ebook img

Strong isomorphism in Marinatto-Weber type quantum games PDF

0.11 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Strong isomorphism in Marinatto-Weber type quantum games

Strong isomorphism in Marinatto-Weber type quantum games PiotrFra˛ckiewicz InstituteofMathematics PomeranianUniversity,Poland January24, 2017 Abstract Ourpurposeistofocusattentiononanewcriterionforquantumschemesbybringingtogetherthenotionsof 7 quantumgameandgameisomorphism. Aquantumgameschemeisrequiredtogeneratetheclassicalgameasa 1 0 specialcase. Now,givenaquantumgameschemeandtwoisomorphicclassicalgames,weadditionallyrequire 2 theresultingquantumgamestobeisomorphicaswell. Weshowhowthisisomorphismconditioninfluencesthe players’strategysets. WeareconcernedwiththeMarinatto-Webertypequantumgameschemeandthestrong n isomorphismbetweengamesinstrategicform. a J 1 1 Introduction 2 ] The Marinatto-Weber (MW) scheme introduced in [1] is a straightforward way to apply the power of quantum T mechanicstoclassicalgametheory. Inthesimplestcaseof2 2games,theplayersmanipulatetheirownqubits G ofatwo-qubitstateeitherwiththeidentity1orthePauliopera×torσ . Therefore,ithasfoundapplicationinmany x . otherbranchesofgametheory: fromevolutionarygametheory[2],[3]toextensive-formgames[4]andduopoly s c examples[5], [6]. Inpaper[7]wepointedoutafewundesirablepropertiesoftheMWschemeandintroduceda [ refinedquantumgamemodel. 1 ThoughitispossibletoextendboththeMWschemeandourrefinementtoconsidermorecomplexgamesthan v 2 2, possible generalizationscan be defined in many differentways. A result concerning3 3 games can be 0 fo×undin[2]and[8].Theauthorsproposedsuitablethree-elementsetsofplayers’strategiestoob×tainageneralized 0 3 3 game. On the other hand, our work [9] providesanotherway to define players’strategy sets thatremains 1 × validforanyfiniten mgames. 6 × 0 Certainly,onecanfindyetotherwaystogeneralizetheMWscheme. Henceitwouldbeinterestingto place . additionalrestrictionsonaquantumgameschemeandexaminehowtheyrefinethequantummodel. Inthispaper 1 weformulateacriterionintermsofisomorphicgames.Giventwoisomorphicgameswerequirethecorresponding 0 7 quantumgamestobeisomorphicaswell. If,forexample,twobimatrixgamesdifferonlyintheorderofplayers’ 1 strategies, they describe the same problem from a game-theoretical point of view. Given a quantum scheme, v: it appears reasonable to assume that the resulting quantum game will not depend on the numbering of players’ i strategiesintheclassicalgame. X r a 2 Preliminaries 2.1 Marinatto-weber type