Bowei Fang, Hao Liu, Wei He, Dexin Li, Chengzhao Liu
Water quality monitoring provides a basis for water quality control and water resources management. Prediction of water quality parameters can plan water use strategies, prevent further water pollution and improve water resource utilization efficiency. We propose a water quality parameter prediction method based on transformer architecture model and multi-sensor data fusion. The proposed multiple water quality parameter prediction model accepts multiple types of water quality parameter data input at the same time. The data embedding module integrates multiple types of water quality parameter information and assigns a unique position code to the data at each time step. The self-attention mechanism of the model mining the potential correlation between different time step data. The model can learn the internal relationship of the fusion data of multiple water quality parameters, and effectively predict the future trend of water quality parameters. The effectiveness of the proposed algorithm is verified by the measured data, and the advantages of the proposed method are verified by comparative experiments.
{"title":"A water quality parameter prediction method based on transformer architecture and multi-sensor data fusion","authors":"Bowei Fang, Hao Liu, Wei He, Dexin Li, Chengzhao Liu","doi":"10.1145/3590003.3590037","DOIUrl":"https://doi.org/10.1145/3590003.3590037","url":null,"abstract":"Water quality monitoring provides a basis for water quality control and water resources management. Prediction of water quality parameters can plan water use strategies, prevent further water pollution and improve water resource utilization efficiency. We propose a water quality parameter prediction method based on transformer architecture model and multi-sensor data fusion. The proposed multiple water quality parameter prediction model accepts multiple types of water quality parameter data input at the same time. The data embedding module integrates multiple types of water quality parameter information and assigns a unique position code to the data at each time step. The self-attention mechanism of the model mining the potential correlation between different time step data. The model can learn the internal relationship of the fusion data of multiple water quality parameters, and effectively predict the future trend of water quality parameters. The effectiveness of the proposed algorithm is verified by the measured data, and the advantages of the proposed method are verified by comparative experiments.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126450644","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}
Threat tree model can clearly organize threat induction information and thus is widely used for risk analysis in software assurance. Threat tree will grow to complicated structures, e.g., the number of nodes and branches, when the threat information grows to a huge volume. To extend the scalability of the threat tree model, we propose a tree model with merged nodes so as to largely decrease the number of nodes and branches. The formal model and dedicated algorithms are proposed in details. The experimental results show the practicality of MergeTree. We also formally analyze the soundness and completeness of the proposed model.
{"title":"MergeTree: a Tree Model with Merged Nodes for Threat Induction","authors":"Ping Chen, Jingjing Hu, Zhitao Wu, Ruoting Xiong, Wei Ren","doi":"10.1145/3590003.3590012","DOIUrl":"https://doi.org/10.1145/3590003.3590012","url":null,"abstract":"Threat tree model can clearly organize threat induction information and thus is widely used for risk analysis in software assurance. Threat tree will grow to complicated structures, e.g., the number of nodes and branches, when the threat information grows to a huge volume. To extend the scalability of the threat tree model, we propose a tree model with merged nodes so as to largely decrease the number of nodes and branches. The formal model and dedicated algorithms are proposed in details. The experimental results show the practicality of MergeTree. We also formally analyze the soundness and completeness of the proposed model.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129220373","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}
The consistency of the dual-channel radar plays a crucial role in the performance of the dual-polarization radar. In theory, the performance of the two channels is required to be completely consistent, but it cannot be completely consistent due to the influence of hardware errors, temperature and noise in practical applications. Therefore, it is necessary to test the consistency of horizontal and vertical channels of radar regularly in business applications. Aiming at this problem, the receiving system of Jinan S-band dual polarization radar is tested by off-line manual testing and online automatic testing. The offline measurement uses two signal sources inside and outside the machine to test separately, it is found that the output power difference between the two channels is too large. After the connection lines of the two channels and the two-channel power divider are exchanged, the output power of the two channels is basically the same. The test results of noise coefficient and echo intensity of the two channels are good and meet the requirement of consistency. CW signal source and TS signal source are used for online automatic test. The amplitude and phase standard deviation of the CW signal and the TS signal meet the requirements of the index. However, TS signal is used to calibrate the received full link, which increases the loss of azimuth rotation joint, so its amplitude and phase standard difference are higher than CW signal. Therefore, it is necessary to test and correct the deviation caused by the rotation joint regularly after running for a long time for the dual polarization radar. The two measurement methods in this paper can effectively detect the dual-channel consistency of radar.
