Pub Date : 2022-12-01DOI: 10.1109/CSE57773.2022.00017
Xiao Ma, Qiumiao Deng, Yi Ye, Tingting Yang, Jiangfeng Zeng
In real academic networks, there exist multiple types of entities(authors, papers, terms, conferences) and links between them. Therefore, the academic networks are generally considered as heterogeneous information networks(HINs). Existing collabo-rator recommendation methods in heterogeneous networks are generally based on the embeddings of nodes and links with re-spect to some given meta-paths. However, they seldom learn meta-paths representations which can provide important interaction information. What's more, the impact of different meta-paths on recommendation are neglected. In order to deal with these unsolved problems, we propose an attention based collaborator recommendation method in the setting of heterogeneous academic networks. Firstly, we select some meta-paths according to the HIN schema. Secondly, the embeddings of nodes and meta-path instances are generated by employing the Skip-gram and Convolutional Neural Network(CNN) models respectively. Thirdly, the attention mechanism is devised to integrate the multiple sources of embeddings so as to produce the author representations and meta-path based context representations. Finally, the Multi-Layer Perceptron is utilized for recommendation task. Comparative experiments conducted on the DBLP dataset demonstrate the effectiveness of our proposed method.
{"title":"Attention based Collaborator Recommendation in Heterogeneous Academic Networks","authors":"Xiao Ma, Qiumiao Deng, Yi Ye, Tingting Yang, Jiangfeng Zeng","doi":"10.1109/CSE57773.2022.00017","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00017","url":null,"abstract":"In real academic networks, there exist multiple types of entities(authors, papers, terms, conferences) and links between them. Therefore, the academic networks are generally considered as heterogeneous information networks(HINs). Existing collabo-rator recommendation methods in heterogeneous networks are generally based on the embeddings of nodes and links with re-spect to some given meta-paths. However, they seldom learn meta-paths representations which can provide important interaction information. What's more, the impact of different meta-paths on recommendation are neglected. In order to deal with these unsolved problems, we propose an attention based collaborator recommendation method in the setting of heterogeneous academic networks. Firstly, we select some meta-paths according to the HIN schema. Secondly, the embeddings of nodes and meta-path instances are generated by employing the Skip-gram and Convolutional Neural Network(CNN) models respectively. Thirdly, the attention mechanism is devised to integrate the multiple sources of embeddings so as to produce the author representations and meta-path based context representations. Finally, the Multi-Layer Perceptron is utilized for recommendation task. Comparative experiments conducted on the DBLP dataset demonstrate the effectiveness of our proposed method.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"67 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133229153","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 : 2022-12-01DOI: 10.1109/CSE57773.2022.00021
Abdulmonem Alshahrani, John A. Clark
Configuring intrusion detection systems (IDSs) in large networks may involve balancing multiple criteria, e.g. detection rate, number of probes, and power consumption at each node. The tradeoffs become particularly acute when the nodes are resource-constrained, as is often the case in the Internet of Things (IoT) networks. A genetic algorithm based optimisation approach is outlined to address this task. However, the fitness function is evaluated in part via a computationally expensive simulation. We show how a neural network, trained over a set of IDS configurations, can be used as a surrogate fitness function, providing better results more cheaply.
{"title":"Neural Network Approximation of Simulation-based IDS Fitness Evaluation","authors":"Abdulmonem Alshahrani, John A. Clark","doi":"10.1109/CSE57773.2022.00021","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00021","url":null,"abstract":"Configuring intrusion detection systems (IDSs) in large networks may involve balancing multiple criteria, e.g. detection rate, number of probes, and power consumption at each node. The tradeoffs become particularly acute when the nodes are resource-constrained, as is often the case in the Internet of Things (IoT) networks. A genetic algorithm based optimisation approach is outlined to address this task. However, the fitness function is evaluated in part via a computationally expensive simulation. We show how a neural network, trained over a set of IDS configurations, can be used as a surrogate fitness function, providing better results more cheaply.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114826679","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 : 2022-12-01DOI: 10.1109/CSE57773.2022.00016
Hongyan Zhao, Dong Yu, Biao Wang
In transonic wind tunnel, anomalous data that are often referred to as outliers or anomalies have severe impact on system identification. To address such a problem, outliers should be detected and new substitutions should be provided before system identification. The combined request for outlier detection and compensation makes it suitable to develop a regression-based outlier mining algorithm. To enhance the effectiveness of traditional regression-based algorithm, this paper proposes a novel one based on ensemble learning. In our outlier ensemble, the base regression models are learnt on a two-level ensemble structure. The aim of the first level is to enhance the robustness to unknown outliers by homogeneous ensemble. The goal of the second level is to improve the robustness to base regression model. In order to verify the effectiveness of the proposed hybrid outlier ensemble, we use several real-world datasets from transonic wind tunnel and compare it with several underlying competitors. The experimental results have shown that the proposed outlier ensembles could outperform its competitors with respect to both outlier mining and the improvement of system identification.
