Domain experts in different areas have a large number of options for approaching their specific data analysis problem. In exploration of large data sets on HPC systems, choosing which method to use, or how to tune the parameters of an algorithm to achieve good results are challenging tasks for data analysts themselves. In this paper, we propose a recommendation module for a distributed machine learning environment aiming at helping the end-users to obtain optimized results for their data analysis / machine learning problem.
{"title":"Ontology-Based Recommender for Distributed Machine Learning Environment","authors":"Daniel Pop, Caius Bogdanescu","doi":"10.1109/SYNASC.2013.76","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.76","url":null,"abstract":"Domain experts in different areas have a large number of options for approaching their specific data analysis problem. In exploration of large data sets on HPC systems, choosing which method to use, or how to tune the parameters of an algorithm to achieve good results are challenging tasks for data analysts themselves. In this paper, we propose a recommendation module for a distributed machine learning environment aiming at helping the end-users to obtain optimized results for their data analysis / machine learning problem.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126044013","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}
Online auction sites have to deal with a enormous amount of product listings, of which a fraction is fraudulent. Although small in proportion, fraudulent listings are costly for site operators, buyers and legitimate sellers. Fraud prediction in this scenario can benefit significantly from machine learning techniques, although interpretability of model predictions is a concern. In this work we extend an unsupervised learning technique -- Self-Organizing Maps -- to use labeled data for binary classification under a constraint on the proportion of false positives. The resulting technique was applied to the prediction of non-delivery fraud, achieving good results while being easier to interpret.
{"title":"Using Self-Organizing Maps for Fraud Prediction at Online Auction Sites","authors":"Vinicius Almendra, D. Enachescu","doi":"10.1109/SYNASC.2013.44","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.44","url":null,"abstract":"Online auction sites have to deal with a enormous amount of product listings, of which a fraction is fraudulent. Although small in proportion, fraudulent listings are costly for site operators, buyers and legitimate sellers. Fraud prediction in this scenario can benefit significantly from machine learning techniques, although interpretability of model predictions is a concern. In this work we extend an unsupervised learning technique -- Self-Organizing Maps -- to use labeled data for binary classification under a constraint on the proportion of false positives. The resulting technique was applied to the prediction of non-delivery fraud, achieving good results while being easier to interpret.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"03 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131320003","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}
Marius Barat, Dumitru-Bogdan Prelipcean, Dragos Gavrilut
In the increasing number of online threats and shape-shifting malware, the use of machine learning techniques has a good impact. To keep the efficiency of these techniques, the training and adaptation schedule must be constant. In this paper we study the behaviour of an automatic updating perceptron, with variable training frequency and using as input samples with increasing freshness. Other variable parameters are the features set and training set dimensions. The collected samples, clean and malicious are from the last year. We conclude with the observed optimal parameters which can be used to obtain a good proactivity.
{"title":"An Automatic Updating Perceptron-Based System for Malware Detection","authors":"Marius Barat, Dumitru-Bogdan Prelipcean, Dragos Gavrilut","doi":"10.1109/SYNASC.2013.47","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.47","url":null,"abstract":"In the increasing number of online threats and shape-shifting malware, the use of machine learning techniques has a good impact. To keep the efficiency of these techniques, the training and adaptation schedule must be constant. In this paper we study the behaviour of an automatic updating perceptron, with variable training frequency and using as input samples with increasing freshness. Other variable parameters are the features set and training set dimensions. The collected samples, clean and malicious are from the last year. We conclude with the observed optimal parameters which can be used to obtain a good proactivity.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131640424","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}
We continue a Coxeter spectral study of finite posets and edge-bipartite graphs (a class of signed graphs in the sense of Harary and Zaslavsky). Here we are interested in two problems. First: whether the incidence matrices CI and CJ of two connected positive posets I and J are Z-congruent if and only if the Coxeter spectra of I and J coincide. Second: the problem if any square integer matrix A E Mn(Z) is Z-congruent with its transpose Atr. We show that these problems can be effectively solved using the right action * : Mn(Z) × Gl(n, Z)D → Mn(Z), A → A * B := Btr · A · B, of the isotropy group Gl(n, Z)D of a simply laced Dynkin diagram D E {An, Dn, E6, E7, E8}. We present an efficient algorithm for computing the isotropy group Gl(n, Z)D. In particular, we show that symbolic and numerical computer calculations in Python and Cython allow us to present a complete description of the isotropy group Gl(n, Z)D with |D| ≤ 10. Furthermore, we discuss optimisation techniques that are important from the calculation efficiency point of view.