{"title":"Two-channel Conformance Test Analysis of S-band Dual-polarization Radar","authors":"Yuxin Gong, Qian Zhang, Weijia Sun, Chuancheng Ma, Yucheng Gong, Juxiu Wu, Xiqiang Yuan","doi":"10.1145/3590003.3590035","DOIUrl":"https://doi.org/10.1145/3590003.3590035","url":null,"abstract":"The consistency of the dual-channel radar plays a crucial role in the performance of the dual-polarization radar. In theory, the performance of the two channels is required to be completely consistent, but it cannot be completely consistent due to the influence of hardware errors, temperature and noise in practical applications. Therefore, it is necessary to test the consistency of horizontal and vertical channels of radar regularly in business applications. Aiming at this problem, the receiving system of Jinan S-band dual polarization radar is tested by off-line manual testing and online automatic testing. The offline measurement uses two signal sources inside and outside the machine to test separately, it is found that the output power difference between the two channels is too large. After the connection lines of the two channels and the two-channel power divider are exchanged, the output power of the two channels is basically the same. The test results of noise coefficient and echo intensity of the two channels are good and meet the requirement of consistency. CW signal source and TS signal source are used for online automatic test. The amplitude and phase standard deviation of the CW signal and the TS signal meet the requirements of the index. However, TS signal is used to calibrate the received full link, which increases the loss of azimuth rotation joint, so its amplitude and phase standard difference are higher than CW signal. Therefore, it is necessary to test and correct the deviation caused by the rotation joint regularly after running for a long time for the dual polarization radar. The two measurement methods in this paper can effectively detect the dual-channel consistency of radar.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127579495","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}
Domain adaptation (DA), the sub-realm of the transfer learning, attempts to deal with machine learning tasks on an unprocessed data domain with the different, but related labeled source domain. However, the classical DA can not efficiently deal with the cross-domain tasks in quantum mechanical scenarios. In this paper, the quantum kernel subspace alignment algorithm is proposed to achieve the procedure of DA by extracting the non-linear features with the quantum kernel method and aligning the two domains with the unitary evolution. The method presented in our work can be implemented on the universal quantum computer with the quantum basic linear algebra subroutines. Based on the algorithmic complexity analysis, the procedure of the QKSA can be implemented with at least quadratic quantum speedup compared with the classical DA algorithms.
{"title":"Quantum kernel subspace alignment for unsupervised domain adaptation","authors":"Xi He, Feiyu Du","doi":"10.1145/3590003.3590054","DOIUrl":"https://doi.org/10.1145/3590003.3590054","url":null,"abstract":"Domain adaptation (DA), the sub-realm of the transfer learning, attempts to deal with machine learning tasks on an unprocessed data domain with the different, but related labeled source domain. However, the classical DA can not efficiently deal with the cross-domain tasks in quantum mechanical scenarios. In this paper, the quantum kernel subspace alignment algorithm is proposed to achieve the procedure of DA by extracting the non-linear features with the quantum kernel method and aligning the two domains with the unitary evolution. The method presented in our work can be implemented on the universal quantum computer with the quantum basic linear algebra subroutines. Based on the algorithmic complexity analysis, the procedure of the QKSA can be implemented with at least quadratic quantum speedup compared with the classical DA algorithms.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"73 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123216542","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}
Monte Carlo Tree Search (MCTS) continues to confront the issue of exponential complexity growth in certain tasks when the planning horizon is excessively long, causing the trajectory’s past to grow exponentially. Our study presents Hierarchical MCTS Latent Skill Planner, an algorithm based on skill discovery that automatically identifies skills based on intrinsic rewards and integrates them with MCTS, enabling efficient decision-making at a higher level. In the grid world maze domain, we found that latent skill search outperformed the standard MCTS approach that do not contain skills in terms of efficiency and performance.
