6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

Norm(mathematics)
FromWikipedia,thefreeencyclopedia

Inlinearalgebra,functionalanalysis,andrelatedareasofmathematics,anormisafunctionthatassignsastrictly
positivelengthorsizetoeachvectorinavectorspacesaveforthezerovector,whichisassignedalengthof
zero.Aseminorm,ontheotherhand,isallowedtoassignzerolengthtosomenonzerovectors(inadditiontothe
zerovector).
Anormmustalsosatisfycertainpropertiespertainingtoscalabilityandadditivitywhicharegivenintheformal
definitionbelow.
Asimpleexampleisthe2dimensionalEuclideanspaceR2equippedwiththeEuclideannorm.Elementsinthis
vectorspace(e.g.,(3,7))areusuallydrawnasarrowsina2dimensionalcartesiancoordinatesystemstartingatthe
origin(0,0).TheEuclideannormassignstoeachvectorthelengthofitsarrow.Becauseofthis,theEuclidean
normisoftenknownasthemagnitude.
Avectorspaceonwhichanormisdefinediscalledanormedvectorspace.Similarly,avectorspacewitha
seminormiscalledaseminormedvectorspace.Itisoftenpossibletosupplyanormforagivenvectorspacein
morethanoneway.

Contents
1 Definition
2 Notation
3 Examples
3.1 Absolutevaluenorm
3.2 Euclideannorm
3.2.1 Euclideannormofacomplexnumber
3.3 TaxicabnormorManhattannorm
3.4 pnorm
3.5 Maximumnorm(specialcaseof:infinitynorm,uniformnorm,orsupremumnorm)
3.6 Zeronorm
3.6.1 Hammingdistanceofavectorfromzero
3.7 Othernorms
3.8 Infinitedimensionalcase
4 Properties
5 Classificationofseminorms:absolutelyconvexabsorbingsets
6 Generalizations
7 Seealso
8 Notes
9 References

Definition
GivenavectorspaceVoverasubfieldFofthecomplexnumbers,anormonVisafunctionp:VRwiththe
followingproperties:[1]
ForallaFandallu,vV,
https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

1/9

6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

1.p(av)=|a|p(v),(absolutehomogeneityorabsolutescalability).
2.p(u+v)p(u)+p(v)(triangleinequalityorsubadditivity).
3.Ifp(v)=0thenvisthezerovector(separatespoints).
Bythefirstaxiom,absolutehomogeneity,wehavep(0)=0andp(v)=p(v),sothatbythetriangleinequality
p(v)0(nonnegativity).
AseminormonVisafunctionp:VRwiththeproperties1.and2.above.
EveryvectorspaceVwithseminormpinducesanormedspaceV/W,calledthequotientspace,whereWisthe
subspaceofVconsistingofallvectorsvinVwithp(v)=0.TheinducednormonV/Wisclearlywelldefinedand
isgivenby:
p(W+v)=p(v).
Twonorms(orseminorms)pandqonavectorspaceVareequivalentifthereexisttworealconstantscandC,
withc>0suchthat
foreveryvectorvinV,onehasthat:cq(v)p(v)Cq(v).
Atopologicalvectorspaceiscallednormable(seminormable)ifthetopologyofthespacecanbeinducedbya
norm(seminorm).

Notation
Ifanormp:VRisgivenonavectorspaceVthenthenormofavectorvVisusuallydenotedbyenclosingit
withindoubleverticallines:v=p(v).Suchnotationisalsosometimesusedifpisonlyaseminorm.
ForthelengthofavectorinEuclideanspace(whichisanexampleofanorm,asexplainedbelow),thenotation|v|
withsingleverticallinesisalsowidespread.
InUnicode,thecodepointofthe"doubleverticalline"characterisU+2016.Thedoubleverticallineshouldnot
beconfusedwiththe"parallelto"symbol,UnicodeU+2225().Thisisusuallynotaproblembecausetheformer
isusedinparenthesislikefashion,whereasthelatterisusedasaninfixoperator.Thedoubleverticallineusedhere
shouldalsonotbeconfusedwiththesymbolusedtodenotelateralclicks,UnicodeU+01C1().Thesingle
verticalline|iscalled"verticalline"inUnicodeanditscodepointisU+007C.

