Scalable Computing-Practice and Experience最新文献

{"title":"A Novel Hybrid Model Based on Dual Attention Networks for Significant Wave Height Forecast","authors":"Jiaming Tan, Junxing Zhu, Kaijun Ren, Xiaoyong Li, Renze Dong, Y. Lan","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00132","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00132","url":null,"abstract":"Extreme waves pose a severe threat to human life and property. Timely and accurate wave forecasting can help humans take appropriate measures in advance to avoid the risks caused by extreme waves. However, it is challenging to accurately forecast ocean waves due to their non-linear and non-smooth characteristics. To overcome this difficulty, we propose a significant wave height prediction method based on feature engineering and dual attention networks. Specifically, in feature engineering, we first decompose the original wave signal by the discrete wavelet transform to obtain several wavelets, after which we add the decomposed wavelets to the original data set for data augmentation, and finally, we use feature selection to determine the features of the final input network. We construct a sequence-to-sequence network with a dual attention mechanism, including the attention at the input layer and the encoder-decoder layer. Extensive experiments are conducted to verify the effectiveness of our method on 24-h and 48-h predictions. The results show that the proposed method outperforms the other methods compared.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"38 1","pages":"872-879"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80953089","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}

{"title":"Joint Global and Local Feature Learning Based on Facial StO2 for Stress Recognition","authors":"Dong Chen, Xinyu Liu, Tong Chen, Dairong Peng, Jiaxiu Wang","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00229","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00229","url":null,"abstract":"Stress is an emotional state that is inevitable in social life, it is of great importance to accurately identify different types of stress. In this paper, we use human facial tissue oxygen saturation (StO2) to identify four states of an individual’s baseline, emotional stress, high-intensity physical stress and low-intensity physical stress. For the stress classification, we proposed a network called GLNet that combines global and local StO2 features. Specifically, GLNet learns local features from facial regions that provide the effective stress information and global features from the whole face. Then, the two features are fused at decision-level and classified. In addition, a new data augmentation method was proposed to address the data imbalance, which generates new data by constructing a mixed network called MixNet that fuses the depth features of multiple data. Experimental results show that MixNet can effectively alleviate the problem of data imbalance and GLNet combined with the data expanded by MixNet achieved the best performance compared to previous methods of StO2 stress recognition.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"1 1","pages":"1209-1215"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79947377","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}

{"title":"Edge Resource Autoscaling for Hierarchical Federated Learning Over Public Edge Platforms","authors":"Mingliao Zhao, Kongyange Zhao, Zhi Zhou, Xu Chen","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00123","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00123","url":null,"abstract":"Federated learning promises to empower ubiquitous end devices to collaboratively learn a shared model in a privacy-preserving manner. To reduce the enormous and expensive wide-area-network (WAN) traffic incurred by the traditional two-tiered cloud-device federated learning, hierarchical federated learning over cloud-edge-device has been proposed recently. With hierarchical federated learning, edge servers are leveraged as intermediaries to perform local model aggregations to reduce the model updates aggregated by the centralized cloud. Considering the emerging public edge platforms such as Aliyun Edge Node Service that rent edge servers to users in an on-demand manner, we present AutoEdge, an edge server autoscaling framework for hierarchical federated learning. The goal of AutoEdge is to autoscale edge servers against dynamical device participants in a cost-efficient manner. Achieving this goal is challenging since the underlying long-term optimization problem is NP-hard involves the future system information. To attack these challenges, AutoEdge first applies regularization technique to decompose the long-term problem into a set of solvable fractional subproblems. Then, adopting a randomized dependent rounding scheme, AutoEdge further rounds the fractional solutions to a near-optimal and feasible integral solution. AutoEdge achieves outstanding performance guarantee, as verified by both rigorous theoretical analysis and extensive trace-driven simulations.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"70 1","pages":"806-814"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79953119","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}

{"title":"Performance Evaluation of Hierarchical Federated Learning Networks Based on Stochastic Network Calculus","authors":"Yashi Dang, Zhuo Li, Xin Chen","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00255","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00255","url":null,"abstract":"Analyzing the key factors affecting the delay of hierarchical federated learning and reducing the generation of delay is an important issue to be addressed. In this paper, we analyze the hierarchical federated learning network in the case of simultaneous access of mobile devices and model the arrival process and service process of data streams satisfying Poisson distribution. This paper analyzes the delay bound of the hierarchical federated learning network under a round of global updates using stochastic network calculus. We model a more realistic service model by considering the service rate variation of edge servers due to channel fading and other factors when analyzing the delay bound of the wireless access network. Finally, we analyze the parameters affecting the end-to-end delay performance of the hierarchical federated learning network in numerical analysis. The factors that affect the latency are the number of mobile nodes, the number of edge nodes, and the arrival rate of the data flow.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"34 1","pages":"1790-1795"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81454991","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}