quantum gamescheme Inpaper[7]and[10]wepresentedarefinementoftheMarinatto-Weberscheme[1].Themotivationofconstructing ourschemewastwofold. Ourmodelenablestheplayerstochoosebetweenplayingafixedquantumstrategyand classicalstrategies.Thesecondaimwastoconstructtheschemethatgeneratestheclassicalgamebymanipulating theplayers’strategiesratherthantheinitialquantumstate. Inwhatfollows,werecalltheschemeforthecaseof 2 2bimatrixgame, × l r t (a ,b ) (a ,b ) 00 00 01 01 , where(a ,b ) R. (1) b (a10,b10) (a11,b11)! ij ij ∈ Definition1 Thequantumschemeforgame(1)isdefinedonaninnerproductspace(C2) 4bythetriple ⊗ Γ =(H,(S ,S ),(M ,M )), (2) Q 1 2 1 2 where 1 H isapositiveoperator, • H =(1 1 11 11) 00 00 + 11 11 Ψ Ψ, (3) ⊗ −| ih | ⊗| ih | | ih |⊗| ih | and Ψ =α00 +β01 +γ10 +δ11 C2 C2 (4) | i | i | i | i | i∈ ⊗ suchthat Ψ =1, k| ik S = P(1) U(3),i, j=0,1 , S = P(2) U(4),k,l=0,1 are the players’ strategy sets, and the upper • 1 i ⊗ j 2 k ⊗ l indicesnidentifythesubspaceoC2of(Cn2) 4onwhichtheoperoators ⊗ P = 0 0, P = 1 1, U =1, U =σ , (5) 0 1 0 1 x | ih | | ih | aredefined, M andM arethemeasurementoperators 1 2 • M =1 1 a (b )xy xy (6) 1(2) ⊗ ⊗ xy xy | ih | x,Xy=0,1  thatdependonthepayoffsa andb from(1). xy xy TheschemeproceedsinthesimilarwayastheMWscheme–theplayersdeterminethefinalstatebychoosingtheir strategiesandactingonoperatorH. Asaresult,theydeterminethefollowingdensityoperator: ρ = P(1) P(2) U(3) U(4) H P(1) P(2) U(3) U(4) f i ⊗ k ⊗ j ⊗ l i ⊗ k ⊗ j ⊗ l (cid:16) (cid:17) (cid:16) (cid:17) 11 11 U(3) U(4)Ψ ΨU(3) U(4) ifi= j=1, = | ih |⊗ j ⊗ l | ih | j ⊗ l (7) ij ij U(cid:16)(3) U(4)00 00U(3) U(4)(cid:17) ifotherwise. | ih |⊗ j ⊗ l | ih | j ⊗ l Next,thepayoffsforplayer1and2are (cid:16) (cid:17) tr(ρ M ) and tr(ρ M ). (8) f 1 f 2 Asitwasshownin[10],scheme(2)canbesummarizedbythefollowingmatrixgame P(2) 1(4) P(2) σ(4) P(2) 1(4) P(2) σ(4) 0 ⊗ 0 ⊗ x 1 ⊗ 1 ⊗ x P(1) 1(3) X X X X 0 ⊗ 00 01 00 01 P(1) σ(3)  X X X X  PP(1(1011))⊗⊗⊗σ1((xx44))  XX101000 XX101111 ∆∆101000 ∆∆101111 , (9) where X =(a ,b ), for i, j=0,1 ij ij ij ∆ = α2X + β2X + γ2X + δ2X , 00 00 01 10 11 | | | | | | | | ∆ = α2X + β2X + γ2X + δ2X , (10) 01 01 00 11 10 | | | | | | | | ∆ = α2X + β2X + γ2X + δ2X , 10 10 11 00 01 | | | | | | | | ∆ = α2X + β2X + γ2X + δ2X . 