Pub Date : 2020-12-01DOI: 10.1109/CSCI51800.2020.00265
Hyo-Kyun Kim, Tae-Sun Chung
This paper introduces an improved ANN (All Nearest Neighbor) algorithm using the SCL (Standard Clustered Loop) algorithm to reduce the consumption of computing resources that can occur when searching for the data object nearest to the query object in the process of executing the algorithm. Additionally, a method to improve ANN algorithm is proposed. When the algorithm is executed, it is a situation in which the user finds a data object adjacent to the user. In this case, our technique applies the criteria set provided by users.
{"title":"All Nearest Neighbors Query Including Scores Road Network","authors":"Hyo-Kyun Kim, Tae-Sun Chung","doi":"10.1109/CSCI51800.2020.00265","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00265","url":null,"abstract":"This paper introduces an improved ANN (All Nearest Neighbor) algorithm using the SCL (Standard Clustered Loop) algorithm to reduce the consumption of computing resources that can occur when searching for the data object nearest to the query object in the process of executing the algorithm. Additionally, a method to improve ANN algorithm is proposed. When the algorithm is executed, it is a situation in which the user finds a data object adjacent to the user. In this case, our technique applies the criteria set provided by users.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131070713","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00103
Helton Agbewonou Yawovi, Tadachika Ozono, T. Shintani
With the increasing number of motorized vehicles, road accidents are increasing year by year all over the world.. After an accident, the police investigate the circumstances of the incident and determine each actor’s responsibilities. Our goal is to create a police support system. We focused on a multi-agent system that predicts each actor’s responsibility in a road accident (especially crossroad accidents). The system uses the driving recorder video of a vehicle as the input data source, and it outputs the prediction of the responsibility of each actor in the accident. It consists of three agents: (1) Crash time detection and crash video split into images; (2) Traffic signs detection in the crash video; (3) Responsibility prediction using a knowledge system.
{"title":"Crossroad Accident Responsibility Prediction Based on a Multi-agent System","authors":"Helton Agbewonou Yawovi, Tadachika Ozono, T. Shintani","doi":"10.1109/CSCI51800.2020.00103","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00103","url":null,"abstract":"With the increasing number of motorized vehicles, road accidents are increasing year by year all over the world.. After an accident, the police investigate the circumstances of the incident and determine each actor’s responsibilities. Our goal is to create a police support system. We focused on a multi-agent system that predicts each actor’s responsibility in a road accident (especially crossroad accidents). The system uses the driving recorder video of a vehicle as the input data source, and it outputs the prediction of the responsibility of each actor in the accident. It consists of three agents: (1) Crash time detection and crash video split into images; (2) Traffic signs detection in the crash video; (3) Responsibility prediction using a knowledge system.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"134 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129300083","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00039
Ahmed Al Guqhaiman, Oluwatobi Akanbi, Amer Aljaedi, C. E. Chow
Underwater Wireless Sensor Networks (UWSNs) are liable to malicious attacks due to limited bandwidth, limited power, high propagation delay, path loss, and variable speed. The major differences between UWSNs and Terrestrial Wireless Sensor Networks (TWSNs) necessitate a new mechanism to secure UWSNs. The existing Media Access Control (MAC) and routing protocols have addressed the network performance of UWSNs, but are vulnerable to several attacks. The secure MAC and routing protocols must exist to detect Sybil, Blackhole, Wormhole, Hello Flooding, Acknowledgment Spoofing, Selective Forwarding, Sinkhole, and Exhaustion attacks. These attacks can disrupt or disable the network connection. Hence, these attacks can degrade the network performance and total loss can be catastrophic in some applications, like monitoring oil/gas spills. Several researchers have studied the security of UWSNs, but most of the works detect malicious attacks solely based on a certain predefined threshold. It is not optimal to detect malicious attacks after the threshold value is met. In this paper, we propose a multi-factor authentication model that is based on zero-knowledge proof to detect malicious activities and secure UWSNs from several attacks.