{"title":"Improving the System Identification of Transonic Wind Tunnel by a Regression Ensemble-Based Outlier Mining Method","authors":"Hongyan Zhao, Dong Yu, Biao Wang","doi":"10.1109/CSE57773.2022.00016","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00016","url":null,"abstract":"In transonic wind tunnel, anomalous data that are often referred to as outliers or anomalies have severe impact on system identification. To address such a problem, outliers should be detected and new substitutions should be provided before system identification. The combined request for outlier detection and compensation makes it suitable to develop a regression-based outlier mining algorithm. To enhance the effectiveness of traditional regression-based algorithm, this paper proposes a novel one based on ensemble learning. In our outlier ensemble, the base regression models are learnt on a two-level ensemble structure. The aim of the first level is to enhance the robustness to unknown outliers by homogeneous ensemble. The goal of the second level is to improve the robustness to base regression model. In order to verify the effectiveness of the proposed hybrid outlier ensemble, we use several real-world datasets from transonic wind tunnel and compare it with several underlying competitors. The experimental results have shown that the proposed outlier ensembles could outperform its competitors with respect to both outlier mining and the improvement of system identification.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117068634","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}
A particular type of dynamic LED visual marker was designed in this study to address the shortcomings of the existing visual marker of the multi-robot positioning system that uses an external camera. Moreover, this dynamic LED visual marker was proposed using the tracking and positioning algorithm. This marker can distinguish and detect the positions of all the robots with LED visual markers in the image. Dynamic LED visual markers use colourful LEDs as carriers, which are arranged in the order of red, green and blue colours to communicate information. Moreover, a coding rule based on ternary trees was also developed. The tracking and positioning algorithm applied the dual-thread design of the tracking and detection threads. The former completes coding verification using the Kalman filtering algorithm while tracking the LED markers in images. The latter positions LED and reads encoding information by detecting the initial signal. Such dual-thread design effectively decreases the computation workload and emphasises on accurate positioning and fast response. The experimental results suggest that the proposed visual marker has a smaller volume and a more extended sphere of influence than the existing ARTag visual marker method. The tracking and positioning algorithm completes the visual positioning task of a multi-robot system with high accuracy and robustness.
{"title":"LED Dynamic Marker and Tracking Algorithm for External Camera Positioning","authors":"Jianxu Mao, Zhiqiang Zou, Caiping Liu, Junfei Yi, Ziming Tao, Yaonan Wang","doi":"10.1109/CSE57773.2022.00012","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00012","url":null,"abstract":"A particular type of dynamic LED visual marker was designed in this study to address the shortcomings of the existing visual marker of the multi-robot positioning system that uses an external camera. Moreover, this dynamic LED visual marker was proposed using the tracking and positioning algorithm. This marker can distinguish and detect the positions of all the robots with LED visual markers in the image. Dynamic LED visual markers use colourful LEDs as carriers, which are arranged in the order of red, green and blue colours to communicate information. Moreover, a coding rule based on ternary trees was also developed. The tracking and positioning algorithm applied the dual-thread design of the tracking and detection threads. The former completes coding verification using the Kalman filtering algorithm while tracking the LED markers in images. The latter positions LED and reads encoding information by detecting the initial signal. Such dual-thread design effectively decreases the computation workload and emphasises on accurate positioning and fast response. The experimental results suggest that the proposed visual marker has a smaller volume and a more extended sphere of influence than the existing ARTag visual marker method. The tracking and positioning algorithm completes the visual positioning task of a multi-robot system with high accuracy and robustness.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115265417","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 : 2022-12-01DOI: 10.1109/CSE57773.2022.00010
Guanglie Ouyang, Tinghao Qi, Lixiao Wei, Bang Wang
Fingerprint-based indoor localization methods usually use received signal strength (RSS) and channel status information (CSI) as the localization fingerprint, which suffers from time-consuming and labor-intensive site survey. In this paper, we propose an indoor localization method based on sparse time difference of arrival (TDOA) fingerprints. This method constructs the localization fingerprints by TDOA, which is calibrated by the straight line fitting method and the beacon estimation method. In order to get the dense fingerprint database, we propose a TDOA interpolation method based on distance relation. Experiments on field measurements validate the effectiveness of the proposed method. In the case of only sampling three reference points (RPs), the average localization error (ALE) of the proposed method reaches 0.824 m, which obtains a 48.8 % improvement compared with the traditional TDOA algorithm,