Examples
Allnormsareseminorms.
Thetrivialseminormhasp(x)=0forallxinV.
Everylinearformfonavectorspacedefinesaseminormbyx|f(x)|.

Absolutevaluenorm
Theabsolutevalue

isanormontheonedimensionalvectorspacesformedbytherealorcomplexnumbers.
https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

2/9

6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

Euclideannorm
OnanndimensionalEuclideanspaceRn,theintuitivenotionoflengthofthevectorx=(x1,x2,...,xn)iscaptured
bytheformula

Thisgivestheordinarydistancefromtheorigintothepointx,aconsequenceofthePythagoreantheorem.The
EuclideannormisbyfarthemostcommonlyusednormonRn,butthereareothernormsonthisvectorspaceas
willbeshownbelow.Howeverallthesenormsareequivalentinthesensethattheyalldefinethesametopology.
OnanndimensionalcomplexspaceCnthemostcommonnormis

Inbothcaseswecanalsoexpressthenormasthesquarerootoftheinnerproductofthevectoranditself:

wherexisrepresentedasacolumnvector([x1x2...xn]),andxdenotesitsconjugatetranspose.
Thisformulaisvalidforanyinnerproductspace,includingEuclideanandcomplexspaces.ForEuclideanspaces,
theinnerproductisequivalenttothedotproduct.Hence,inthisspecificcasetheformulacanbealsowrittenwith
thefollowingnotation:

TheEuclideannormisalsocalledtheEuclideanlength,L2distance,2distance,L2norm,or2normseeLp
space.
ThesetofvectorsinRn+1whoseEuclideannormisagivenpositiveconstantformsannsphere.
Euclideannormofacomplexnumber
TheEuclideannormofacomplexnumberistheabsolutevalue(alsocalledthemodulus)ofit,ifthecomplex
planeisidentifiedwiththeEuclideanplaneR2.Thisidentificationofthecomplexnumberx+iyasavectorinthe
Euclideanplane,makesthequantity

(asfirstsuggestedbyEuler)theEuclideannormassociatedwith

thecomplexnumber.

TaxicabnormorManhattannorm

Thenamerelatestothedistanceataxihastodriveinarectangularstreetgridtogetfromtheorigintothepointx.

https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

3/9

6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

Thesetofvectorswhose1normisagivenconstantformsthesurfaceofacrosspolytopeofdimensionequivalent
tothatofthenormminus1.TheTaxicabnormisalsocalledthe 1norm.Thedistancederivedfromthisnormis
calledtheManhattandistanceor 1distance.
The1normissimplythesumoftheabsolutevaluesofthecolumns.
Incontrast,

isnotanormbecauseitmayyieldnegativeresults.

pnorm
Letp1bearealnumber.

Forp=1wegetthetaxicabnorm,forp=2wegettheEuclideannorm,andaspapproaches thepnorm
approachestheinfinitynormormaximumnorm.Thepnormisrelatedtothegeneralizedmeanorpowermean.
Thisdefinitionisstillofsomeinterestfor0<p<1,buttheresultingfunctiondoesnotdefineanorm,[2]becauseit
violatesthetriangleinequality.Whatistrueforthiscaseof0<p<1,eveninthemeasurableanalog,isthatthe
correspondingLpclassisavectorspace,anditisalsotruethatthefunction

(withoutpthroot)definesadistancethatmakesLp(X)intoacompletemetrictopologicalvectorspace.These
spacesareofgreatinterestinfunctionalanalysis,probabilitytheory,andharmonicanalysis.However,outside
trivialcases,thistopologicalvectorspaceisnotlocallyconvexandhasnocontinuousnonzerolinearforms.Thus
thetopologicaldualspacecontainsonlythezerofunctional.
Thepartialderivativeofthepnormisgivenby

Thederivativewithrespecttox,therefore,is

where denotesHadamardproductand

isusedforabsolutevalueofeachcomponentofthevector.