{"title":"Lightweight Multi-factor Authentication for Underwater Wireless Sensor Networks","authors":"Ahmed Al Guqhaiman, Oluwatobi Akanbi, Amer Aljaedi, C. E. Chow","doi":"10.1109/CSCI51800.2020.00039","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00039","url":null,"abstract":"Underwater Wireless Sensor Networks (UWSNs) are liable to malicious attacks due to limited bandwidth, limited power, high propagation delay, path loss, and variable speed. The major differences between UWSNs and Terrestrial Wireless Sensor Networks (TWSNs) necessitate a new mechanism to secure UWSNs. The existing Media Access Control (MAC) and routing protocols have addressed the network performance of UWSNs, but are vulnerable to several attacks. The secure MAC and routing protocols must exist to detect Sybil, Blackhole, Wormhole, Hello Flooding, Acknowledgment Spoofing, Selective Forwarding, Sinkhole, and Exhaustion attacks. These attacks can disrupt or disable the network connection. Hence, these attacks can degrade the network performance and total loss can be catastrophic in some applications, like monitoring oil/gas spills. Several researchers have studied the security of UWSNs, but most of the works detect malicious attacks solely based on a certain predefined threshold. It is not optimal to detect malicious attacks after the threshold value is met. In this paper, we propose a multi-factor authentication model that is based on zero-knowledge proof to detect malicious activities and secure UWSNs from several attacks.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"85 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130868500","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00019
Abdulaziz A. Alsulami, S. Zein-Sabatto
In recent years, there has been a rapid expansion in the development of Cyber-Physical Systems (CPS), which allows the physical components and the cyber components of a system to be fully integrated and interacted with each other and with the physical world. The commercial aviation industry is shifting towards Aviation Cyber-Physical Systems (ACPS) framework because it allows real-time monitoring and diagnostics, real-time data analytics, and the use of Artificial Intelligent technologies in decision making. Inevitably, ACPS is not immune to cyber-attacks due to integrating a network system, which introduces serious security threats. False Data Injection (FDI) attack is widely used against CPS. It is a serious threat to the integrity of the connected physical components. In this paper, we propose a novel security algorithm for detecting FDI attacks in the communication network of ACPS using Artificial Immune System (AIS). The algorithm was developed based on the negative selection approach. The negative selection algorithm is used to detect malicious network packets and drop them. Then, a Nonlinear Autoregressive Exogenous (NARX) network is used to predict packets that dropped by the negative selection algorithm. The developed algorithm was implemented and tested on a networked control system of commercial aircraft as an Aviation Cyber-physical system.
{"title":"Detection and Defense from False Data Injection Attacks In Aviation Cyber-Physical Systems Using Artificial Immune Systems","authors":"Abdulaziz A. Alsulami, S. Zein-Sabatto","doi":"10.1109/CSCI51800.2020.00019","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00019","url":null,"abstract":"In recent years, there has been a rapid expansion in the development of Cyber-Physical Systems (CPS), which allows the physical components and the cyber components of a system to be fully integrated and interacted with each other and with the physical world. The commercial aviation industry is shifting towards Aviation Cyber-Physical Systems (ACPS) framework because it allows real-time monitoring and diagnostics, real-time data analytics, and the use of Artificial Intelligent technologies in decision making. Inevitably, ACPS is not immune to cyber-attacks due to integrating a network system, which introduces serious security threats. False Data Injection (FDI) attack is widely used against CPS. It is a serious threat to the integrity of the connected physical components. In this paper, we propose a novel security algorithm for detecting FDI attacks in the communication network of ACPS using Artificial Immune System (AIS). The algorithm was developed based on the negative selection approach. The negative selection algorithm is used to detect malicious network packets and drop them. Then, a Nonlinear Autoregressive Exogenous (NARX) network is used to predict packets that dropped by the negative selection algorithm. The developed algorithm was implemented and tested on a networked control system of commercial aircraft as an Aviation Cyber-physical system.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131153767","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00184
Hussain Aljafer, Gary Cantrell
A topic of debate among Java teachers is whether to use an Integrated Development Environment (IDE) or a text editor paired with a command line compiler to teach introductory Java programming courses. Is it really in the student’s favor to start their programming journey using an IDE or is it better to just use any text editor to write the code and then compile and run the code using the command line? Which approach will help the students really understand the programming concepts and be able to debug their code when they have errors? In this paper we discuss the advantages and disadvantages of both approaches and we compare the performance of two groups of students in which the first group used a text editor with command line commands and the other group used Eclipse IDE.
{"title":"Learning and Teaching Undergraduate Introductory Programming Courses in Java – The Use of an IDE VS Command Line","authors":"Hussain Aljafer, Gary Cantrell","doi":"10.1109/CSCI51800.2020.00184","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00184","url":null,"abstract":"A topic of debate among Java teachers is whether to use an Integrated Development Environment (IDE) or a text editor paired with a command line compiler to teach introductory Java programming courses. Is it really in the student’s favor to start their programming journey using an IDE or is it better to just use any text editor to write the code and then compile and run the code using the command line? Which approach will help the students really understand the programming concepts and be able to debug their code when they have errors? In this paper we discuss the advantages and disadvantages of both approaches and we compare the performance of two groups of students in which the first group used a text editor with command line commands and the other group used Eclipse IDE.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127779954","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00306
José Prades, A. Salazar, G. Safont, L. Vergara
Some applications require knowing how many materials are present in the scene represented by a hyperspectral Image. In a previous paper, we presented an algorithm that estimated the number of materials in the scene using clustering principles. The proposed algorithm obtains a hierarchy of image partitions and selects a partition using a validation Index; the estimated number of materials is set to the number of dusters of the selected partition. In this algorithm, the user must provide the Image and the maximum number of materials that can be estimated (P). In this paper, we have extended our algorithm so that It does not require P as input parameter. The proposed method Iteratively performs the estimation for several increasing values of P and stops the process when a certain condition is met. The results obtained with five hyperspectral Images show that our algorithm approximately estimates the number of materials in that images.
