Pub Date : 2021-04-12DOI: 10.1080/0952813X.2021.1908433
Xianhui Zhang
ABSTRACT This paper focuses on the convergence analysis for a patch structure Nicholson’s blowflies system involving an oscillating death rate and multiple different time-varying delays. By using inequality techniques and concise mathematical analysis proof, some sufficient criteria are established to guarantee the global exponential convergence of the zero equilibrium point for the addressed system. Our results are novel and supplement some existing ones. Furthermore, the effectiveness and feasibility of the obtained results are demonstrated by some numerical simulations.
{"title":"Convergence analysis of a patch structure Nicholson’s blowflies system involving an oscillating death rate","authors":"Xianhui Zhang","doi":"10.1080/0952813X.2021.1908433","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1908433","url":null,"abstract":"ABSTRACT This paper focuses on the convergence analysis for a patch structure Nicholson’s blowflies system involving an oscillating death rate and multiple different time-varying delays. By using inequality techniques and concise mathematical analysis proof, some sufficient criteria are established to guarantee the global exponential convergence of the zero equilibrium point for the addressed system. Our results are novel and supplement some existing ones. Furthermore, the effectiveness and feasibility of the obtained results are demonstrated by some numerical simulations.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"50 1","pages":"663 - 672"},"PeriodicalIF":2.2,"publicationDate":"2021-04-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86814642","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-04-06DOI: 10.1080/0952813X.2021.1908432
S. Gao, Changzhu Zhang, Zhuping Wang, Hao Zhang, Chao Huang
ABSTRACT Real-time semantic segmentation aims to generate high-quality prediction in limited time. Recently, with the development of many related potential applications, such as autonomous driving, robot sensing and augmented reality devices, semantic segmentation is desirable to make a trade-off between accuracy and inference speed with limited computation resources. This paper introduces a novel effective and light-weighted network based on Separable Pyramid Module (SPM) to achieve competitive accuracy and inference speed with fewer parameters and computation. Our proposed SPM unit utilises factorised convolution and dilated convolution in the form of a feature pyramid to build a bottleneck structure, which extracts local and context information in a simple but effective way. Experiments on Cityscapes and Camvid datasets demonstrate our superior trade-off between speed and precision. Without pre-training or any additional processing, our SPMNet achieves 71.22% mIoU on Cityscapes test set at the speed of 94 FPS on a single GTX 1080Ti GPU card.
{"title":"SPMNet: A light-weighted network with separable pyramid module for real-time semantic segmentation","authors":"S. Gao, Changzhu Zhang, Zhuping Wang, Hao Zhang, Chao Huang","doi":"10.1080/0952813X.2021.1908432","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1908432","url":null,"abstract":"ABSTRACT Real-time semantic segmentation aims to generate high-quality prediction in limited time. Recently, with the development of many related potential applications, such as autonomous driving, robot sensing and augmented reality devices, semantic segmentation is desirable to make a trade-off between accuracy and inference speed with limited computation resources. This paper introduces a novel effective and light-weighted network based on Separable Pyramid Module (SPM) to achieve competitive accuracy and inference speed with fewer parameters and computation. Our proposed SPM unit utilises factorised convolution and dilated convolution in the form of a feature pyramid to build a bottleneck structure, which extracts local and context information in a simple but effective way. Experiments on Cityscapes and Camvid datasets demonstrate our superior trade-off between speed and precision. Without pre-training or any additional processing, our SPMNet achieves 71.22% mIoU on Cityscapes test set at the speed of 94 FPS on a single GTX 1080Ti GPU card.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"17 1","pages":"651 - 662"},"PeriodicalIF":2.2,"publicationDate":"2021-04-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73175055","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-04-05DOI: 10.1080/0952813X.2021.1908429
Saejoon Kim, Hyuksoo Kim
ABSTRACT Trend following strategies are well-known to exhibit excellent excess return performance across a wide range of asset classes in various global markets. For the equity asset class in particular, while the securities selection part is relatively a straightforward procedure, the weight allocation part is more debatable and it has traditionally been identified with the equal-weighted allocation strategy. In this paper, we examine security’s own return-based weight allocation strategy for trend following investing and find that this strategy generates superior returns to several well-established weight allocation schemes. In particular, if the true return of the holding period is used ex ante for weight allocation, it is found that this strategy can generate absolutely huge excess returns. Motivated by this finding, we investigate the efficacy of machine learning techniques for regression of securities’ returns to improve the weight calculation in this framework. Empirical results indicate that deep learning provides the means of regression with which largest excess return gains are possible. In particular, it is demonstrated that the return-based weight allocation strategy defined by our proposed deep learning architecture produces substantial abnormal returns outperforming all other broadly recognised weight allocation schemes compared in this paper.