{"title":"Indoor Localization Based on Sparse TDOA Fingerprints","authors":"Guanglie Ouyang, Tinghao Qi, Lixiao Wei, Bang Wang","doi":"10.1109/CSE57773.2022.00010","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00010","url":null,"abstract":"Fingerprint-based indoor localization methods usually use received signal strength (RSS) and channel status information (CSI) as the localization fingerprint, which suffers from time-consuming and labor-intensive site survey. In this paper, we propose an indoor localization method based on sparse time difference of arrival (TDOA) fingerprints. This method constructs the localization fingerprints by TDOA, which is calibrated by the straight line fitting method and the beacon estimation method. In order to get the dense fingerprint database, we propose a TDOA interpolation method based on distance relation. Experiments on field measurements validate the effectiveness of the proposed method. In the case of only sampling three reference points (RPs), the average localization error (ALE) of the proposed method reaches 0.824 m, which obtains a 48.8 % improvement compared with the traditional TDOA algorithm,","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132304975","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 : 2022-12-01DOI: 10.1109/CSE57773.2022.00018
Jing Zhao, Will Zhao, Yimin Yang, A. Safaei, Ruizhong Wei
In the past few decades, with the explosion of information, a large number of computer scientists have devoted themselves to analyzing collected data and applying these findings to many disciplines. Natural language processing (NLP) has been one of the most popular areas for data analysis and pattern recognition. A significantly large amount of data is obtained in text format due to the ease of access nowadays. Most modern techniques focus on exploring large sets of textual data to build forecasting models; they tend to ignore the importance of temporal information which is often the main ingredient to determine the performance of analysis, especially in the public policy view. The contribution of this paper is two-fold. First, a dataset called COVID-News is collected from three news agencies, which consists of article segments related to wearing masks during the COVID-19 pandemic. Second, we propose a long-short term memory (LSTM)-based learning model to predict the attitude of the articles from the three news agencies towards wearing a mask with both temporal and textural information. Experimental results on COVID-News dataset show the effectiveness of the proposed LSTM-based algorithm.
{"title":"To Mask or Not To Mask? A Machine Learning Approach to Covid News Coverage Attitude Prediction Based on Time Series and Text Content","authors":"Jing Zhao, Will Zhao, Yimin Yang, A. Safaei, Ruizhong Wei","doi":"10.1109/CSE57773.2022.00018","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00018","url":null,"abstract":"In the past few decades, with the explosion of information, a large number of computer scientists have devoted themselves to analyzing collected data and applying these findings to many disciplines. Natural language processing (NLP) has been one of the most popular areas for data analysis and pattern recognition. A significantly large amount of data is obtained in text format due to the ease of access nowadays. Most modern techniques focus on exploring large sets of textual data to build forecasting models; they tend to ignore the importance of temporal information which is often the main ingredient to determine the performance of analysis, especially in the public policy view. The contribution of this paper is two-fold. First, a dataset called COVID-News is collected from three news agencies, which consists of article segments related to wearing masks during the COVID-19 pandemic. Second, we propose a long-short term memory (LSTM)-based learning model to predict the attitude of the articles from the three news agencies towards wearing a mask with both temporal and textural information. Experimental results on COVID-News dataset show the effectiveness of the proposed LSTM-based algorithm.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132174807","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 : 2022-12-01DOI: 10.1109/CSE57773.2022.00023
Weiwei Yu, Jacques Bangamwabo, Zidi Wang, XiaoXu Yang, Min Jiang, Yanen Wang
E-learning offers the flexibility of learning time and location for many students than traditional education. However, not all learning courses can be easily learned through online platforms and effectively meet the students' needs. As a result, lead to a loss of learning motivation and concentration on the student's side. While most recent studies have considered facial emotions as an effective tool for interpreting learning experiences in learners, but they have ignored the characteristic of courses. And learner engagement and performance on specific knowledge units is an effective method to analyze the student learning process. Therefore, this study proposed a hybrid approach for analyzing student engagement in video-based online courses, including a knowledge map to represent knowledge units and their relationships, and a convolutional neural network to examine the learners' facial expressions to interpret their learning effect on each concept during e-learning. The classification process has been adapted to identify different grades of learning emotions at knowledge units. The knowledge map partition method has been proposed, and by analyzing and visualizing the different divisions, the teacher can better understand the student's understanding levels based on his emotion within the course. The study demonstrated how personalized reports of knowledge unit understanding from this model could serve as a basis for future course content modifications and the organization and optimizing teaching materials.