{"title":"Determining the number of endmembers of hyperspectral images using clustering","authors":"José Prades, A. Salazar, G. Safont, L. Vergara","doi":"10.1109/CSCI51800.2020.00306","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00306","url":null,"abstract":"Some applications require knowing how many materials are present in the scene represented by a hyperspectral Image. In a previous paper, we presented an algorithm that estimated the number of materials in the scene using clustering principles. The proposed algorithm obtains a hierarchy of image partitions and selects a partition using a validation Index; the estimated number of materials is set to the number of dusters of the selected partition. In this algorithm, the user must provide the Image and the maximum number of materials that can be estimated (P). In this paper, we have extended our algorithm so that It does not require P as input parameter. The proposed method Iteratively performs the estimation for several increasing values of P and stops the process when a certain condition is met. The results obtained with five hyperspectral Images show that our algorithm approximately estimates the number of materials in that images.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128384033","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00316
L. Deligiannidis
In this work we present an inexpensive, yet accurate, solution of measuring human temperature from a distance. The need for such solution is essential during pandemics. During COVID-19, one of the most common symptoms, for those who develop symptoms, is fever. We believe a tool that measures multiple peoples’ temperature from a safe distance can be valuable. As people enter buildings, airports, hospitals, etc. they can be scanned automatically from a safe distance. The system can alert the authorities for further assessment. Even though such a tool does not prevent the spread of a virus by itself, it can help contain the virus following additional measures such as wearing a face mask, frequent hand washing, and social distancing.
{"title":"Human Temperature Scanning from a Distance","authors":"L. Deligiannidis","doi":"10.1109/CSCI51800.2020.00316","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00316","url":null,"abstract":"In this work we present an inexpensive, yet accurate, solution of measuring human temperature from a distance. The need for such solution is essential during pandemics. During COVID-19, one of the most common symptoms, for those who develop symptoms, is fever. We believe a tool that measures multiple peoples’ temperature from a safe distance can be valuable. As people enter buildings, airports, hospitals, etc. they can be scanned automatically from a safe distance. The system can alert the authorities for further assessment. Even though such a tool does not prevent the spread of a virus by itself, it can help contain the virus following additional measures such as wearing a face mask, frequent hand washing, and social distancing.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128408152","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00043
I. Avramovic, D. Richards
Perpetual gossiping is an all-to-all communication problem on social networks, or any coordinated distributed system in general. In perpetual gossiping, a state of universal reachability is maintained by a continuous sequence of communications between participants. Unlike the well-understood static case, perpetual gossiping is a difficult problem, with some NP-complete classes of solutions. A basic question which remains open is whether an optimal scheme of contiguous calls is guaranteed to exist for a tree. This paper presents a series of theoretical tools directed towards answering the question.
{"title":"Rules for Optimal Perpetual Gossiping Schemes","authors":"I. Avramovic, D. Richards","doi":"10.1109/CSCI51800.2020.00043","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00043","url":null,"abstract":"Perpetual gossiping is an all-to-all communication problem on social networks, or any coordinated distributed system in general. In perpetual gossiping, a state of universal reachability is maintained by a continuous sequence of communications between participants. Unlike the well-understood static case, perpetual gossiping is a difficult problem, with some NP-complete classes of solutions. A basic question which remains open is whether an optimal scheme of contiguous calls is guaranteed to exist for a tree. This paper presents a series of theoretical tools directed towards answering the question.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125857064","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00121
Taishi Nemoto, T. Fujimoto
Since the 2000s, the third artificial intelligence boom has occurred. Research on machine learning and deep learning is progressing, but challenges remain regarding realizing so-called ‘human-like’ general-purpose AI (Artificial Intelligence). In recent years, artificial intelligence research has been linked to cognitive science, and the question of what ‘humanity’ is and how to design ‘humanity’ has been raised as issues. The more robots resemble humans, the more the ‘uncanny valley phenomenon’ increases and the more people feel uncomfortable. Even though technology has advanced and realistic textures can be expressed, robots seem to be just ‘artifacts.’ The point of this study is to determine at what point humans feel ‘human-like’ and how to reproduce ‘human-like’ using a computer. In this study, in order to express human personality and characteristic gestures, we generate an ‘artificial personality’ (AP), and let people find the human touch that a real person possesses through that AP. For example, artificial reproduction of the intelligence of a deceased person is difficult with today's technology. However, AP enables to extract the characteristics of a person's gestures, routines, habits, and facial expressions in his or her lifetime, and to digitally recreate the person's personality based on the accumulation of ‘flesh and blood’ data. This study discusses the two elements and basic mechanisms that are necessary for AP research.