{"title":"A Language-Agnostic Framework with Bidirectional Syntactic Graph Convolutional Networks for Cross-Lingual Aspect Term Extraction","authors":"Yaxin Cui, Baojie Tian, Junlin Wang, Yan Zhou, Songlin Hu","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00215","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00215","url":null,"abstract":"Aspect term extraction is a vital sub-task of sentiment analysis, which aims to extract explicit product attributes in customer reviews. Unfortunately, many languages lack sufficient labeled data, so researchers focus on Cross-lingual Aspect Term Extraction (XATE) to fully use sufficient data in other languages. Most recent cross-lingual methods focus on semantic alignment and data augmentation, but lack research on language structure, including syntax and lexicality. To this end, we propose a Language-Agnostic framework with Bidirectional Syntactic Graph Convolutional Networks (LA-BSGCN) for XATE. It is based on the idea that the topological structures of syntactic dependencies and the lexical tags across different languages are similar. We design a multi-layer bidirectional GCN, which can encode the syntactic tree more accurately. Furthermore, to reduce the lexicality semantic gap between different languages, we encode named entity recognition (NER) and part of speech (POS) information into our model. We conduct six pairs of cross-lingual experiments on SemEval2016 Task5 datasets. The results show that our LA-BSGCN significantly reduces the semantic gap and outperforms the state-of-the-art methods. For reproducibility, our code for this paper is available at github.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"33 1","pages":"1488-1495"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81894781","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}

{"title":"Multimodal Sarcasm Detection Based on Multimodal Sentiment Co-training","authors":"Yi Liu, Zengwei Zheng, Binbin Zhou, Jianhua Ma, Lin Sun, Ruichen Xia","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00090","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00090","url":null,"abstract":"Sarcasm detection is a difficult task in sentiment analysis because sarcasm often includes both positive and negative sentiments, making it difficult to identify. In recent years, visual information has been used to study sarcasm in social media data. Based on sentiment contrast in image and text, this paper proposes a Multimodal Sentiment and Sarcasm Gradient Co-training (MSSGC) model. The model uses text and image feature sharing networks to explicitly learn image and text sentimental features from image and text sentiment datasets and integrates a cross-modal fusion module for Multimodal Sarcasm Detection (MSD). The training algorithm uses the sentimental features for sarcasm detection by weighting the sentiment and sarcasm classification gradients. Extensive experiments, including case studies, are performed to evaluate the MSSGC model. The results illustrate that the proposed model outperforms recent MSD models. The code is available at: https://github.com/vantree/MSSGC.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"11 1","pages":"508-515"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86572453","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}

{"title":"Minimum Delay Optimization for Message Scheduling in In-Vehicle Applications Based on Pheromone Resetting Strategy","authors":"Junqiang Jiang, Lunxin Xie, Duqun Zhou, Bo Fan","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00298","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00298","url":null,"abstract":"In-vehicle applications generally have latency requirements; serious safety accidents will likely be caused due to the applications failing to take the correct actions within a specified time frame. In this study, a Pheromone Resetting Ant Colony Optimization (PRACO) algorithm is proposed to address the calculation of the minimum response time of an application whose messages are transmitted by using the Controller Area Network with flexible data rates (CAN FD) bus. A Random Popup (RP) algorithm is equipped in PRACO to quickly obtain the valid message sequence, followed by resetting the pheromones on all paths if ants find a new optimal valid message sequence path. The minimum response delay of an in-vehicle application can be further got through a continuous iterative search and pheromone update. A Directed Acyclic Graph (DAG) workflow scheduling example and an Adaptive Cruise Control (ACC) application are used to conduct the simulation experiment. The results show that our PRACO algorithm significantly outperforms other static scheduling algorithms in obtaining the lowest response latency.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"1 1","pages":"2061-2068"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"89519270","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}

{"title":"GAS: GPU Allocation Strategy for Deep Learning Training Tasks","authors":"Yingwen Chen, Jianchen Han, Huan Zhou, Chen Chen","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00133","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00133","url":null,"abstract":"Nowadays, with the significant increasement of the deep learning training (DLT) task workload in GPU clusters, the number and the scale of GPU clusters grow rapidly. A crucial question is how to efficiently schedule DLT tasks with limited cluster resources. Existing GPU schedulers do not fully consider the connection between users and clusters, and few methods optimize the GPU allocation of DLT tasks. In this study, we propose a scheduling framework for GPU clusters, which improves performance and reduces energy consumption of clusters. We first analyze the relationship between the characteristics of performance and energy consumption and the task configurations for DLT tasks. Then, we propose a prediction method to predict the completion time and energy consumption of DLT tasks. To make better use of cluster resources, based on the prediction model, we propose GAS, which adopts the GPU Allocation Strategy by specifying the parallelism for DLT tasks. Compared to FIFO and SJF schedulers, GAS reduces the makespan by 19.6%-19.8%, reduces the average queueing time by 84.4%-93.9% and reduces the energy consumption by 22.2%22.5%. For users, GAS also reduces the cost of users by 21.3%21.6%. The large-scale simulation experiment further illustrates the effectiveness and scalability of GAS.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"16 1","pages":"880-887"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87304164","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}

