First EuroNGI Workshop New Trends in Modelling, Quantitative Methods and Measurements.pdfVIP

下载本文档

2
0
约1.7万字
约 10页
2017-04-11 发布于江苏
举报
版权申诉

First EuroNGI Workshop New Trends in Modelling, Quantitative Methods and Measurements.pdf

1、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。。
2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
3、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
4、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
5、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
6、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
7、VIP文档为合作方或网友上传，每下载1次，网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档

First EuroNGI Workshop New Trends in Modelling, Quantitative Methods and Measurements

First EuroNGI Workshop: New Trends in Modelling, Quantitative Methods and Measurements The WZ factorization and the use of the BLAS library in the numerical solution of Markov chains BEATA BYLINA a a Department of Computer Science, Marie Curie-Sk?odowska University, Pl. M. Curie-Sk?odowskiej 1, 20-031 Lublin, Poland, e-mail: beatas@golem.umcs.lublin.pl 1. Introduction A system modelled with a Markov process assumes exactly one state (from the whole set) in each time moment. The system evolution is described with the probabilities of transitions from one state to another. In such a model we are usually interested in probabilities of particular states in a given time moment . In this paper we are interested in contium-time Markov chains (CTMC). We have got a transitoin rate matrix (or an infinitesimal generator matrix) and we get a linear equation system with some constraints: (1) (where is a stationary probability vector and ) which we must solve. The matrix is a square matrix (usually a huge one – but we consider only finite state spaces), of the rank ! (in interesting for us cases), with the dominant diagonal. It is singular and sparse, so we need appropriate approaches to the system (1). There are indirect methods that are widely used – like iterative or projection methods. However, sometimes we need a direct method (when we need the most accurate results, for example). The most popular of the direct methods are the LU factorization method and its relatives. In this article we want to present another direct method for solving (1), namely the WZ factorization method. We show how to speed up (without changing accuracy) the computations with the use of vector and parallel algorithms of the WZ factorization. 2. The WZ Factorization In this section we describe shortly the WZ factorization and the method for solving a linear system #%$ )(+*,-* #/ 7 $8.9032 (2) with the WZ factorizat