{"title":"Design of Humanity by the Concept of Artificial Personalities","authors":"Taishi Nemoto, T. Fujimoto","doi":"10.1109/CSCI51800.2020.00121","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00121","url":null,"abstract":"Since the 2000s, the third artificial intelligence boom has occurred. Research on machine learning and deep learning is progressing, but challenges remain regarding realizing so-called ‘human-like’ general-purpose AI (Artificial Intelligence). In recent years, artificial intelligence research has been linked to cognitive science, and the question of what ‘humanity’ is and how to design ‘humanity’ has been raised as issues. The more robots resemble humans, the more the ‘uncanny valley phenomenon’ increases and the more people feel uncomfortable. Even though technology has advanced and realistic textures can be expressed, robots seem to be just ‘artifacts.’ The point of this study is to determine at what point humans feel ‘human-like’ and how to reproduce ‘human-like’ using a computer. In this study, in order to express human personality and characteristic gestures, we generate an ‘artificial personality’ (AP), and let people find the human touch that a real person possesses through that AP. For example, artificial reproduction of the intelligence of a deceased person is difficult with today's technology. However, AP enables to extract the characteristics of a person's gestures, routines, habits, and facial expressions in his or her lifetime, and to digitally recreate the person's personality based on the accumulation of ‘flesh and blood’ data. This study discusses the two elements and basic mechanisms that are necessary for AP research.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114522630","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 : 2020-12-01DOI: 10.1109/CSCI51800.2020.00230
Adedolapo Okanlawon, Huichen Yang, Avishek Bose, W. Hsu, Dan Andresen, Mohammed Tanash
We present a machine learning framework and a new test bed for data mining from the Slurm Workload Manager for high-performance computing (HPC) clusters. The focus was to find a method for selecting features to support decisions: helping users decide whether to resubmit failed jobs with boosted CPU and memory allocations or migrate them to a computing cloud. This task was cast as both supervised classification and regression learning, specifically, sequential problem solving suitable for reinforcement learning. Selecting relevant features can improve training accuracy, reduce training time, and produce a more comprehensible model, with an intelligent system that can explain predictions and inferences. We present a supervised learning model trained on a Simple Linux Utility for Resource Management (Slurm) data set of HPC jobs using three different techniques for selecting features: linear regression, lasso, and ridge regression. Our data set represented both HPC jobs that failed and those that succeeded, so our model was reliable, less likely to overfit, and generalizable. Our model achieved an R2 of 95% with 99% accuracy. We identified five predictors for both CPU and memory properties.
{"title":"Feature Selection for Learning to Predict Outcomes of Compute Cluster Jobs with Application to Decision Support","authors":"Adedolapo Okanlawon, Huichen Yang, Avishek Bose, W. Hsu, Dan Andresen, Mohammed Tanash","doi":"10.1109/CSCI51800.2020.00230","DOIUrl":"https://doi.org/10.1109/CSCI51800.2020.00230","url":null,"abstract":"We present a machine learning framework and a new test bed for data mining from the Slurm Workload Manager for high-performance computing (HPC) clusters. The focus was to find a method for selecting features to support decisions: helping users decide whether to resubmit failed jobs with boosted CPU and memory allocations or migrate them to a computing cloud. This task was cast as both supervised classification and regression learning, specifically, sequential problem solving suitable for reinforcement learning. Selecting relevant features can improve training accuracy, reduce training time, and produce a more comprehensible model, with an intelligent system that can explain predictions and inferences. We present a supervised learning model trained on a Simple Linux Utility for Resource Management (Slurm) data set of HPC jobs using three different techniques for selecting features: linear regression, lasso, and ridge regression. Our data set represented both HPC jobs that failed and those that succeeded, so our model was reliable, less likely to overfit, and generalizable. Our model achieved an R2 of 95% with 99% accuracy. We identified five predictors for both CPU and memory properties.","PeriodicalId":336929,"journal":{"name":"2020 International Conference on Computational Science and Computational Intelligence (CSCI)","volume":"67 1-2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123587156","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：