{"title":"Accuracy Indoor Localization Based on Fuzzy Transfer Learning Model","authors":"Sheng Wu, Yanhu Ji, Licai Zhu, Liang Zhao, Hao Yang","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00064","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00064","url":null,"abstract":"Location-based services greatly facilitate people’s daily life, which puts forward higher requirements for the location calculation of target objects in different environments. Since the fingerprint positioning method does not require additional special equipment and easy to implement, it has become one of the most attractive solutions. In order to ensure the positioning accuracy, this method requires complete sampling of the fingerprints of the positioning area, so a lot of sampling costs are required. In particular, when sampling buildings with multiple floors, the labor and time of the entire sampling process will increase dramatically. At the same time, certain floors or rooms may not be allowed to open, so their fingerprints cannot be sampled. In fact, the floor structures of buildings are mostly similar or the same, such as office buildings, hotels. Therefore, this paper proposes a fuzzy transfer learning model and builds the corresponding prototype system FTLoc. On the premise of ensuring the positioning accuracy of different floors, the system greatly reduces the sampling cost of the entire building. First, for the complete fingerprint data (source domain) of a certain floor, we mine the fingerprint features fine-grained to generate a short-term feature set for each sampling point. Then, according to the sparsity and timing of short-term features, we design an optimized SELSTM, and obtain an effective localization model as the PreModel for transfer learning. Finally, fuzzy clustering is used to add category labels to the source domain data and target domain data, and input them into PreModel to realize the localization model transfer, so as to avoid their data distribution differences affecting the transfer effect as much as possible. FTLoc is fully validated in a multi-storey building. According to the experimental results, when using the first floor sampling data as the source domain, the errors of the FTLoc system on the adjacent floor (second floor) are 1.38 meters (sampling rate = 80%) and 2.33 meters (sampling rate = 30%). The average errors in non-adjacent layers (three, four, five) are 1.92 meters (sampling rate = 80%), 2.87 meters (sampling rate = 30%). Compared with traditional migration, the FTLoc system increased by 18.1% and 12.6% respectively. At the same time, the experiment verified that the error jitter multiple of the FTLoc system under different devices and different sampling densities does not exceed 1.5. Therefore, the FTLoc system designed in this paper ensures the transfer learning effect of different floors, and has good robustness and reliability. In actual positioning applications, the system greatly reduces the sampling cost and achieves high-precision positioning at the same time.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"4 1","pages":"293-300"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87667512","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}

{"title":"Lightweight Collaborative Inferencing for Real-Time Intrusion Detection in IoT Networks","authors":"Gabriel A. Morales, Jingye Xu, Dakai Zhu, Rocky Slavin","doi":"10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00076","DOIUrl":"https://doi.org/10.1109/SmartWorld-UIC-ATC-ScalCom-DigitalTwin-PriComp-Metaverse56740.2022.00076","url":null,"abstract":"The security in Internet-of-Things (IoT) networks becomes increasingly important with the growing popularity of IoT devices and their wide applications (e.g., critical infrastructure monitoring). However, traditional intrusion detection systems (IDS) are not suitable for IoT networks due to their large resource requirements. Moreover, IoT networks tend to have multiple access points for IoT devices and thus benefit from a distributed framework to enable collaborative prevention of potential attacks. To this end, we propose a lightweight collaborative distributed network IDS (NIDS) based on widely-utilized machine learning (ML) models, which are trained through a federated learning framework with two known datasets. We evaluate the distributed NIDS using the trained ML models on an IoT network testbed under seven types of attacks in comparison with Snort (a state-of-the-art IDS) and a centralized implementation of our proposed NIDS. An offline benchmark is also designed to measure the system’s performance with regard to resource usage and response time. Our results show that the proposed distributed NIDS outperforms Snort in identifying malicious traffic and achieves a much lower false positive rate compared to the centralized version in real-time for all seven types of network attacks tested.","PeriodicalId":43791,"journal":{"name":"Scalable Computing-Practice and Experience","volume":"20 1","pages":"392-400"},"PeriodicalIF":1.1,"publicationDate":"2022-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87876975","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}