{"title":"Hierarchical Monte Carlo Tree Search for Latent Skill Planning","authors":"Yue Pei","doi":"10.1145/3590003.3590005","DOIUrl":"https://doi.org/10.1145/3590003.3590005","url":null,"abstract":"Monte Carlo Tree Search (MCTS) continues to confront the issue of exponential complexity growth in certain tasks when the planning horizon is excessively long, causing the trajectory’s past to grow exponentially. Our study presents Hierarchical MCTS Latent Skill Planner, an algorithm based on skill discovery that automatically identifies skills based on intrinsic rewards and integrates them with MCTS, enabling efficient decision-making at a higher level. In the grid world maze domain, we found that latent skill search outperformed the standard MCTS approach that do not contain skills in terms of efficiency and performance.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128165688","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}
The research of linear vibration has been very mature, but the gap nonlinear element widely exists in the actual structure and vibration system. The general linear vibration theory can not meet the needs of solving the nonlinear dynamic problems of clearance, and the research on the nonlinear vibration of clearance is essential. Based on the description function method, the forced response calculation of linear and nonlinear systems and the nonlinear detection method of clearance systems are studied in this paper. The forced response calculation of linear and clearance nonlinear systems mainly depends on the digital filtering method to realize the corresponding simulation of the system with clearance nonlinear. Based on the definition of linear system, a concrete scheme for judging whether the system contains the nonlinear characteristics of clearance is given. The position identification of the clearance nonlinear system based on the description function and the polynomial fitting inversion of the description function is studied and verified by an example.
{"title":"Research on Identification Method of Gap Nonlinear Vibration","authors":"Jialiang Sun, Jingying Liu","doi":"10.1145/3590003.3590138","DOIUrl":"https://doi.org/10.1145/3590003.3590138","url":null,"abstract":"The research of linear vibration has been very mature, but the gap nonlinear element widely exists in the actual structure and vibration system. The general linear vibration theory can not meet the needs of solving the nonlinear dynamic problems of clearance, and the research on the nonlinear vibration of clearance is essential. Based on the description function method, the forced response calculation of linear and nonlinear systems and the nonlinear detection method of clearance systems are studied in this paper. The forced response calculation of linear and clearance nonlinear systems mainly depends on the digital filtering method to realize the corresponding simulation of the system with clearance nonlinear. Based on the definition of linear system, a concrete scheme for judging whether the system contains the nonlinear characteristics of clearance is given. The position identification of the clearance nonlinear system based on the description function and the polynomial fitting inversion of the description function is studied and verified by an example.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124913953","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}
Abstract:Proof of work(POW) is one of the most widely used consensus method of bitcoin. In some chains, because of the large number of users, the huge amount of information interaction data, equipment hardware failure or malicious attacks on some nodes may cause communication delay, All of these may engender forks on the blockchain resulting data lose. Therefore, whether the blockchain protocol can achieve sufficient security in the asynchronous network environment with long delay, so how to reduce fork and user's data loss caused by long delay attack are important issue related to blockchain protocol. In this study,we optimized the existing model and proposed TOD to describe the evolution state of the main chain accurately.