Forthespecialcaseofp=2,thisbecomes
https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

4/9

6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

or

Maximumnorm(specialcaseof:infinitynorm,uniformnorm,orsupremumnorm)
If issomevectorsuchthat

,then:

Thesetofvectorswhoseinfinitynormisagivenconstant,c,formsthesurfaceofa
hypercubewithedgelength2c.

Zeronorm
Inprobabilityandfunctionalanalysis,thezeronorminducesacompletemetric
topologyforthespaceofmeasurablefunctionsandfortheFspaceofsequenceswith
Fnorm
,whichisdiscussedbyStefanRolewiczinMetricLinearSpaces.[3]Here
wemeanbyFnormsomerealvaluedfunction
onanFspacewithdistanced,suchthat
shouldnotethattheFnormdescribedaboveisnotanormintheusualsensebecauseitlackstherequired
homogeneityproperty.

.One

Hammingdistanceofavectorfromzero
Inmetricgeometry,thediscretemetrictakesthevalueonefordistinctpointsandzerootherwise.Whenapplied
coordinatewisetotheelementsofavectorspace,thediscretedistancedefinestheHammingdistance,whichis
importantincodingandinformationtheory.Inthefieldofrealorcomplexnumbers,thedistanceofthediscrete
metricfromzeroisnothomogeneousinthenonzeropointindeed,thedistancefromzeroremainsoneasitsnon
zeroargumentapproacheszero.However,thediscretedistanceofanumberfromzerodoessatisfytheother
propertiesofanorm,namelythetriangleinequalityandpositivedefiniteness.Whenappliedcomponentwiseto
vectors,thediscretedistancefromzerobehaveslikeanonhomogeneous"norm",whichcountsthenumberofnon
zerocomponentsinitsvectorargumentagain,thisnonhomogeneous"norm"isdiscontinuous.
Insignalprocessingandstatistics,DavidDonohoreferredtothezero"norm"withquotationmarks.Following
Donoho'snotation,thezero"norm"ofxissimplythenumberofnonzerocoordinatesofx,ortheHamming
distanceofthevectorfromzero.Whenthis"norm"islocalizedtoaboundedset,itisthelimitofpnormsasp
approaches0.Ofcourse,thezero"norm"isnottrulyanorm,becauseitisnotpositivehomogeneous.Indeed,itis
notevenanFnorminthesensedescribedabove,sinceitisdiscontinuous,jointlyandseverally,withrespecttothe
scalarargumentinscalarvectormultiplicationandwithrespecttoitsvectorargument.Abusingterminology,
someengineersomitDonoho'squotationmarksandinappropriatelycallthenumberofnonzerosfunctiontheL0
norm,echoingthenotationfortheLebesguespaceofmeasurablefunctions.

Othernorms
OthernormsonRncanbeconstructedbycombiningtheaboveforexample
https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

5/9

6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

isanormonR4.
ForanynormandanyinjectivelineartransformationAwecandefineanewnormofx,equalto

In2D,withAarotationby45andasuitablescaling,thischangesthetaxicabnormintothemaximumnorm.In
2D,eachAappliedtothetaxicabnorm,uptoinversionandinterchangingofaxes,givesadifferentunitball:a
parallelogramofaparticularshape,sizeandorientation.In3Dthisissimilarbutdifferentforthe1norm
(octahedrons)andthemaximumnorm(prismswithparallelogrambase).
AlltheaboveformulasalsoyieldnormsonCnwithoutmodification.

