Abstract Behavioural biometrics provides an extra layer of security for user authentication mechanisms. Among behavioural biometrics, mouse dynamics provides a non-intrusive layer of security. In this paper we propose a novel convolutional neural network for extracting the features from the time series of users’ mouse movements. The effect of two preprocessing methods on the performance of the proposed architecture were evaluated. Different training types of the model, namely transfer learning and training from scratch, were investigated. Results for both authentication and identification systems are reported. The Balabit public data set was used for performance evaluation, however for transfer learning we used the DFL data set. Comprehensive experimental evaluations suggest that our model performed better than other deep learning models. In addition, transfer learning contributed to the better performance of both identification and authentication systems.
{"title":"Mouse dynamics based user recognition using deep learning","authors":"M. Antal, Norbert Fejér","doi":"10.2478/ausi-2020-0003","DOIUrl":"https://doi.org/10.2478/ausi-2020-0003","url":null,"abstract":"Abstract Behavioural biometrics provides an extra layer of security for user authentication mechanisms. Among behavioural biometrics, mouse dynamics provides a non-intrusive layer of security. In this paper we propose a novel convolutional neural network for extracting the features from the time series of users’ mouse movements. The effect of two preprocessing methods on the performance of the proposed architecture were evaluated. Different training types of the model, namely transfer learning and training from scratch, were investigated. Results for both authentication and identification systems are reported. The Balabit public data set was used for performance evaluation, however for transfer learning we used the DFL data set. Comprehensive experimental evaluations suggest that our model performed better than other deep learning models. In addition, transfer learning contributed to the better performance of both identification and authentication systems.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"119 1","pages":"39 - 50"},"PeriodicalIF":0.3,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73694330","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 The paper focuses on providing an insight into the current state of computational modeling regarding reactive magnetron sputtering systems. A detailed compilation of developed models is gathered and grouped into categories based on the phenomena being modeled. The survey covers models developed for the analysis of magnetron discharges, particle-surface interactions at the target and the substrate, as well as macroscopic models. Corresponding software packages available online are also presented. After gaining the necessary insight into the current state of research, a list of the most challenging tasks is given, comparing diffierent approaches, that have been used to combat the encountered difficulties. The challenges associated with modeling tasks range from analytical complexity, mathematical know-how used for model approximation and reduction, as well as optimization between computational load and result accuracy. As a conclusion, the future challenges are compiled into a list and a probable direction in modeling is given, that is likely to be further pursued.
{"title":"Modeling reactive magnetron sputtering: a survey of different modeling approaches","authors":"R. Madarász, A. Kelemen, P. Kádár","doi":"10.2478/ausi-2020-0008","DOIUrl":"https://doi.org/10.2478/ausi-2020-0008","url":null,"abstract":"Abstract The paper focuses on providing an insight into the current state of computational modeling regarding reactive magnetron sputtering systems. A detailed compilation of developed models is gathered and grouped into categories based on the phenomena being modeled. The survey covers models developed for the analysis of magnetron discharges, particle-surface interactions at the target and the substrate, as well as macroscopic models. Corresponding software packages available online are also presented. After gaining the necessary insight into the current state of research, a list of the most challenging tasks is given, comparing diffierent approaches, that have been used to combat the encountered difficulties. The challenges associated with modeling tasks range from analytical complexity, mathematical know-how used for model approximation and reduction, as well as optimization between computational load and result accuracy. As a conclusion, the future challenges are compiled into a list and a probable direction in modeling is given, that is likely to be further pursued.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"3 1","pages":"112 - 136"},"PeriodicalIF":0.3,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80604322","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 Closing a deal in a business to business environment implies a series of orchestrated actions that the sales representatives are taking to take a prospective buyer from first contact to a closed sale. The actions, such as meetings, emails, phone calls happen in succession and in different points in time relative to the first interaction. Time-series are ordered sequences of discrete-time data. In this work, we are examining the relationship between the actions as time series and the final win outcome for each deal. To assess whether the behavior of the salespeople have a direct influence on the final outcome of the current deal, we used histogram analysis, dynamic time warping and string edit distance on a real-world Customer Relationship Management System data set. The results are discussed and included in this paper.