{"title":"Analysis of student e-learning engagement using learning affect: Hybrid of facial emotions and domain model","authors":"Weiwei Yu, Jacques Bangamwabo, Zidi Wang, XiaoXu Yang, Min Jiang, Yanen Wang","doi":"10.1109/CSE57773.2022.00023","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00023","url":null,"abstract":"E-learning offers the flexibility of learning time and location for many students than traditional education. However, not all learning courses can be easily learned through online platforms and effectively meet the students' needs. As a result, lead to a loss of learning motivation and concentration on the student's side. While most recent studies have considered facial emotions as an effective tool for interpreting learning experiences in learners, but they have ignored the characteristic of courses. And learner engagement and performance on specific knowledge units is an effective method to analyze the student learning process. Therefore, this study proposed a hybrid approach for analyzing student engagement in video-based online courses, including a knowledge map to represent knowledge units and their relationships, and a convolutional neural network to examine the learners' facial expressions to interpret their learning effect on each concept during e-learning. The classification process has been adapted to identify different grades of learning emotions at knowledge units. The knowledge map partition method has been proposed, and by analyzing and visualizing the different divisions, the teacher can better understand the student's understanding levels based on his emotion within the course. The study demonstrated how personalized reports of knowledge unit understanding from this model could serve as a basis for future course content modifications and the organization and optimizing teaching materials.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115100666","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 : 2022-12-01DOI: 10.1109/CSE57773.2022.00011
Bruce Belson, B. Philippa
We demonstrate a novel method of speeding up large iterative tasks such as machine learning inference. Our approach is to improve the memory access pattern, taking advantage of coroutines as a programming language feature to minimise the developer effort and reduce code complexity. We evaluate our approach using a comprehensive set of bench-marks run on three hardware platforms (one ARM and two Intel CPUs). The best observed performance boosts were 65% for scanning the nodes in a B+ tree, 34% for support vector machine inference, 12% for image pixel normalisation, and 15.5% for two dimensional convolution. Performance varied with data size, numeric type, and other factors, but overall the method is practical and can lead to significant improvements for edge computing.
{"title":"Speeding up Machine Learning Inference on Edge Devices by Improving Memory Access Patterns using Coroutines","authors":"Bruce Belson, B. Philippa","doi":"10.1109/CSE57773.2022.00011","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00011","url":null,"abstract":"We demonstrate a novel method of speeding up large iterative tasks such as machine learning inference. Our approach is to improve the memory access pattern, taking advantage of coroutines as a programming language feature to minimise the developer effort and reduce code complexity. We evaluate our approach using a comprehensive set of bench-marks run on three hardware platforms (one ARM and two Intel CPUs). The best observed performance boosts were 65% for scanning the nodes in a B+ tree, 34% for support vector machine inference, 12% for image pixel normalisation, and 15.5% for two dimensional convolution. Performance varied with data size, numeric type, and other factors, but overall the method is practical and can lead to significant improvements for edge computing.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"135 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123753393","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 : 2022-12-01DOI: 10.1109/CSE57773.2022.00022
Rong Gu, Bo Li, Dingjin Liu, Zhaokang Wang, Suhui Wangzhang, Shulin Wang, Haipeng Dai, Yihua Huang
The reverse-time migration (RTM) imaging algorithm is widely used in petroleum seismic exploration analysis. It is one of the most accurate imaging algorithms but is also computation-intensive and thus time-consuming. In this paper, we focus on improving the parallel execution performance of the reverse-time migration imaging algorithm. Firstly, we analyze the performance bottlenecks of the reverse-time migration imaging algorithm with program profiling techniques. Based on the program profiling and performance analysis, we propose three effective performance improvement strategies, including the pipeline-based iterative propagation computation, the fine-grained data compression, and the GPU memory specification-based data transmission, to eliminate the performance bottle-necks. Extensive experiments on physical clusters and real-world datasets show that the proposed pipeline-based and fine-grained parallel RTM algorithm can reduce the running time by an average of 58.42% compared with the existing solutions. In addition, the proposed algorithm has been used for over one year in the real-world production environment in Sinopec, which is one of the world's largest petroleum exploration companies.