Infinitedimensionalcase
ThegeneralizationoftheabovenormstoaninfinitenumberofcomponentsleadstotheLpspaces,withnorms

(forcomplexvaluedsequencesxresp.functionsfdefinedon
measure).

),whichcanbefurthergeneralized(seeHaar

Anyinnerproductinducesinanaturalwaythenorm
OtherexamplesofinfinitedimensionalnormedvectorspacescanbefoundintheBanachspacearticle.

Properties
Theconceptofunitcircle(thesetofallvectorsofnorm1)isdifferentindifferentnorms:forthe1normtheunit
circleinR2isasquare,forthe2norm(Euclideannorm)itisthewellknownunitcircle,whilefortheinfinity
normitisadifferentsquare.Foranypnormitisasuperellipse(withcongruentaxes).Seetheaccompanying
illustration.Duetothedefinitionofthenorm,theunitcircleisalwaysconvexandcentrallysymmetric(therefore,
forexample,theunitballmaybearectanglebutcannotbeatriangle).
Intermsofthevectorspace,theseminormdefinesatopologyonthespace,andthisisaHausdorfftopology
preciselywhentheseminormcandistinguishbetweendistinctvectors,whichisagainequivalenttotheseminorm
beinganorm.Thetopologythusdefined(byeitheranormoraseminorm)canbeunderstoodeitherintermsof
sequencesoropensets.Asequenceofvectors
issaidtoconvergeinnormto if
as
.
Equivalently,thetopologyconsistsofallsetsthatcanberepresentedasaunionofopenballs.
TwonormsandonavectorspaceVarecalledequivalentifthereexistpositiverealnumbersCandDsuch
thatforallxinV

https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

6/9

6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

Forinstance,on

,ifp>r>0,then

Inparticular,

Ifthevectorspaceisafinitedimensionalrealorcomplexone,allnormsareequivalent.
Ontheotherhand,inthecaseofinfinitedimensionalvectorspaces,notallnormsare
equivalent.
Equivalentnormsdefinethesamenotionsofcontinuityandconvergenceandformany
purposesdonotneedtobedistinguished.Tobemoreprecisetheuniformstructure
definedbyequivalentnormsonthevectorspaceisuniformlyisomorphic.
Every(semi)normisasublinearfunction,whichimpliesthateverynormisaconvex
function.Asaresult,findingaglobaloptimumofanormbasedobjectivefunctionis
oftentractable.
Givenafinitefamilyofseminormspionavectorspacethesum

Illustrationsofunit
circlesindifferent
norms.

isagainaseminorm.
ForanynormponavectorspaceV,wehavethatforalluandvV:
p(uv)|p(u)p(v)|.
Proof:Applyingthetriangularinequalitytoboth

and

Thus,p(uv)|p(u)p(v)|.
If and arenormedspacesand
transposeof areequal.[4]

isacontinuouslinearmap,thenthenormof andthenormofthe

Forthelpnorms,wehaveHlder'sinequality[5]

AspecialcaseofthisistheCauchySchwarzinequality:[5]
https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

7/9

6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

Classificationofseminorms:absolutelyconvexabsorbingsets
AllseminormsonavectorspaceVcanbeclassifiedintermsofabsolutelyconvexabsorbingsetsinV.Toeach
suchset,A,correspondsaseminormpAcalledthegaugeofA,definedas
pA(x):=inf{:>0,xA}
withthepropertythat
{x:pA(x)<1}A{x:pA(x)1}.
Conversely:
Anylocallyconvextopologicalvectorspacehasalocalbasisconsistingofabsolutelyconvexsets.Acommon
methodtoconstructsuchabasisistouseafamily(p)ofseminormspthatseparatespoints:thecollectionofall
finiteintersectionsofsets{p<1/n}turnsthespaceintoalocallyconvextopologicalvectorspacesothateverypis
continuous.
Suchamethodisusedtodesignweakandweak*topologies.
normcase:
Supposenowthat(p)containsasinglep:since(p)isseparating,pisanorm,andA={p<1}isitsopenunit
ball.ThenAisanabsolutelyconvexboundedneighbourhoodof0,andp=pAiscontinuous.
TheconverseisduetoKolmogorov:anylocallyconvexandlocallyboundedtopologicalvectorspaceis
normable.Precisely:
IfVisanabsolutelyconvexboundedneighbourhoodof0,thegaugegV(sothatV={gV<1})isanorm.

