMNIST), while still maintaining the state-of-the-art performance on IID data.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114801403","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674137
Y. Feng, Haiyin Lin
The leaf area index reflects the growth of vegetation. The hemispheric image method is a common method for measuring the leaf area index. However, fisheye lenses, containing a large number of mixed cells, would change the luminosity of images. Traditional threshold methods will increase the error in calculating the key variable of the leaf area index, the canopy porosity. So, they cannot distinguish the sky and the leaves efficiently. This paper proposes the Angular Block Otsu algorithm, an improved algorithm based on Otsu for canopy fisheye images. Compared with previous methods, it can retain or highlight the original detail information of the image better, so that the accuracy of the canopy porosity calculation is greatly improved.
{"title":"The Angular Block OTSU for Canopy Porosity of Hemisphere Method","authors":"Y. Feng, Haiyin Lin","doi":"10.1109/ICCWAMTIP53232.2021.9674137","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674137","url":null,"abstract":"The leaf area index reflects the growth of vegetation. The hemispheric image method is a common method for measuring the leaf area index. However, fisheye lenses, containing a large number of mixed cells, would change the luminosity of images. Traditional threshold methods will increase the error in calculating the key variable of the leaf area index, the canopy porosity. So, they cannot distinguish the sky and the leaves efficiently. This paper proposes the Angular Block Otsu algorithm, an improved algorithm based on Otsu for canopy fisheye images. Compared with previous methods, it can retain or highlight the original detail information of the image better, so that the accuracy of the canopy porosity calculation is greatly improved.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128633179","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674073
Bai Wei, Zhuang Yan
As online fora increasingly become the main media for argument and debate, the automatic processing of such data is rapidly becoming more and more important. Stance classification, which aims to classify the stance of the claims towards the given topic, can be applied in many application areas such as users' feelings about services and products. We propose a ensemble model for stance classification with data augment for small sample scenarios, multi-sample dropout for low training speed scenarios, focal loss for imbalance sample scenarios, pseudo labels for self-supervised training scenarios, adversarial training for low robustness scenarios, and all the above can be used in normal scenarios. Besides, the ensemble model is composed of task-specific RoBERTa and MacBERT, which can make more reasonable predictions. We used dataset from NLPCC to validate the model and it worked well.
{"title":"Claim Stance Classification Optimized by Data Augment","authors":"Bai Wei, Zhuang Yan","doi":"10.1109/ICCWAMTIP53232.2021.9674073","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674073","url":null,"abstract":"As online fora increasingly become the main media for argument and debate, the automatic processing of such data is rapidly becoming more and more important. Stance classification, which aims to classify the stance of the claims towards the given topic, can be applied in many application areas such as users' feelings about services and products. We propose a ensemble model for stance classification with data augment for small sample scenarios, multi-sample dropout for low training speed scenarios, focal loss for imbalance sample scenarios, pseudo labels for self-supervised training scenarios, adversarial training for low robustness scenarios, and all the above can be used in normal scenarios. Besides, the ensemble model is composed of task-specific RoBERTa and MacBERT, which can make more reasonable predictions. We used dataset from NLPCC to validate the model and it worked well.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129394590","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674071
Isaac Osei Agyemang, Xiaoling Zhang, Isaac Adjei-Mensah, B. L. Y. Agbley, Linda Delali Fiasam, Bernard Cobbinah Mawuli, Collins Sey
Resource-constrained edge nodes are ubiquitous in industrial settings yet are challenged by limited computing resources. Leveraging computational advantage and perceptual awareness of Gabor filters, a hybrid classifier to mitigate computational requirements in the context of classification of key components of civil structures which are essential in the proactive structural assessment, specific repairs, and maintenance post-construction is given. Deployment of the hybrid classifier using the CoreML framework exhibits favorable classification accuracy and robustness as compared to contemporary state-of-the-art classifiers.