11 11 10 01 00 | | | | | | | | 2.2 Strong isomorphism Thenotionofstrongisomorphismdefinesclassesofgamesthatarethesameuptothenumberingoftheplayers andthe orderofplayers’strategies. Thefollowingdefinitionsaretakenfrom[11](see also [12], [13] and[14]). Thefirstonedefinesamappingthatassociatesplayersandtheiractionsinonegamewithplayersandtheiractions intheothergame. Definition2 LetΓ=(N,(S ) ,(u) )andΓ =(N,(S ) ,(u) )begamesinstrategicform.Agamemapping i i N i i N ′ i′ i N ′i i N f fromΓtoΓ isatuple f =(η∈,(ϕ) ∈),whereηisabijecti∈onfrom∈NtoNandforanyi N,ϕ isabijectionfrom ′ i i N i ∈ ∈ S toS . i η′(i) 2 Ingeneralcase,themapping f from(N,(S ) ,(u) )to(N,(S ) ,(u) )identifiesplayeri N withplayer i i∈N i i∈N i′ i∈N ′i i∈N ∈ η(i) and maps S to S . This means that a strategy profile (s ,...,s ) S S is mapped into profile i η(i) 1 n 1 n (s ,...,s )thatsatisfiesequations =ϕ(s)fori N. ∈ ×···× ′1 ′n ′η(i) i i ∈ The notion of game mapping is a basis for the definition of game isomorphism. Depending on how rich structureofthegameistobepreservedwecandistinguishvarioustypesofgameisomorphism.Onethatpreserves theplayers’payofffunctionsiscalledastrongisomorphism.Theformaldefinitionisasfollows: Definition3 Given two strategic games Γ = (N,(S ) ,(u) ) and Γ = (N,(S ) ,(u) ), a game mapping i i N i i N ′ i′ i N ′i i N f =(η,(ϕ) )iscalledastrongisomorphismifrelatio∈nu(s)∈=u (f(s))holdsfor∈eachi ∈N andeachstrategy i i∈N i ′η(i) ∈ profiles S S . 1 n ∈ ×···× FromtheabovedefinitionitmaybeconcludedthatifthereisastrongisomorphismbetweengamesΓandΓ,they ′ maydiffermerelybythenumberingofplayersandtheorderoftheirstrategies. Thefollowinglemmashowsthatrelabelingplayersandtheirstrategiesdonotaffectthegamewith regardto Nashequilibria. If f isastrongisomorphismbetweengamesΓandΓ,onemayexpectthattheNashequilibriain ′ ΓmaptoonesinΓ under f. ′ Lemma1 Let f beastrongisomorphismbetweengamesΓandΓ. Strategyprofiles =(s ,...,s ) S S isaNashequilibriumingameΓifandonlyif f(s ) S S′ isaNashequilib∗rium∗1inΓ. ∗n ∈ 1×···× n ∗ ∈ 1′ ×···× n′ ′ 3 Application of game isomorphism to Marinatto-Weber type quantum game schemes ItisnothardtoseethatwecandefineawidevarietyofschemesbasedontheMWapproach.Wecanmodifyoper- ator(3)andtheplayers’strategiestoconstructanotherschemestillsatisfyingtherequirementaboutgeneralization oftheinputgame.Thefollowingexampleofsuchaschemeisparticularlyinteresting. Letusconsideratriple Γ =(H ,(S ,S ),(M ,M )) (11) ′Q ′ 1′ 2′ 1 2 withthecomponentsdefinedasfollows: H isapositiveoperator, ′ • H = 00 00 00 00 + 01 01 0 0 ρ + 10 10 ρ 0 0 + 11 11 Ψ Ψ, (12) ′ 2 1 | ih |⊗| ih | | ih |⊗| ih |⊗ | ih |⊗ ⊗| ih | | ih |⊗| ih | where Ψ C2 C2 such that Ψ = 1, ρ and ρ are the reduced density operators of Ψ Ψ, i.e., 1 2 ρ =tr|(Ψi ∈Ψ)an⊗dρ =tr (Ψ Ψk|),ik | ih | 1 2 2 1 | ih | | ih | S = P(1) 1(3),P(1) σ(3),P(1) 1(3) andS = P(2) 1(4),P(2) σ(4),P(2) 1(4) aretheplayers’strat- • 1′ 0 ⊗ 0 ⊗ x 1 ⊗ 2′ 0 ⊗ 0 ⊗ x 1 ⊗ egysents, o n o M andM arethemeasurementoperatorsdefinedbyequation(6). 