{"title":"Efficient Computation of the Isotropy Group of a Finite Graph: A Combinatorial Approach","authors":"Marcin Gąsiorek","doi":"10.1109/SYNASC.2013.21","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.21","url":null,"abstract":"We continue a Coxeter spectral study of finite posets and edge-bipartite graphs (a class of signed graphs in the sense of Harary and Zaslavsky). Here we are interested in two problems. First: whether the incidence matrices CI and CJ of two connected positive posets I and J are Z-congruent if and only if the Coxeter spectra of I and J coincide. Second: the problem if any square integer matrix A E Mn(Z) is Z-congruent with its transpose Atr. We show that these problems can be effectively solved using the right action * : M<sub>n</sub>(Z) × Gl(n, Z)<sub>D</sub> → M<sub>n</sub>(Z), A → A * B := Btr · A · B, of the isotropy group Gl(n, Z)D of a simply laced Dynkin diagram D E {A<sub>n</sub>, D<sub>n</sub>, E<sub>6</sub>, E<sub>7</sub>, E<sub>8</sub>}. We present an efficient algorithm for computing the isotropy group Gl(n, Z)D. In particular, we show that symbolic and numerical computer calculations in Python and Cython allow us to present a complete description of the isotropy group Gl(n, Z)D with |D| ≤ 10. Furthermore, we discuss optimisation techniques that are important from the calculation efficiency point of view.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129662193","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}
We offer an experimental proof that the application of compression to data files can be used as a evaluation technique for minability of the data. This is based on the fact that the presence of patterns embedded in data has an influence of compressibility.
{"title":"Minability through Compression","authors":"D. Simovici","doi":"10.1109/SYNASC.2013.11","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.11","url":null,"abstract":"We offer an experimental proof that the application of compression to data files can be used as a evaluation technique for minability of the data. This is based on the fact that the presence of patterns embedded in data has an influence of compressibility.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130040769","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}
We introduce a Matlab mprec arbitrary precision library with applications to numerical analysis. For maximum efficiency arithmetic operators and algebraic functions are implemented in the mpreal class. The examples are chosen to reflect the diversity of types of problems for which multiple precision can play a useful role.
{"title":"Software Implementation of Numerical Algorithms in Arbitrary Precision","authors":"Z. Krougly, D. J. Jeffrey, Dina Tsarapkina","doi":"10.1109/SYNASC.2013.25","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.25","url":null,"abstract":"We introduce a Matlab mprec arbitrary precision library with applications to numerical analysis. For maximum efficiency arithmetic operators and algebraic functions are implemented in the mpreal class. The examples are chosen to reflect the diversity of types of problems for which multiple precision can play a useful role.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128749546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper we discuss cosmological N-Body/SPH simulations on parallel computing systems with distributed memory using GADGET-2. GADGET-2 (GAlaxies with Dark matter and Gas intEracT) is a novel cosmological simulation code, written in C++ and publicly available, developed by Volker Springel at the Max-Planck-Institute for Astrophysics in Munchen, Germany as an improved version of GADGET. It is a massively parallel code that uses an explicit communication model implemented with the standardized MPI communication interface. Our contribution consists of a scalability and performance analysis, done on different types of simulations ranging from thousands to hundreds of millions of particles. The size of the simulation is limited by the amount of available physical memory (RAM) when provided with sufficient computing time. Based on our study, we propose an inferior and superior limit to astrophysical simulations than can be run on a specific parallel computing system while maintaining a specific efficiency. We also outline a trend stating the computational resources that are necessary depending on the size of the simulation.
{"title":"N-Body Simulations with GADGET-2","authors":"V. Spiridon, E. Slusanschi","doi":"10.1109/SYNASC.2013.75","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.75","url":null,"abstract":"In this paper we discuss cosmological N-Body/SPH simulations on parallel computing systems with distributed memory using GADGET-2. GADGET-2 (GAlaxies with Dark matter and Gas intEracT) is a novel cosmological simulation code, written in C++ and publicly available, developed by Volker Springel at the Max-Planck-Institute for Astrophysics in Munchen, Germany as an improved version of GADGET. It is a massively parallel code that uses an explicit communication model implemented with the standardized MPI communication interface. Our contribution consists of a scalability and performance analysis, done on different types of simulations ranging from thousands to hundreds of millions of particles. The size of the simulation is limited by the amount of available physical memory (RAM) when provided with sufficient computing time. Based on our study, we propose an inferior and superior limit to astrophysical simulations than can be run on a specific parallel computing system while maintaining a specific efficiency. We also outline a trend stating the computational resources that are necessary depending on the size of the simulation.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"43 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121411682","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}
We continue the Coxeter spectral study of finite connected loop-free edge-bipartite graphs Δ, with n ≥ 2 vertices (a class of signed graphs), started in [SIAM J. Discrete Math., 27(2013), 827-854] by means of the complex Coxeter spectrum speccΔ ⊆ ℂ. Here, we discuss Coxeter spectral analysis problems of non-negative edge-bipartite graphs of corank s ≤ n-1, which means that the symmetric Gram matrix GΔ ∈ Mn(ℤ) is positive semi-definite of rank n-s ≤ n. In particular, we study in details the loop-free edge-bipartite graphs of corank s = n - 1. We present algorithms that generate all such edge-bipartite graphs of a given size and, using symbolic and numerical computer calculations in Python, and we obtain their complete classification in relation with Diophantine geometry problems. We also construct algorithms that allow us to classify all connected loop-free non-negative edge-bipartite graphs Δ, with a fixed number n ≥ 2 of vertices, by means of their Coxeter spectra speccΔ.