{"title":"Accelerating Classification on Resource-Constrained Edge Nodes Towards Automated Structural Health Monitoring","authors":"Isaac Osei Agyemang, Xiaoling Zhang, Isaac Adjei-Mensah, B. L. Y. Agbley, Linda Delali Fiasam, Bernard Cobbinah Mawuli, Collins Sey","doi":"10.1109/ICCWAMTIP53232.2021.9674071","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674071","url":null,"abstract":"Resource-constrained edge nodes are ubiquitous in industrial settings yet are challenged by limited computing resources. Leveraging computational advantage and perceptual awareness of Gabor filters, a hybrid classifier to mitigate computational requirements in the context of classification of key components of civil structures which are essential in the proactive structural assessment, specific repairs, and maintenance post-construction is given. Deployment of the hybrid classifier using the CoreML framework exhibits favorable classification accuracy and robustness as compared to contemporary state-of-the-art classifiers.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134101064","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674105
Wang Jianjun
Smart education is the fundamental means to realize “student-centered”, and the smart learning environment is the technical support and condition guarantee for smart education. Explains how to build a smart learning environment based on the three-tier architecture of the Internet of Things, that is, the perception layer, network layer, and application layer, to achieve the deep integration of human, machine, and things, to create a contextual learning environment for learners to satisfy learners Personalized learning needs.
{"title":"Research on the Application Practice of the Internet of Things in the Smart Learning Environment","authors":"Wang Jianjun","doi":"10.1109/ICCWAMTIP53232.2021.9674105","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674105","url":null,"abstract":"Smart education is the fundamental means to realize “student-centered”, and the smart learning environment is the technical support and condition guarantee for smart education. Explains how to build a smart learning environment based on the three-tier architecture of the Internet of Things, that is, the perception layer, network layer, and application layer, to achieve the deep integration of human, machine, and things, to create a contextual learning environment for learners to satisfy learners Personalized learning needs.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116645750","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674139
Xiang Qian, Wang Yan-Wu
Although convolutional neural networks (CNN) have notably improved the effect of image denoising, removal of non-Gaussian noise remain a challenging problem. In this work, the statistical characteristic of image residuals is investigated and used as auxiliary information for better removing complex type noise via multi-task learning method. We propose an improved algorithm for denoising CNN (DCNN) by optimizing the training of the DCNN and it can achieve a Pareto optimal solution. Extensive experiments on benchmark data sets with different noise models demonstrate that the proposed method can effectively improve the quality of denoised images both in Gaussian and non-Gaussian noise, even when the network architecture is left unchanged.
{"title":"Image Denoising Via Multi-Task Learning","authors":"Xiang Qian, Wang Yan-Wu","doi":"10.1109/ICCWAMTIP53232.2021.9674139","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674139","url":null,"abstract":"Although convolutional neural networks (CNN) have notably improved the effect of image denoising, removal of non-Gaussian noise remain a challenging problem. In this work, the statistical characteristic of image residuals is investigated and used as auxiliary information for better removing complex type noise via multi-task learning method. We propose an improved algorithm for denoising CNN (DCNN) by optimizing the training of the DCNN and it can achieve a Pareto optimal solution. Extensive experiments on benchmark data sets with different noise models demonstrate that the proposed method can effectively improve the quality of denoised images both in Gaussian and non-Gaussian noise, even when the network architecture is left unchanged.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125458485","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 : 2021-12-17DOI: 10.1109/ICCWAMTIP53232.2021.9674102
F. A. Acheampong, H. Nunoo-Mensah, Wenyu Chen
The use of ensembles has given rise to improved performance in various machine learning tasks. Following the performance of major transformer-based language models in detecting emotions from written texts, the paper investigates the ensemble's performance of the RoBERTa and XLNet transformer-based language models in recognizing emotions from the ISEAR dataset. Finally, the results obtained outperformed the F1-scores of current works in literature with a higher F1-score of 0.75 in detecting emotions from the ISEAR text data.
{"title":"Recognizing Emotions from Texts Using an Ensemble of Transformer-Based Language Models","authors":"F. A. Acheampong, H. Nunoo-Mensah, Wenyu Chen","doi":"10.1109/ICCWAMTIP53232.2021.9674102","DOIUrl":"https://doi.org/10.1109/ICCWAMTIP53232.2021.9674102","url":null,"abstract":"The use of ensembles has given rise to improved performance in various machine learning tasks. Following the performance of major transformer-based language models in detecting emotions from written texts, the paper investigates the ensemble's performance of the RoBERTa and XLNet transformer-based language models in recognizing emotions from the ISEAR dataset. Finally, the results obtained outperformed the F1-scores of current works in literature with a higher F1-score of 0.75 in detecting emotions from the ISEAR text data.","PeriodicalId":358772,"journal":{"name":"2021 18th International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129972504","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：