{"title":"Towards Efficient Reverse-time Migration Imaging Computation by Pipeline and Fine-grained Execution Parallelization","authors":"Rong Gu, Bo Li, Dingjin Liu, Zhaokang Wang, Suhui Wangzhang, Shulin Wang, Haipeng Dai, Yihua Huang","doi":"10.1109/CSE57773.2022.00022","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00022","url":null,"abstract":"The reverse-time migration (RTM) imaging algorithm is widely used in petroleum seismic exploration analysis. It is one of the most accurate imaging algorithms but is also computation-intensive and thus time-consuming. In this paper, we focus on improving the parallel execution performance of the reverse-time migration imaging algorithm. Firstly, we analyze the performance bottlenecks of the reverse-time migration imaging algorithm with program profiling techniques. Based on the program profiling and performance analysis, we propose three effective performance improvement strategies, including the pipeline-based iterative propagation computation, the fine-grained data compression, and the GPU memory specification-based data transmission, to eliminate the performance bottle-necks. Extensive experiments on physical clusters and real-world datasets show that the proposed pipeline-based and fine-grained parallel RTM algorithm can reduce the running time by an average of 58.42% compared with the existing solutions. In addition, the proposed algorithm has been used for over one year in the real-world production environment in Sinopec, which is one of the world's largest petroleum exploration companies.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"141 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124478904","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 : 2022-12-01DOI: 10.1109/CSE57773.2022.00014
Wei Xu, Ruoxuan Zhou, Qiuming Liu
Electroencephalogram signals (EEG) has been widely used in emotion recognition because of its authenticity and unforgeability. Therefore, EEG emotion recognition has become one of the main technologies of emotion computing. EEG signals are composed of complex time domain, frequency domain and spatial domain (TFS) related information. Aiming at the problems of insufficient mining of TFS feature information and low recognition rate in EEG emotion recognition. This paper presents a Multi-Task Joint Neural Network (MT-2DCNN-LSTM) model constructed by two-dimensional convolutional neural network (2DCNN) and long short-term memory neural network (LSTM). In this paper, frequency domain and spatial domain features are used to construct 3D feature matrix graph, and time domain features are used to construct 2D sequence information. Then these two features are used as input of the model to fully extract the TFS feature information of EEG signals. In order to verify the recognition ability of the model for EEG signals, a multivariate classification experiment was carried out on the DEAP dataset, a well-known dataset for comparison purposes. Among them, the average accuracy of emotion recognition of arousal and valence is 97.29% and 97.72%, respectively. The results show that MT-2DCNN-LSTM has excellent performance.
{"title":"Electroencephalogram Emotion Recognition Based on Three-Dimensional Feature Matrix and Multivariate Neural Network","authors":"Wei Xu, Ruoxuan Zhou, Qiuming Liu","doi":"10.1109/CSE57773.2022.00014","DOIUrl":"https://doi.org/10.1109/CSE57773.2022.00014","url":null,"abstract":"Electroencephalogram signals (EEG) has been widely used in emotion recognition because of its authenticity and unforgeability. Therefore, EEG emotion recognition has become one of the main technologies of emotion computing. EEG signals are composed of complex time domain, frequency domain and spatial domain (TFS) related information. Aiming at the problems of insufficient mining of TFS feature information and low recognition rate in EEG emotion recognition. This paper presents a Multi-Task Joint Neural Network (MT-2DCNN-LSTM) model constructed by two-dimensional convolutional neural network (2DCNN) and long short-term memory neural network (LSTM). In this paper, frequency domain and spatial domain features are used to construct 3D feature matrix graph, and time domain features are used to construct 2D sequence information. Then these two features are used as input of the model to fully extract the TFS feature information of EEG signals. In order to verify the recognition ability of the model for EEG signals, a multivariate classification experiment was carried out on the DEAP dataset, a well-known dataset for comparison purposes. Among them, the average accuracy of emotion recognition of arousal and valence is 97.29% and 97.72%, respectively. The results show that MT-2DCNN-LSTM has excellent performance.","PeriodicalId":165085,"journal":{"name":"2022 IEEE 25th International Conference on Computational Science and Engineering (CSE)","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132889272","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：