我们继续有限连通无环边二部图Δ的Coxeter谱研究，n≥2个顶点(一类有符号图)，开始于[SIAM J.离散数学]。， 27(2013)， 827-854]利用复Coxeter谱speccΔ本文讨论了corank s≤n-1的非负边二部图的Coxeter谱分析问题，这意味着对称Gram矩阵GΔ∈Mn(n)是秩n-s≤n的正半定的。特别地，我们详细研究了corank s = n-1的无环边二部图。我们提出了生成给定大小的所有这些边二部图的算法，并使用Python中的符号和数值计算机计算，我们获得了与丢芬图几何问题相关的完整分类。我们还构建了一种算法，允许我们对所有连接的无环非负边二部图Δ进行分类，这些图具有固定数目n≥2个顶点，通过它们的Coxeter谱speccΔ。
{"title":"On Computing Non-negative Loop-Free Edge-Bipartite Graphs","authors":"Grzegorz Marczak, D. Simson, Katarzyna Zając","doi":"10.1109/SYNASC.2013.16","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.16","url":null,"abstract":"We continue the Coxeter spectral study of finite connected loop-free edge-bipartite graphs Δ, with n ≥ 2 vertices (a class of signed graphs), started in [SIAM J. Discrete Math., 27(2013), 827-854] by means of the complex Coxeter spectrum specc<sub>Δ</sub> ⊆ ℂ. Here, we discuss Coxeter spectral analysis problems of non-negative edge-bipartite graphs of corank s ≤ n-1, which means that the symmetric Gram matrix G<sub>Δ</sub> ∈ M<sub>n</sub>(ℤ) is positive semi-definite of rank n-s ≤ n. In particular, we study in details the loop-free edge-bipartite graphs of corank s = n - 1. We present algorithms that generate all such edge-bipartite graphs of a given size and, using symbolic and numerical computer calculations in Python, and we obtain their complete classification in relation with Diophantine geometry problems. We also construct algorithms that allow us to classify all connected loop-free non-negative edge-bipartite graphs Δ, with a fixed number n ≥ 2 of vertices, by means of their Coxeter spectra specc<sub>Δ</sub>.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127029246","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}
It has long been known that cylindrical algebraic decompositions (CADs) can in theory be used for robot motion planning. However, in practice even the simplest examples can be too complicated to tackle. We consider in detail a ``Piano Mover's Problem'' which considers moving an infinitesimally thin piano (or ladder) through a right-angled corridor. Producing a CAD for the original formulation of this problem is still infeasible after 25 years of improvements in both CAD theory and computer hardware. We review some alternative formulations in the literature which use differing levels of geometric analysis before input to a CAD algorithm. Simpler formulations allow CAD to easily address the question of the existence of a path. We provide a new formulation for which both a CAD can be constructed and from which an actual path could be determined if one exists, and analyse the CADs produced using this approach for variations of the problem. This emphasises the importance of the precise formulation of such problems for CAD. We analyse the formulations and their CADs considering a variety of heuristics and general criteria, leading to conclusions about tackling other problems of this form.
{"title":"A \"Piano Movers\" Problem Reformulated","authors":"D. Wilson, J. Davenport, M. England, R. Bradford","doi":"10.1109/SYNASC.2013.14","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.14","url":null,"abstract":"It has long been known that cylindrical algebraic decompositions (CADs) can in theory be used for robot motion planning. However, in practice even the simplest examples can be too complicated to tackle. We consider in detail a ``Piano Mover's Problem'' which considers moving an infinitesimally thin piano (or ladder) through a right-angled corridor. Producing a CAD for the original formulation of this problem is still infeasible after 25 years of improvements in both CAD theory and computer hardware. We review some alternative formulations in the literature which use differing levels of geometric analysis before input to a CAD algorithm. Simpler formulations allow CAD to easily address the question of the existence of a path. We provide a new formulation for which both a CAD can be constructed and from which an actual path could be determined if one exists, and analyse the CADs produced using this approach for variations of the problem. This emphasises the importance of the precise formulation of such problems for CAD. We analyse the formulations and their CADs considering a variety of heuristics and general criteria, leading to conclusions about tackling other problems of this form.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115948443","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}
A simple and expressive formalism called TiMo is presented as a simplified version of timed distributed pi-calculus. TiMo aims to bridge the gap between the existing theoretical approach of process calculi and forthcoming realistic languages for multi-agent systems.
{"title":"TiMo: Timed Mobility in Distributed Systems","authors":"Gabriel Ciobanu","doi":"10.1109/SYNASC.2013.8","DOIUrl":"https://doi.org/10.1109/SYNASC.2013.8","url":null,"abstract":"A simple and expressive formalism called TiMo is presented as a simplified version of timed distributed pi-calculus. TiMo aims to bridge the gap between the existing theoretical approach of process calculi and forthcoming realistic languages for multi-agent systems.","PeriodicalId":293085,"journal":{"name":"2013 15th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133548036","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：