1 2 • Itisimmediatethattheresultingfinalstateρ isadensityoperatorforeach(pureormixed)strategyprofile. For ′f example,player1’sstrategyP(1) σ(3) andplayer2’sstrategyP(2) 1(4)imply 0 ⊗ x 1 ⊗ ρ = P(1) P(2) σ(3) 1(4) H P(1) P(2) σ(3) 1(4) = 01 01 1 1 ρ . (13) ′f 0 ⊗ 1 ⊗ x ⊗ ′ 0 ⊗ 1 ⊗ x ⊗ | ih |⊗| ih |⊗ 2 (cid:16) (cid:17) (cid:16) (cid:17) Asaresult,theplayers’payofffunctifonsu andu givenbytr(ρ M )andtr(ρ M ),respectively,arewell-defined. ′1 ′2 ′f 1 ′f 2 It is also clear that scheme (11) produces the classical game in a similar way to scheme (2). The players play the classical game as long as they choose the strategies P 1 and P σ . This can be seen by determin- 0 0 x ⊗ ⊗ ing tr(ρ M ) for each strategy profile and arranging the obtained values into a matrix. As an example, let us ′f 1(2) determinetr(ρ M )forthefinalstateρ givenby(13). Let Ψ representageneraltwoqubitstate, ′f 1(2) ′f | i Ψ =α00 +β01 +γ10 +δ11 . (14) | i | i | i | i | i Since ρ =(α2+ β2)0 0 +(αγ +βδ )0 1 +(γα +δβ )1 0 +(γ2+ δ2)1 1, 1 ∗ ∗ ∗ ∗ | | | | | ih | | ih | | ih | | | | | | ih | (15) ρ =(α2+ γ2)0 0 +(αβ +γδ )0 1 +(βα +δγ )1 0 +(β2+ δ2)1 1, 2 ∗ ∗ ∗ ∗ | | | | | ih | | ih | | ih | | | | | | ih | theplayers’strategiesP(1) σ(3) andP(2) 1(4)generatethefollowingformofthefinalstate: 0 ⊗ x 1 ⊗ ρ = 01 01 (α2+ γ2)10 10 +(αβ +γδ )10 11 +(βα +δγ )11 10 +(β2+ δ2)11 11 . (16) ′f | ih |⊗ | | | | | ih | ∗ ∗ | ih | ∗ ∗ | ih | | | | | | ih | (cid:16) (cid:17) 3 Hence (tr(ρ M ),tr(ρ M ))=(α2+ γ2)(a ,b )+(β2+ δ2)(a ,b ). (17) ′f 1 ′f 2 | | | | 10 10 | | | | 11 11 Thevalues(tr(ρ M ),tr(ρ M ))forallstrategycombinationsaregivenbythefollowingmatrix: ′f 1 ′f 2 P(2) 1(4) P(2) σ(4) P(2) 1(4) 0 ⊗ 0 ⊗ x 1 ⊗ P(1) 1(3) X X ∆ 0 ⊗ 00 01 02 P(1) σ(3)  X X ∆  (18) P0(11)⊗⊗1(x3)  ∆1200 ∆1211 ∆1222  where X =(a ,b ), for i, j=0,1 ij ij ij ∆ =(α2+ γ2)X +(β2+ δ2)X ; 02 00 01 | | | | | | | | ∆ =(α2+ γ2)X +(β2+ δ2)X ; 12 10 11 | | | | | | | | ∆ =(α2+ β2)X +(γ2+ δ2)X ; 20 00 10 | | | | | | | | ∆ =(α2+ β2)X +(γ2+ δ2)X ; 21 01 11 | | | | | | | | ∆ = α2X + β2X + γ2X + δ2X . (19) 22 00 01 10 11 | | | | | | | | Itfollowseasilythatmatrixgame(18)isagenuineextensionof(1). Althoughpayoffprofiles∆ , ∆ arealso ij 22 achievablein(1),theplayers,ingeneral,arenotabletoobtain∆ whenchoosingtheir(mixed)strategies. 