Generalizations
Thereareseveralgeneralizationsofnormsandseminorms.Ifpisabsolutehomogeneitybutinplaceof
subadditivitywerequirethat
2. thereisa

suchthat

forall

thenpsatisfiesthetriangleinequalitybutiscalledaquasiseminormandthesmallestvalueofbforwhichthis
holdsiscalledthemultiplierofpifinadditionpseparatespointsthenitiscalledaquasinorm.
Ontheotherhand,ifpsatisfiesthetriangleinequalitybutinplaceofabsolutehomogeneitywerequirethat
1. thereexistsaksuchthat

andforall

andscalars :

thenpiscalledakseminorm.
Wehavethefollowingrelationshipbetweenquasiseminormsandkseminorms:
SupposethatqisaquasiseminormonavectorspaceXwithmultiplierb.If
https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

thenthereexists
8/9

6/9/2016

Norm(mathematics)Wikipedia,thefreeencyclopedia

kseminormponXequivalenttoq.

Seealso
Normedvectorspace
Asymmetricnorm
Matrixnorm
Gowersnorm
Mahalanobisdistance
Manhattandistance
Relationofnormsandmetrics

Notes
1.Prugoveki1981,page20(http://books.google.com/books?id=GxmQxn2PF3IC&pg=PA20)
2.ExceptinR1,whereitcoincideswiththeEuclideannorm,andR0,whereitistrivial.
3.Rolewicz,Stefan(1987),Functionalanalysisandcontroltheory:Linearsystems,MathematicsanditsApplications(East
EuropeanSeries)29(TranslatedfromthePolishbyEwaBednarczuked.),DordrechtWarsaw:D.ReidelPublishingCo.
PWNPolishScientificPublishers,pp.xvi,524,ISBN9027721866,MR920371,OCLC13064804
4.Trevespp.242243
5.Golub,GeneVanLoan,CharlesF.(1996).MatrixComputations(Thirded.).Baltimore:TheJohnsHopkinsUniversity
Press.p.53.ISBN080185413X.

References
Bourbaki,Nicolas(1987)."Chapters15".Topologicalvectorspaces.Springer.ISBN3540136274.
Prugoveki,Eduard(1981).QuantummechanicsinHilbertspace(2nded.).AcademicPress.p.20.ISBN0
12566060X.
Trves,Franois(1995).TopologicalVectorSpaces,DistributionsandKernels.AcademicPress,Inc.
pp.136149,195201,240252,335390,420433.ISBN0486453529.
Khaleelulla,S.M.(1982).CounterexamplesinTopologicalVectorSpaces.LectureNotesinMathematics
936.SpringerVerlag.pp.35.ISBN9783540115656.Zbl0482.46002.
Retrievedfrom"https://en.wikipedia.org/w/index.php?
title=Norm_(mathematics)&oldid=722493647#Euclidean_norm"
Categories: Norms(mathematics) Linearalgebra
Thispagewaslastmodifiedon28May2016,at14:09.
TextisavailableundertheCreativeCommonsAttributionShareAlikeLicenseadditionaltermsmayapply.
Byusingthissite,youagreetotheTermsofUseandPrivacyPolicy.Wikipediaisaregisteredtrademark
oftheWikimediaFoundation,Inc.,anonprofitorganization.

https://en.wikipedia.org/wiki/Norm_(mathematics)#Euclidean_norm

9/9