{"title":"Deep asset allocation for trend following investing","authors":"Saejoon Kim, Hyuksoo Kim","doi":"10.1080/0952813X.2021.1908429","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1908429","url":null,"abstract":"ABSTRACT Trend following strategies are well-known to exhibit excellent excess return performance across a wide range of asset classes in various global markets. For the equity asset class in particular, while the securities selection part is relatively a straightforward procedure, the weight allocation part is more debatable and it has traditionally been identified with the equal-weighted allocation strategy. In this paper, we examine security’s own return-based weight allocation strategy for trend following investing and find that this strategy generates superior returns to several well-established weight allocation schemes. In particular, if the true return of the holding period is used ex ante for weight allocation, it is found that this strategy can generate absolutely huge excess returns. Motivated by this finding, we investigate the efficacy of machine learning techniques for regression of securities’ returns to improve the weight calculation in this framework. Empirical results indicate that deep learning provides the means of regression with which largest excess return gains are possible. In particular, it is demonstrated that the return-based weight allocation strategy defined by our proposed deep learning architecture produces substantial abnormal returns outperforming all other broadly recognised weight allocation schemes compared in this paper.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"28 1","pages":"599 - 619"},"PeriodicalIF":2.2,"publicationDate":"2021-04-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81650328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-04-03DOI: 10.1080/0952813X.2021.1907795
Amit Singh, R. Ranjan, A. Tiwari
ABSTRACT Credit card fraud is one of the biggest cybercrimes faced by users. Intelligent machine learning based fraudulent transaction detection systems are very effective in real-world scenarios. However, while designing these systems, machine learning approaches suffer from the problem of imbalanced data, i.e. imbalanced class distribution. Therefore, balancing the dataset becomes an imperative sub-task. Investigation of state-of-the-art approaches reveals that there is a need for a systematic study of class imbalance handling strategies to design an intelligent and capable system to detect the fraudulent transaction. This work aims to provide a comparative study of different class imbalance handling methods. To compare the effectiveness and efficiency of different class imbalance approaches in conjunction with state-of-the-art classification approaches, we have performed an extensive experimental study. We compared these methods on many performance indicators such as Precision, Recall, K-fold Cross-validation, AUC-ROC curve and execution time. In this study, we found that the Oversampling followed by Undersampling methods performs well for ensemble classification models such as AdaBoost, XGBoost and Random Forest.