{"title":"Optimized model analysis of blockchain PoW procotol under long delay attack","authors":"Tao Feng, Yufeng Liu","doi":"10.1145/3590003.3590008","DOIUrl":"https://doi.org/10.1145/3590003.3590008","url":null,"abstract":"Abstract:Proof of work(POW) is one of the most widely used consensus method of bitcoin. In some chains, because of the large number of users, the huge amount of information interaction data, equipment hardware failure or malicious attacks on some nodes may cause communication delay, All of these may engender forks on the blockchain resulting data lose. Therefore, whether the blockchain protocol can achieve sufficient security in the asynchronous network environment with long delay, so how to reduce fork and user's data loss caused by long delay attack are important issue related to blockchain protocol. In this study,we optimized the existing model and proposed TOD to describe the evolution state of the main chain accurately.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128727454","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}
In nowadays using different tools and apps is a basic need of people's behavior in life, but the security issues arising from the existence of source code plagiarism among tools and apps are likely to bring huge losses to companies and even countries, so detecting the existence of vulnerabilities or malicious code in software becomes an important part of protecting information and detecting software security. This project is based on the aspect of JAVA code vulnerability detection based on graph representation learning and software homology comparison to carry out research. This project will be based on the content of deep learning, using a large number of vulnerable source code, extracting its features, and transforming it into a graph so that it can be tested source code for comparison and report the vulnerability content. The main work and results of this project are as follows: 1.By extracting the example dataset and generating json files to save the feature information of relevant java code; by generating vector files, bytecode files and dot files, and batch extracting nodes and edges based on the contents of the dot files for subsequent machine learning use, the before and after steps and operations form a logical self-consistency to ensure the integrity of the project. 2.Through the study of graph neural networks and graph convolutional neural networks, relevant models are selected for machine learning using predecessor files and manual model tuning is performed to ensure good learning results and feedback for the machine learning part of the project. 3.This project training dataset negative samples for sard above the shared dataset, which contains 46636 java vulnerability source code, and dataset support environment, test dataset negative samples dataset also from sard, positive samples dataset are generated from the relevant person in charge. 4.Based on Graph Neural Network (GNN) and Graph Convolutional Neural Network (GCN), this project will design and implement a whole set of automated vulnerability detection system for java code. 5. All the related contents of this project, after the human extensive search of domestic and foreign related papers and materials, there are not all projects or contents similar to this project, the same papers and materials appear, all the problems involved in this project and related ideas are for the project this group of people thinking, looking for solutions.
{"title":"Graph representation learning and software homology matching based A study of JAVA code vulnerability detection techniques","authors":"Yibin Yang, Xin Bo, Zitong Wang, Xinrui Shao, Xinjie Xie","doi":"10.1145/3590003.3590028","DOIUrl":"https://doi.org/10.1145/3590003.3590028","url":null,"abstract":"In nowadays using different tools and apps is a basic need of people's behavior in life, but the security issues arising from the existence of source code plagiarism among tools and apps are likely to bring huge losses to companies and even countries, so detecting the existence of vulnerabilities or malicious code in software becomes an important part of protecting information and detecting software security. This project is based on the aspect of JAVA code vulnerability detection based on graph representation learning and software homology comparison to carry out research. This project will be based on the content of deep learning, using a large number of vulnerable source code, extracting its features, and transforming it into a graph so that it can be tested source code for comparison and report the vulnerability content. The main work and results of this project are as follows: 1.By extracting the example dataset and generating json files to save the feature information of relevant java code; by generating vector files, bytecode files and dot files, and batch extracting nodes and edges based on the contents of the dot files for subsequent machine learning use, the before and after steps and operations form a logical self-consistency to ensure the integrity of the project. 2.Through the study of graph neural networks and graph convolutional neural networks, relevant models are selected for machine learning using predecessor files and manual model tuning is performed to ensure good learning results and feedback for the machine learning part of the project. 3.This project training dataset negative samples for sard above the shared dataset, which contains 46636 java vulnerability source code, and dataset support environment, test dataset negative samples dataset also from sard, positive samples dataset are generated from the relevant person in charge. 4.Based on Graph Neural Network (GNN) and Graph Convolutional Neural Network (GCN), this project will design and implement a whole set of automated vulnerability detection system for java code. 5. All the related contents of this project, after the human extensive search of domestic and foreign related papers and materials, there are not all projects or contents similar to this project, the same papers and materials appear, all the problems involved in this project and related ideas are for the project this group of people thinking, looking for solutions.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128598786","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}
Zhiyong Wang, Xuefu Xiang, Kan Zeng, Zhenyu Zhang, Yanan Li, Dengpan Song
For the infrared search and tracking system, it is necessary to increase the ability to detect small infrared targets against complex backgrounds. YOLOX is a high-performance detector, but its detection performance is constrained when it uses data from low-resolution infrared images with small targets. However, occasionally design constraints and budgetary restraints will prevent the optical system and sensor resolution from being increased enough to improve image quality. Real-ESRGAN is used to solve this issue by reconstructing a high-resolution infrared image from its low-resolution counterpart, which will be used as YOLOX-S's input. Also, the YOLOX-S training strategy is modified further to make it appropriate for the detection of infrared small targets, including the Mosaic and MixUp data augmentation and the size of ground-truth. The average precision achieved by the suggested method in this work increases from 63.70% to 77.19%, which shows a considerable improvement in infrared small target detection when compared with the original model by inputting original images.
