资源描述:
《approximate strong separation with application in fractional graph coloring and preemptivenew》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、Approximatestrongseparationwithappli
ationinfra
tionalgraph
oloringandpreemptives
hedulingylausansennstitutfurnformatikundpraktis
heathematik,Universitatzuiel,Germanykjinformatik.uni-kiel.deAbstra
tnthispaperweshowthatapproximationalgorithmsfortheweight
2、edindependentsetands-dimensionalknapsa
kproblemwithratioa
anbeturnedintoapproximationalgorithmswiththesameratioforfra
tionalweightedgraph
oloringandpreemptiveresour
e
onstraineds
heduling.nordertoobtaintheseresults,wegeneralizeknownresultsbyGrots
hel,ovaszandS
3、
hrijveronseparation,non-emptinesstest,optimizationandviolationinthedire
tionofapproximability.1ntrodu
tionThefra
tionalgraph
oloringisdenedasfollows(seealso[10,15,24,26,27,28,29℄).etG=(V;E)beagraphwithapositiveweightwforea
hvertexv2V.etbethesetofvallindepe
4、ndentsetsofG.Theweightedfra
tional
oloringproblem
onsistsofassigninganon-negativerealvaluextoea
hindependentsetofGsu
hthatea
hvertexv2Vis
ompletely
overedbyindependentsets
ontainingv(i.e.thesumoftheirvaluesisatleastw)andthevtotalvaluexisminimized.Grots
heleta
5、l.[15℄provedthattheweightedfra
tional
oloringproblemis-hardforgeneralgraphs,but
anbesolvedinpolynomialtimeforperfe
tgraphs.Theyhaveprovedthefollowinginterestingresult:Foranygraph
lassG,iftheproblemof
omputing(G;w)(theweightofthelargestweightedindependentseti
6、nG)forgraphsG2Gis-hard,thentheproblemofdeterminingtheweightedfra
tionalhromati
number(G;w)fisalso-hard.Thisgivesanegativeresultoftheweightedfra
tional
oloringproblemevenforplanar
ubi
graphsandunitdiskgraphs.Furthermore,iftheweightedindependentsetproblemforg
7、raphsinGispolynomial-timesolvable,thentheweightedfra
tional
oloringproblemforG
analsobesolvedinpolynomialtime.Therstinapproximabilityresultfortheunweightedversionoftheproblem(i.e.wherew=1vforea
hvertexv2V)wasobtainedbyundandYannakakis[24℄whoprovedthatthereexis
8、tsaÆ>0su
hthatthereisnopolynomial-timeapproximationalgorithmfortheproblemwithÆapproximationration,unless=.Feigeandilian[10℄showedthatthefra
tionalhromati
1 numbe