{"title":"Credit Card Fraud Detection under Extreme Imbalanced Data: A Comparative Study of Data-level Algorithms","authors":"Amit Singh, R. Ranjan, A. Tiwari","doi":"10.1080/0952813X.2021.1907795","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1907795","url":null,"abstract":"ABSTRACT Credit card fraud is one of the biggest cybercrimes faced by users. Intelligent machine learning based fraudulent transaction detection systems are very effective in real-world scenarios. However, while designing these systems, machine learning approaches suffer from the problem of imbalanced data, i.e. imbalanced class distribution. Therefore, balancing the dataset becomes an imperative sub-task. Investigation of state-of-the-art approaches reveals that there is a need for a systematic study of class imbalance handling strategies to design an intelligent and capable system to detect the fraudulent transaction. This work aims to provide a comparative study of different class imbalance handling methods. To compare the effectiveness and efficiency of different class imbalance approaches in conjunction with state-of-the-art classification approaches, we have performed an extensive experimental study. We compared these methods on many performance indicators such as Precision, Recall, K-fold Cross-validation, AUC-ROC curve and execution time. In this study, we found that the Oversampling followed by Undersampling methods performs well for ensemble classification models such as AdaBoost, XGBoost and Random Forest.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"1 1","pages":"571 - 598"},"PeriodicalIF":2.2,"publicationDate":"2021-04-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90988237","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-04-01DOI: 10.1080/0952813X.2021.1907793
K. Boudjit, N. Ramzan
ABSTRACT Recent advancements in the field of Artificial Intelligence (AI) have provided an opportunity to create autonomous devices, robots, and machines characterised particularly with the ability to make decisions and perform tasks without human mediation. One of these devices, Unmanned Aerial Vehicles (UAVs) or drones are widely used to perform tasks like surveillance, search and rescue, object detection and target tracking, and many more. Efficient real-time object detection in aerial videos is an urgent need, especially with the increasing use of UAV in various fields. The sensitivity in performing said tasks demands that drones must be efficient and reliable. This paper presents our research progress in the development of applications for the identification and detection of person using the convolutional neural networks (CNN) YOLO-v2 based on the camera of drone. The position and state of the person are determined with deep-learning-based computer vision. The person detection results show that YOLO-v2 detects and classifies object with a high level of accuracy. For real-time tracking, the tracking algorithm responds faster than conventionally used approaches, efficiently tracking the detected person without losing it from sight.
{"title":"Human detection based on deep learning YOLO-v2 for real-time UAV applications","authors":"K. Boudjit, N. Ramzan","doi":"10.1080/0952813X.2021.1907793","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1907793","url":null,"abstract":"ABSTRACT Recent advancements in the field of Artificial Intelligence (AI) have provided an opportunity to create autonomous devices, robots, and machines characterised particularly with the ability to make decisions and perform tasks without human mediation. One of these devices, Unmanned Aerial Vehicles (UAVs) or drones are widely used to perform tasks like surveillance, search and rescue, object detection and target tracking, and many more. Efficient real-time object detection in aerial videos is an urgent need, especially with the increasing use of UAV in various fields. The sensitivity in performing said tasks demands that drones must be efficient and reliable. This paper presents our research progress in the development of applications for the identification and detection of person using the convolutional neural networks (CNN) YOLO-v2 based on the camera of drone. The position and state of the person are determined with deep-learning-based computer vision. The person detection results show that YOLO-v2 detects and classifies object with a high level of accuracy. For real-time tracking, the tracking algorithm responds faster than conventionally used approaches, efficiently tracking the detected person without losing it from sight.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"17 1","pages":"527 - 544"},"PeriodicalIF":2.2,"publicationDate":"2021-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87826323","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-03-15DOI: 10.1080/0952813X.2021.1882001
Jagat Sesh Challa, Poonam Goyal, Ajinkya Kokandakar, D. Mantri, Pranet Verma, S. Balasubramaniam, Navneet Goyal
ABSTRACT Clustering of data streams has become very popular in recent times, owing to rapid rise of real-time streaming utilities that produce large amounts of data at varying inter-arrival rates. We propose AnyClus, a framework for anytime clustering of data streams. AnyClus uses a proposed variant of R-tree, AnyRTree, to capture the incoming stream objects arriving at variable rate, and to index them in the form of micro-clusters of hierarchical fashion. The leaf-level micro-clusters produced are aggregated and stored in a logarithmic tilted-time window framework (TTWF). Our extensive experimental analysis shows (i) the capability of AnyClus in handling variable stream speeds (upto 250k objects/second); (ii) its ability to produce micro-clusters of high purity (≈1) and compactness; (iii) effectiveness of AnyRTree in handling noise, capturing concept drift and preservation of spatial locality in the indexing of micro-clusters, when compared to the existing methods. We also propose a parallel framework, Any-MP-Clus, for anytime clustering of multiport data streams over commodity clusters. Any-MP-Clus uses AnyRTree at each computing node of the cluster (for each stream-port) and maintains the aggregated micro-clusters in TTWF. The experimental results on datasets of billions scale show that Any-MP-Clus is scalable, efficient and produces clustering of higher quality.