{"title":"Opportunity activity sequence investigations in B2B CRM systems","authors":"Doru Rotovei","doi":"10.2478/ausi-2020-0005","DOIUrl":"https://doi.org/10.2478/ausi-2020-0005","url":null,"abstract":"Abstract Closing a deal in a business to business environment implies a series of orchestrated actions that the sales representatives are taking to take a prospective buyer from first contact to a closed sale. The actions, such as meetings, emails, phone calls happen in succession and in different points in time relative to the first interaction. Time-series are ordered sequences of discrete-time data. In this work, we are examining the relationship between the actions as time series and the final win outcome for each deal. To assess whether the behavior of the salespeople have a direct influence on the final outcome of the current deal, we used histogram analysis, dynamic time warping and string edit distance on a real-world Customer Relationship Management System data set. The results are discussed and included in this paper.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"88 1","pages":"70 - 83"},"PeriodicalIF":0.3,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73202091","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 One of the most studied aspect of complex graphs is identifying the most influential nodes. There are some local metrics like degree centrality, which is cost-effiective and easy to calculate, although using global metrics like betweenness centrality or closeness centrality can identify influential nodes more accurately, however calculating these values can be costly and each measure has it’s own limitations and disadvantages. There is an ever-growing interest in calculating such metrics in time-varying graphs (TVGs), since modern complex networks can be best modelled with such graphs. In this paper we are investigating the effectiveness of a new centrality measure called efficiency centrality in TVGs. To evaluate the performance of the algorithm Independent Cascade Model is used to simulate infection spreading in four real networks. To simulate the changes in the network we are deleting and adding nodes based on their degree centrality. We are investigating the Time-Constrained Coverage and the magnitude of propagation resulted by the use of the algorithm.
{"title":"Efficiency centrality in time-varying graphs","authors":"Péter Marjai, A. Kiss","doi":"10.2478/ausi-2020-0001","DOIUrl":"https://doi.org/10.2478/ausi-2020-0001","url":null,"abstract":"Abstract One of the most studied aspect of complex graphs is identifying the most influential nodes. There are some local metrics like degree centrality, which is cost-effiective and easy to calculate, although using global metrics like betweenness centrality or closeness centrality can identify influential nodes more accurately, however calculating these values can be costly and each measure has it’s own limitations and disadvantages. There is an ever-growing interest in calculating such metrics in time-varying graphs (TVGs), since modern complex networks can be best modelled with such graphs. In this paper we are investigating the effectiveness of a new centrality measure called efficiency centrality in TVGs. To evaluate the performance of the algorithm Independent Cascade Model is used to simulate infection spreading in four real networks. To simulate the changes in the network we are deleting and adding nodes based on their degree centrality. We are investigating the Time-Constrained Coverage and the magnitude of propagation resulted by the use of the algorithm.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"56 1","pages":"21 - 5"},"PeriodicalIF":0.3,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80476582","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 Given a graph G = (V, E), with respect to a vertex partition 𝒫 we associate a matrix called 𝒫-matrix and define the 𝒫-energy, E𝒫 (G) as the sum of 𝒫-eigenvalues of 𝒫-matrix of G. Apart from studying some properties of 𝒫-matrix, its eigenvalues and obtaining bounds of 𝒫-energy, we explore the robust(shear) 𝒫-energy which is the maximum(minimum) value of 𝒫-energy for some families of graphs. Further, we derive explicit formulas for E𝒫 (G) of few classes of graphs with different vertex partitions.