{"title":"Infrared small target detection based on the combination of single image super-resolution reconstruction and YOLOX","authors":"Zhiyong Wang, Xuefu Xiang, Kan Zeng, Zhenyu Zhang, Yanan Li, Dengpan Song","doi":"10.1145/3590003.3590104","DOIUrl":"https://doi.org/10.1145/3590003.3590104","url":null,"abstract":"For the infrared search and tracking system, it is necessary to increase the ability to detect small infrared targets against complex backgrounds. YOLOX is a high-performance detector, but its detection performance is constrained when it uses data from low-resolution infrared images with small targets. However, occasionally design constraints and budgetary restraints will prevent the optical system and sensor resolution from being increased enough to improve image quality. Real-ESRGAN is used to solve this issue by reconstructing a high-resolution infrared image from its low-resolution counterpart, which will be used as YOLOX-S's input. Also, the YOLOX-S training strategy is modified further to make it appropriate for the detection of infrared small targets, including the Mosaic and MixUp data augmentation and the size of ground-truth. The average precision achieved by the suggested method in this work increases from 63.70% to 77.19%, which shows a considerable improvement in infrared small target detection when compared with the original model by inputting original images.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130243717","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}
With the rapid development of the industrial internet of things(IIoT), the anomalies will cause significant damage to the ordinary operation of the industry. Anomaly detection work has increasingly become a hot spot. Although many related kinds of research exist, some problems still need to be solved. This paper proposes an efficient and robust semi-supervised capsule network (ADCapsNet) for anomaly detection by changing the convolution structure to better extract the features of the data and adding a new SecondaryCaps layer to better extract spatial relationships. Besides, we optimize the vector selection for dynamic anomaly detection routing and propose the scoring operation, the modified probability mechanism. The modified probability mechanism can widen the score gap between positive and negative samples. This model can accurately identify and output the spatial relationships. Extensive experiments on four datasets show that the ADCapsNet has good performance in anomaly detection.
{"title":"ADCapsNet: An Efficient and Robust Capsule Network Model for Anomaly Detection","authors":"Xiangyu Cai, Ruliang Xiao, Zhixia Zeng, Ping Gong, Shenmin Zhang","doi":"10.1145/3590003.3590009","DOIUrl":"https://doi.org/10.1145/3590003.3590009","url":null,"abstract":"With the rapid development of the industrial internet of things(IIoT), the anomalies will cause significant damage to the ordinary operation of the industry. Anomaly detection work has increasingly become a hot spot. Although many related kinds of research exist, some problems still need to be solved. This paper proposes an efficient and robust semi-supervised capsule network (ADCapsNet) for anomaly detection by changing the convolution structure to better extract the features of the data and adding a new SecondaryCaps layer to better extract spatial relationships. Besides, we optimize the vector selection for dynamic anomaly detection routing and propose the scoring operation, the modified probability mechanism. The modified probability mechanism can widen the score gap between positive and negative samples. This model can accurately identify and output the spatial relationships. Extensive experiments on four datasets show that the ADCapsNet has good performance in anomaly detection.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"103 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124603164","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：