{"title":"Anytime clustering of data streams while handling noise and concept drift","authors":"Jagat Sesh Challa, Poonam Goyal, Ajinkya Kokandakar, D. Mantri, Pranet Verma, S. Balasubramaniam, Navneet Goyal","doi":"10.1080/0952813X.2021.1882001","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1882001","url":null,"abstract":"ABSTRACT Clustering of data streams has become very popular in recent times, owing to rapid rise of real-time streaming utilities that produce large amounts of data at varying inter-arrival rates. We propose AnyClus, a framework for anytime clustering of data streams. AnyClus uses a proposed variant of R-tree, AnyRTree, to capture the incoming stream objects arriving at variable rate, and to index them in the form of micro-clusters of hierarchical fashion. The leaf-level micro-clusters produced are aggregated and stored in a logarithmic tilted-time window framework (TTWF). Our extensive experimental analysis shows (i) the capability of AnyClus in handling variable stream speeds (upto 250k objects/second); (ii) its ability to produce micro-clusters of high purity (≈1) and compactness; (iii) effectiveness of AnyRTree in handling noise, capturing concept drift and preservation of spatial locality in the indexing of micro-clusters, when compared to the existing methods. We also propose a parallel framework, Any-MP-Clus, for anytime clustering of multiport data streams over commodity clusters. Any-MP-Clus uses AnyRTree at each computing node of the cluster (for each stream-port) and maintains the aggregated micro-clusters in TTWF. The experimental results on datasets of billions scale show that Any-MP-Clus is scalable, efficient and produces clustering of higher quality.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"62 1","pages":"399 - 429"},"PeriodicalIF":2.2,"publicationDate":"2021-03-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83633697","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-03-08DOI: 10.1080/0952813X.2021.1883744
Qian Cao, Guoqiu Wang
ABSTRACT In this manuscript, inertial neural networks with both time- varying and distributed delays are studied. Applying inequality techniques and Lyapunov function approach, a new sufficient condition that guarantees the existence and exponential stability of periodic solutions for the addressed networks is presented. The obtained results supplement some earlier publications that deal with the periodic solutions of inertial neural networks with time- varying delays. Computer simulations are displayed to check the derived analytical results.
{"title":"New findings on global exponential stability of inertial neural networks with both time-varying and distributed delays","authors":"Qian Cao, Guoqiu Wang","doi":"10.1080/0952813X.2021.1883744","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1883744","url":null,"abstract":"ABSTRACT In this manuscript, inertial neural networks with both time- varying and distributed delays are studied. Applying inequality techniques and Lyapunov function approach, a new sufficient condition that guarantees the existence and exponential stability of periodic solutions for the addressed networks is presented. The obtained results supplement some earlier publications that deal with the periodic solutions of inertial neural networks with time- varying delays. Computer simulations are displayed to check the derived analytical results.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"21 1","pages":"469 - 482"},"PeriodicalIF":2.2,"publicationDate":"2021-03-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82430516","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-03-04DOI: 10.1080/0952813x.2020.1735529
L. Safatly, M. Baydoun, Mohamad Alipour, A. Al-Takach, K. Atab, M. Al‐Husseini, A. El-Hajj, H. Ghaziri
ABSTRACT The current landmine clearance methods mostly rely on the manual use of metal detectors (MDs) and on the deminer’s experience in differentiating between the sounds emitted due to the presence of a landmine or of harmless clutter. This process suffers from high false-alarm rates, which renders the demining effort slow and costly. In this paper, we report our attempts in using machine learning for decision making in the demining process. We have created our own database of the MD responses corresponding to landmines and/or clutter. A robotic rail is designed and assembled to accurately measure these responses and build the database. Several machine learning models are then developed using the database with the aim of detecting the presence of landmines and classifying them. It is shown that the classification algorithms lead to accurately discriminating the landmines and distinguishing between different buried objects including mines or other items based on the metal detector delivered data or signature.