摘要:给定图G = (V, E)，对一个顶点划分，我们关联一个矩阵𝒫-matrix，并定义𝒫-energy, E (G)为G的𝒫-matrix的𝒫-eigenvalues的和。除了研究𝒫-matrix的一些性质，它的特征值和𝒫-energy的界外，我们还探讨了图族中𝒫-energy的最大(最小)值的鲁棒性(剪切性)𝒫-energy。进一步，我们导出了几种不同顶点划分的图的显式公式E - p (G)。
{"title":"𝒫-energy of graphs","authors":"Prajakta Bharat Joshi, Mayamma Joseph","doi":"10.2478/ausi-2020-0009","DOIUrl":"https://doi.org/10.2478/ausi-2020-0009","url":null,"abstract":"Abstract Given a graph G = (V, E), with respect to a vertex partition 𝒫 we associate a matrix called 𝒫-matrix and define the 𝒫-energy, E𝒫 (G) as the sum of 𝒫-eigenvalues of 𝒫-matrix of G. Apart from studying some properties of 𝒫-matrix, its eigenvalues and obtaining bounds of 𝒫-energy, we explore the robust(shear) 𝒫-energy which is the maximum(minimum) value of 𝒫-energy for some families of graphs. Further, we derive explicit formulas for E𝒫 (G) of few classes of graphs with different vertex partitions.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"2014 1","pages":"137 - 157"},"PeriodicalIF":0.3,"publicationDate":"2020-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83872182","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 Time-series are ordered sequences of discrete-time data. Due to their temporal dimension, anomaly detection techniques used in time-series have to take into consideration time correlations and other time-related particularities. Generally, in order to evaluate the quality of an anomaly detection technique, the confusion matrix and its derived metrics such as precision and recall are used. These metrics, however, do not take this temporal dimension into consideration. In this paper, we propose three metrics that can be used to evaluate the quality of a classification, while accounting for the temporal dimension found in time-series data.
{"title":"Evaluation metrics for anomaly detection algorithms in time-series","authors":"György Kovács, G. Sebestyen, A. Hangan","doi":"10.2478/ausi-2019-0008","DOIUrl":"https://doi.org/10.2478/ausi-2019-0008","url":null,"abstract":"Abstract Time-series are ordered sequences of discrete-time data. Due to their temporal dimension, anomaly detection techniques used in time-series have to take into consideration time correlations and other time-related particularities. Generally, in order to evaluate the quality of an anomaly detection technique, the confusion matrix and its derived metrics such as precision and recall are used. These metrics, however, do not take this temporal dimension into consideration. In this paper, we propose three metrics that can be used to evaluate the quality of a classification, while accounting for the temporal dimension found in time-series data.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"10 1","pages":"113 - 130"},"PeriodicalIF":0.3,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87953255","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 Graph coloring can be considered as a random experiment with the color of a randomly selected vertex as the random variable. In this paper, we consider the L(2, 1)-coloring of G as the random experiment and we discuss the concept of two fundamental statistical parameters – mean and variance – with respect to the L(2, 1)-coloring of certain fundamental graph classes.
{"title":"On some L(2, 1)-coloring parameters of certain graph classes","authors":"G. Anjali, N. Sudev","doi":"10.2478/ausi-2019-0013","DOIUrl":"https://doi.org/10.2478/ausi-2019-0013","url":null,"abstract":"Abstract Graph coloring can be considered as a random experiment with the color of a randomly selected vertex as the random variable. In this paper, we consider the L(2, 1)-coloring of G as the random experiment and we discuss the concept of two fundamental statistical parameters – mean and variance – with respect to the L(2, 1)-coloring of certain fundamental graph classes.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"32 1","pages":"184 - 205"},"PeriodicalIF":0.3,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90822176","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 In an earlier work [6] the concept of splitting partition of a graph was introduced in connection with the maximum clique problem. A splitting partition of a graph can be used to replace the graph by two smaller graphs in the course of a clique search algorithm. In other words splitting partitions can serve as a branching rule in an algorithm to compute the clique number of a given graph. In the paper we revisit this branching idea. We will describe a technique to construct not necessary optimal splitting partitions. The given graph can be viewed as a metric space and the geometry of this space plays a guiding role. In order to assess the performance of the procedure we carried out numerical experiments.
