\uff2f\uff32\u5b66\u4f1a\u4e2d\u90e8\u652f\u90e8\u7814\u7a76\u4f1a\u3092\u4ee5\u4e0b\u306e\u3088\u3046\u306b\u958b\u50ac\u3044\u305f\u3057\u307e\u3059\uff0e \u7686\u3055\u307e\u306e\u3054\u53c2\u52a0\u3092\u304a\u5f85\u3061\u3057\u3066\u304a\u308a\u307e\u3059\uff0e<\/p>\n
\u30102016\/05\/19\u66f4\u65b0\uff1a\u8b1b\u6f14\u6982\u8981\u3092\u63b2\u8f09\u3011
\n—<\/p>\n
\u65e5\u6642: 6\u670822\u65e5(\u6c34) 17:00-19:00<\/p>\n
\u5834\u6240: \u540d\u53e4\u5c4b\u5927\u5b66 \u6771\u5c71\u30ad\u30e3\u30f3\u30d1\u30b9 \u60c5\u5831\u79d1\u5b66\u7814\u7a76\u79d1\u68df1\u968e \u7b2c4\u8b1b\u7fa9\u5ba4 \u8b1b\u6f14\u8005: Gregor Hendel \u8b1b\u6f14\u8005: Yuji Shinano ——————————————————- Gregor Hendel<\/p>\n How to solve Integer Optimization Problems with SCIP<\/p>\n The\u00a0 general-purpose\u00a0 branch-and-cut\u00a0 solver\u00a0 SCIP\u00a0 is one of the ——————————————————-<\/p>\n Yuji Shinano<\/p>\n Towards Using over a Million CPU Cores to Solve Previously The\u00a0 Ubiquity\u00a0 Generator\u00a0 (UG)\u00a0 is\u00a0 a\u00a0 framework for the external ——————————————————-<\/p>\n —<\/p>\n","protected":false},"excerpt":{"rendered":" \uff2f\uff32\u5b66\u4f1a\u4e2d\u90e8\u652f\u90e8\u7814\u7a76\u4f1a\u3092\u4ee5\u4e0b\u306e\u3088\u3046\u306b\u958b\u50ac\u3044\u305f\u3057\u307e\u3059\uff0e \u7686\u3055\u307e\u306e\u3054\u53c2\u52a0\u3092\u304a\u5f85\u3061\u3057\u3066\u304a\u308a\u307e\u3059\uff0e \u30102016\/05\/19\u66f4\u65b0\uff1a\u8b1b\u6f14\u6982\u8981\u3092\u63b2\u8f09\u3011 — \u65e5\u6642: 6\u670822\u65e5(\u6c34) 17:00-19:00 \u5834\u6240: \u540d\u53e4\u5c4b\u5927 … \u7d9a\u304d\u3092\u8aad\u3080
\nhttp:\/\/www.co.cm.is.nagoya-u.ac.jp\/~yagiura\/access-j.html<\/a><\/p>\n
\nZuse Institute Berlin, \u7814\u7a76\u54e1
\nhttp:\/\/www.zib.de\/members\/hendel<\/a>
\n\u8b1b\u6f14\u984c\u76ee: How to solve Integer Optimization Problems with SCIP<\/p>\n
\nZuse Institute Berlin, \u7814\u7a76\u54e1
\nhttp:\/\/www.zib.de\/members\/shinano<\/a>
\n\u8b1b\u6f14\u984c\u76ee: Towards Using over a Million CPU Cores to Solve Previously
\nUnsolved Mixed Integer Programming Problem Instances<\/p>\n
\n\u8b1b\u6f14\u6982\u8981:
\n——————————————————-<\/p>\n
\nfastest\u00a0 noncommercial\u00a0 tools\u00a0 for\u00a0 solving\u00a0 integer optimization
\nproblems.\u00a0 Its plugin-based system facilitates\u00a0 custom extensions
\nfor specific\u00a0 applications.\u00a0 The goal of this talk\u00a0 is to provide
\nusers\u00a0 with\u00a0 a\u00a0 basic\u00a0 understanding\u00a0 of the SCIP solving process
\nand to\u00a0 illustrate some\u00a0 ways to tackle custom\u00a0 projects.\u00a0 In the
\nfirst part of this talk,\u00a0 I will give an overview\u00a0 of the solving
\nprocess of\u00a0 SCIP and\u00a0 the role\u00a0 of the different plugin types.\u00a0 I
\nwill\u00a0 introduce\u00a0 the\u00a0 most\u00a0 important\u00a0 plugin\u00a0 types of SCIP with
\nspecial focus on its available primal heuristics. The second part
\ndeals with the customization of SCIP at the example of the famous
\nTraveling\u00a0 Salesman\u00a0 Problem.\u00a0 Three\u00a0 alternative\u00a0 approaches\u00a0 to
\nformulate and solve TSP’s will\u00a0 be discussed and their particular
\nadvantages will be highlighted:\u00a0 via the modeling language ZIMPL,
\nthe C\/C++ callable library of SCIP, and the new Python interface.<\/p>\n
\nUnsolved Mixed Integer Programming Problem Instances<\/p>\n
\nparallelization\u00a0 of\u00a0 MIP\u00a0 solvers.\u00a0 UG\u00a0 provides a systematic way
\nto\u00a0 develop\u00a0 a\u00a0 parallel\u00a0 solver\u00a0 that\u00a0 can\u00a0 run\u00a0 on\u00a0 large-scale
\ndistributed memory computing environments. It was used to develop
\nParaSCIP,\u00a0 a\u00a0 distributed\u00a0 memory,\u00a0 massively parallel version of
\nthe open\u00a0 source academic\u00a0 solver SCIP.\u00a0 In this\u00a0 talk we present
\na\u00a0 success\u00a0 story\u00a0 where\u00a0 we\u00a0 solve\u00a0 14\u00a0 open\u00a0 MIP instances from
\nMIPLIB2003 and MIPLIB2010 using ParaSCIP on up to 80,000 cores of
\nsupercomputers.\u00a0 Finally, we introduce ParaXpress,\u00a0 for which one
\nof the fastest commercial MIP solvers, the FICO Xpress-Optimizer,
\nhas been parallelized by UG. Combining the internal shared-memory
\nparallelization of Xpress and the external parallelization of UG,
\nwe aim at\u00a0 a new order of magnitude\u00a0 for supercomputer core-usage
\nin MIP solving.<\/p>\n