{"title":"Detection and classification of landmines using machine learning applied to metal detector data","authors":"L. Safatly, M. Baydoun, Mohamad Alipour, A. Al-Takach, K. Atab, M. Al‐Husseini, A. El-Hajj, H. Ghaziri","doi":"10.1080/0952813x.2020.1735529","DOIUrl":"https://doi.org/10.1080/0952813x.2020.1735529","url":null,"abstract":"ABSTRACT The current landmine clearance methods mostly rely on the manual use of metal detectors (MDs) and on the deminer’s experience in differentiating between the sounds emitted due to the presence of a landmine or of harmless clutter. This process suffers from high false-alarm rates, which renders the demining effort slow and costly. In this paper, we report our attempts in using machine learning for decision making in the demining process. We have created our own database of the MD responses corresponding to landmines and/or clutter. A robotic rail is designed and assembled to accurately measure these responses and build the database. Several machine learning models are then developed using the database with the aim of detecting the presence of landmines and classifying them. It is shown that the classification algorithms lead to accurately discriminating the landmines and distinguishing between different buried objects including mines or other items based on the metal detector delivered data or signature.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"6 5 1","pages":"203 - 226"},"PeriodicalIF":2.2,"publicationDate":"2021-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78491049","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-03-04DOI: 10.1080/0952813X.2020.1725652
Mirsaeid Hosseini Shirvani
ABSTRACT Cloud computing became an inevitable information technology industry. Despite its several plus points such as economy of scale and rapid elasticity, it suffers from vendor lock-in, resource limitation and cybersecurity attacks in which it leads business discontinuity or even business failure. Multi-cloud, on the other hand, can be trustable paradigm to obviate obstacles such as aforesaid unpleasant features of a single cloud. One of the biggest challenges is to know which cloud is commensurate with user’s business process with regards to security objectives. To this end, the new method is presented to quantify the amount of cloud security risk (CSR) in regards to user’s business process. Therefore, in this paper, the web service composition problem is formulated to bi-objective optimisation problem with service cost and multi-cloud risk viewpoints in ever-increasing multi-cloud environment (MCE) in which each provider has its variable pricing policy and different security level. It is obviously an NP-Hard problem. To solve the combinatorial problem, we develop a bi-objective time-varying particle swarm optimisation (BOTV-PSO) algorithm. The parameters are tuned based on elapsed time so a good balance between exploration and exploitation is achieved. To illustrate the effectiveness of proposed algorithm, we defined several scenarios and compared the performance of proposed algorithm with multi-objective GA-based (MOGA) optimiser, a single objective genetic algorithm (SOGA) that only optimises cost function and neglects CSR, and multi-objective simulated annealing algorithm (MOSA). The experimental results showed the superiority of proposed BOTV-PSO against other approaches in terms of convergence, diversity, fitness, performance, and even scalability.