{"title":"Metric space method for constructing splitting partitions of graphs","authors":"S. Szabó","doi":"10.2478/ausi-2019-0009","DOIUrl":"https://doi.org/10.2478/ausi-2019-0009","url":null,"abstract":"Abstract In an earlier work [6] the concept of splitting partition of a graph was introduced in connection with the maximum clique problem. A splitting partition of a graph can be used to replace the graph by two smaller graphs in the course of a clique search algorithm. In other words splitting partitions can serve as a branching rule in an algorithm to compute the clique number of a given graph. In the paper we revisit this branching idea. We will describe a technique to construct not necessary optimal splitting partitions. The given graph can be viewed as a metric space and the geometry of this space plays a guiding role. In order to assess the performance of the procedure we carried out numerical experiments.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"4 1","pages":"131 - 141"},"PeriodicalIF":0.3,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90921612","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 The existence of edges is a huge challenge with regards to determining lucky k-polynomials of simple connected graphs in general. In this paper the lucky 3-polynomials of path and cycle graphs of order, 3 ≤ n ≤ 8 are presented as the basis for the heuristic method to determine the lucky k-polynomials for k-colorable graphs. The difficulty of adjacency with graphs is illustrated through these elementary graph structures. The results are also illustratively compared with the results for null graphs (edgeless graphs). The paper could serve as a basis for finding recurrence results through innovative methodology.
{"title":"Heuristic method to determine lucky k-polynomials for k-colorable graphs","authors":"J. Kok","doi":"10.2478/ausi-2019-0014","DOIUrl":"https://doi.org/10.2478/ausi-2019-0014","url":null,"abstract":"Abstract The existence of edges is a huge challenge with regards to determining lucky k-polynomials of simple connected graphs in general. In this paper the lucky 3-polynomials of path and cycle graphs of order, 3 ≤ n ≤ 8 are presented as the basis for the heuristic method to determine the lucky k-polynomials for k-colorable graphs. The difficulty of adjacency with graphs is illustrated through these elementary graph structures. The results are also illustratively compared with the results for null graphs (edgeless graphs). The paper could serve as a basis for finding recurrence results through innovative methodology.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"19 1","pages":"206 - 214"},"PeriodicalIF":0.3,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74018334","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 A common assumption about neural networks is that they can learn an appropriate internal representation on their own, see e.g. end-to-end learning. In this work we challenge this assumption. We consider two simple tasks and show that the state-of-the-art training algorithm fails, although the model itself is able to represent an appropriate solution. We will demonstrate that encouraging an appropriate internal representation allows the same model to solve these tasks. While we do not claim that it is impossible to solve these tasks by other means (such as neural networks with more layers), our results illustrate that integration of domain knowledge in form of a desired internal representation may improve the generalization ability of neural networks.
{"title":"Encouraging an appropriate representation simplifies training of neural networks","authors":"Krisztián Búza","doi":"10.2478/ausi-2020-0007","DOIUrl":"https://doi.org/10.2478/ausi-2020-0007","url":null,"abstract":"Abstract A common assumption about neural networks is that they can learn an appropriate internal representation on their own, see e.g. end-to-end learning. In this work we challenge this assumption. We consider two simple tasks and show that the state-of-the-art training algorithm fails, although the model itself is able to represent an appropriate solution. We will demonstrate that encouraging an appropriate internal representation allows the same model to solve these tasks. While we do not claim that it is impossible to solve these tasks by other means (such as neural networks with more layers), our results illustrate that integration of domain knowledge in form of a desired internal representation may improve the generalization ability of neural networks.","PeriodicalId":41480,"journal":{"name":"Acta Universitatis Sapientiae Informatica","volume":"23 1","pages":"102 - 111"},"PeriodicalIF":0.3,"publicationDate":"2019-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82168806","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：