10000维线性系统的数值验证速度快10000倍

ARCH@ADHS Pub Date : 2018-09-17 DOI:10.29007/gv5q
Stanley Bak
{"title":"10000维线性系统的数值验证速度快10000倍","authors":"Stanley Bak","doi":"10.29007/gv5q","DOIUrl":null,"url":null,"abstract":"Tool Presentation: We evaluate an improved reachability algorithm for linear (and affine) systems implemented in the continuous branch of the Hylaa tool. While Hylaa’s earlier approach required n simulations to verify an n-dimensional system, the new method takes advantage of additional problem structure to produce the same verification result in significantly less time. If the initial states can be defined in i dimensions, and the output variables related to the property being checked are o-dimensional, the new approach needs only min(i, o) simulations to verify the system, or produce a counter-example. In addition to reducing the number of simulations, a second improvement speeds up individual simulations when the dynamics is sparse by using Krylov subspace methods. At ARCH 2017, we used the original approach to verify nine large linear benchmarks taken from model order reduction. Here, we run the new algorithm on the same set of benchmarks, and get an identical verification result in a fraction of the time. None of the benchmarks need more than tens of seconds to complete. The largest system with 10922 dimensions, which took over 24 hours using last year’s method, is verified in 3.4 seconds.","PeriodicalId":236469,"journal":{"name":"ARCH@ADHS","volume":"25 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2018-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Numerical Verification of 10000-dimensional Linear Systems 10000x Faster\",\"authors\":\"Stanley Bak\",\"doi\":\"10.29007/gv5q\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Tool Presentation: We evaluate an improved reachability algorithm for linear (and affine) systems implemented in the continuous branch of the Hylaa tool. While Hylaa’s earlier approach required n simulations to verify an n-dimensional system, the new method takes advantage of additional problem structure to produce the same verification result in significantly less time. If the initial states can be defined in i dimensions, and the output variables related to the property being checked are o-dimensional, the new approach needs only min(i, o) simulations to verify the system, or produce a counter-example. In addition to reducing the number of simulations, a second improvement speeds up individual simulations when the dynamics is sparse by using Krylov subspace methods. At ARCH 2017, we used the original approach to verify nine large linear benchmarks taken from model order reduction. Here, we run the new algorithm on the same set of benchmarks, and get an identical verification result in a fraction of the time. None of the benchmarks need more than tens of seconds to complete. The largest system with 10922 dimensions, which took over 24 hours using last year’s method, is verified in 3.4 seconds.\",\"PeriodicalId\":236469,\"journal\":{\"name\":\"ARCH@ADHS\",\"volume\":\"25 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2018-09-17\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"ARCH@ADHS\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.29007/gv5q\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"ARCH@ADHS","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.29007/gv5q","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 1

摘要

工具介绍:我们评估了在Hylaa工具的连续分支中实现的线性(和仿射)系统的改进可达性算法。Hylaa之前的方法需要n次模拟来验证一个n维系统,而新方法利用了额外的问题结构,在更短的时间内产生相同的验证结果。如果初始状态可以在i维中定义,并且与被检查的属性相关的输出变量是o维的,则新方法只需要最少(i, o)个模拟来验证系统或生成反例。除了减少模拟次数外,第二个改进还通过使用Krylov子空间方法加快了动态稀疏时的单个模拟。在ARCH 2017上,我们使用原始方法验证了从模型降阶中获得的9个大型线性基准。在这里,我们在相同的基准测试集上运行新算法,并在很短的时间内获得相同的验证结果。没有一个基准测试需要超过几十秒才能完成。最大的10922个维度的系统,使用去年的方法需要24小时以上,在3.4秒内得到验证。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Numerical Verification of 10000-dimensional Linear Systems 10000x Faster
Tool Presentation: We evaluate an improved reachability algorithm for linear (and affine) systems implemented in the continuous branch of the Hylaa tool. While Hylaa’s earlier approach required n simulations to verify an n-dimensional system, the new method takes advantage of additional problem structure to produce the same verification result in significantly less time. If the initial states can be defined in i dimensions, and the output variables related to the property being checked are o-dimensional, the new approach needs only min(i, o) simulations to verify the system, or produce a counter-example. In addition to reducing the number of simulations, a second improvement speeds up individual simulations when the dynamics is sparse by using Krylov subspace methods. At ARCH 2017, we used the original approach to verify nine large linear benchmarks taken from model order reduction. Here, we run the new algorithm on the same set of benchmarks, and get an identical verification result in a fraction of the time. None of the benchmarks need more than tens of seconds to complete. The largest system with 10922 dimensions, which took over 24 hours using last year’s method, is verified in 3.4 seconds.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信