{"title":"Bi-objective web service composition problem in multi-cloud environment: a bi-objective time-varying particle swarm optimisation algorithm","authors":"Mirsaeid Hosseini Shirvani","doi":"10.1080/0952813X.2020.1725652","DOIUrl":"https://doi.org/10.1080/0952813X.2020.1725652","url":null,"abstract":"ABSTRACT Cloud computing became an inevitable information technology industry. Despite its several plus points such as economy of scale and rapid elasticity, it suffers from vendor lock-in, resource limitation and cybersecurity attacks in which it leads business discontinuity or even business failure. Multi-cloud, on the other hand, can be trustable paradigm to obviate obstacles such as aforesaid unpleasant features of a single cloud. One of the biggest challenges is to know which cloud is commensurate with user’s business process with regards to security objectives. To this end, the new method is presented to quantify the amount of cloud security risk (CSR) in regards to user’s business process. Therefore, in this paper, the web service composition problem is formulated to bi-objective optimisation problem with service cost and multi-cloud risk viewpoints in ever-increasing multi-cloud environment (MCE) in which each provider has its variable pricing policy and different security level. It is obviously an NP-Hard problem. To solve the combinatorial problem, we develop a bi-objective time-varying particle swarm optimisation (BOTV-PSO) algorithm. The parameters are tuned based on elapsed time so a good balance between exploration and exploitation is achieved. To illustrate the effectiveness of proposed algorithm, we defined several scenarios and compared the performance of proposed algorithm with multi-objective GA-based (MOGA) optimiser, a single objective genetic algorithm (SOGA) that only optimises cost function and neglects CSR, and multi-objective simulated annealing algorithm (MOSA). The experimental results showed the superiority of proposed BOTV-PSO against other approaches in terms of convergence, diversity, fitness, performance, and even scalability.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"333 1","pages":"179 - 202"},"PeriodicalIF":2.2,"publicationDate":"2021-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72845726","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-02-23DOI: 10.1080/0952813X.2021.1883745
Guangbin Wang, Jixiang Du, Hongbo Zhang
ABSTRACT Describing video content using natural language is an important part of video understanding. It needs to not only understand the spatial information on video, but also capture the motion information. Meanwhile, video captioning is a cross-modal problem between vision and language. Traditional video captioning methods follow the encoder-decoder framework that transfers the video to sentence. But the semantic alignment from sentence to video is ignored. Hence, finding a discriminative visual representation as well as narrowing the semantic gap between video and text has great influence on generating accurate sentences. In this paper, we propose an approach based on multi-feature fusion refine network (MFRN), which can not only capture the spatial information and motion information by exploiting multi-feature fusion, but also can get better semantic aligning of different models by designing a refiner to explore the sentence to video stream. The main novelties and advantages of our method are: (1) multi-feature fusion: Both two-dimension convolutional neural networks and three-dimension convolutional neural networks pre-trained on ImageNet and Kinetic respectively are used to construct spatial information and motion information, and then fused to get better visual representation. (2) Sematic alignment refiner: the refiner is designed to restrain the decoder and reproduce the video features to narrow semantic gap between different modal. Experiments on two widely used datasets demonstrate our approach achieves state-of-the-art performance in terms of BLEU@4, METEOR, ROUGE and CIDEr metrics.
{"title":"Multi-feature fusion refine network for video captioning","authors":"Guangbin Wang, Jixiang Du, Hongbo Zhang","doi":"10.1080/0952813X.2021.1883745","DOIUrl":"https://doi.org/10.1080/0952813X.2021.1883745","url":null,"abstract":"ABSTRACT Describing video content using natural language is an important part of video understanding. It needs to not only understand the spatial information on video, but also capture the motion information. Meanwhile, video captioning is a cross-modal problem between vision and language. Traditional video captioning methods follow the encoder-decoder framework that transfers the video to sentence. But the semantic alignment from sentence to video is ignored. Hence, finding a discriminative visual representation as well as narrowing the semantic gap between video and text has great influence on generating accurate sentences. In this paper, we propose an approach based on multi-feature fusion refine network (MFRN), which can not only capture the spatial information and motion information by exploiting multi-feature fusion, but also can get better semantic aligning of different models by designing a refiner to explore the sentence to video stream. The main novelties and advantages of our method are: (1) multi-feature fusion: Both two-dimension convolutional neural networks and three-dimension convolutional neural networks pre-trained on ImageNet and Kinetic respectively are used to construct spatial information and motion information, and then fused to get better visual representation. (2) Sematic alignment refiner: the refiner is designed to restrain the decoder and reproduce the video features to narrow semantic gap between different modal. Experiments on two widely used datasets demonstrate our approach achieves state-of-the-art performance in terms of BLEU@4, METEOR, ROUGE and CIDEr metrics.","PeriodicalId":15677,"journal":{"name":"Journal of Experimental & Theoretical Artificial Intelligence","volume":"13 1","pages":"483 - 497"},"PeriodicalIF":2.2,"publicationDate":"2021-02-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82416243","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"计算机科学","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：