22 Tosumup, scheme(11)mightseemto beacceptableaslongasscheme(2)isacceptable. Matrixgame(18) includes(1)anddependingontheinitialstate Ψ itmaygiveextraordinaryNashequilibria. Itisworthpointing | i outthattheNashequilibriain(18)correspondtocorrelatedequilibriain(1),(see[15]).Howeverscheme(11)fails toimplytheisomorphicgameswhentheinputgamesareisomorphic. Wecanmakethisclearwiththefollowing example. Example1 Letusconsiderthegameof“Chicken"Γ andits(strongly)isomorphiccounterpartΓ , 1 2 l r l r ′ ′ t (6,6) (2,7) t (2,7) (6,6) Γ : , Γ : ′ . (20) 1 b (7,2) (0,0)! 2 b′ (0,0) (7,2)! Thecorrespondingisomorphism f =(π,ϕ ,ϕ )isdefinedbycomponents 1 2 π(i)=i for i=1,2, ϕ =(t t ,b b),ϕ =(l r ,r l). (21) 1 ′ ′ 2 ′ ′ → → → → Set Ψ =(00 + 01 + 10 )/√3. Using(9)wecanwritequantumapproach(2)togames(20)as | i | i | i | i P(2) 1(4) P(2) σ(4) P(2) 1(4) P(2) σ(4) 0 ⊗ 0 ⊗ x 1 ⊗ 1 ⊗ x P(1) 1(3) (6,6) (2,7) (6,6) (2,7) 0 ⊗   ΓQ1: PPP(1(0(1111)))⊗⊗⊗σσ1(((xx443)))  (((776,,,226))) (((002,,,007))) (4((7531,,,225))32) (2((3032,,,430))31)  (22) and P(2) 1(4) P(2) σ(4) P(2) 1(4) P(2) σ(4) 0 ⊗ 0 ⊗ x 1 ⊗ 1 ⊗ x P(1) 1(3) (2,7) (6,6) (2,7) (6,6) 0 ⊗   It is fairly easy to seeΓthQa2t:gPPPa(1(0(1m111)))e⊗⊗⊗sσσ1(2(((xx4243)))) and((((0022,,,3007))))differ in(((776th,,,226e)))order o(2f((3032th,,,430e))31fi)rst tw(4o((5731s,,,t252ra))32t)egies and the second(t2w3o) strategiesofplayer2.Thus,thegamesarestronglyisomorphic.Moreformally,onecancheckthatagamemapping f˜=(η,ϕ˜ ,ϕ˜ ),where 1 2 ϕ˜ = P(1) 1(3) P(1) 1(3),P(1) σ(3) P(1) σ(3) , 1 (cid:16) i ⊗ → i ⊗ i ⊗ x → i ⊗ x (cid:17) (24) ϕ˜ = P(2) 1(4) P(2) σ(4),P(2) σ(4) P(2) 1(4) , 1 k ⊗ → k ⊗ x k ⊗ x → k ⊗ (cid:16) (cid:17) 4 fori,k=0,1isastrongisomorphism. Inthenextsectionweproveamoregeneralresultaboutscheme(2)). Letusnowconsiderscheme(11). Matrix(18)intermsofinputgames(20)implies P(2) 1(4) P(2) σ(4) P(2) 1(4) 0 ⊗ 0 ⊗ x 1 ⊗ P(1) 1(3) (6,6) (2,7) (42,61) Γ′Q1: PP0(0(111))⊗⊗⊗σ1((x33))  (6(731,,42)32) (1(031,,40)32) (4(5332,,15)331)  (25) and P(2) 1(4) P(2) σ(4) P(2) 1(4) 0 ⊗ 0 ⊗ x 1 ⊗ P(1) 1(3) (2,7) (6,6) (31,62) Γ′Q2: PP0(1(011))⊗⊗⊗1σ((x33))  (1(031,,40)32) (6(731,,42)32) ((2233231,,432331)) . (26) With Lemma 1 we can show that games (25) and (26) are not isomorphic. Comparing the sets of pure Nash equilibriainbothgameswefindtheequilibriumprofiles (P(1) 1(3),P(2) σ(4)),(P(1) σ(3),P(2) 1(4)),(P(1) 1(3),P(2) 1(4)) (27) { 0 ⊗ 0 ⊗ x 0 ⊗ x 0 ⊗ 1 ⊗ 1 ⊗ } inthefirstgameand (P(1) 1(3),P(2) 1(4)),(P(1) σ(3),P(2) σ(4)) (28) { 0 ⊗ 0 ⊗ 0 ⊗ x 0 ⊗ x } inthesecondone. 4 ApplicationofgameisomorphismtogeneralizedMarinatto-Weberquan- tum game scheme Additionalcriteriaforaquantumgameschememayhaveasignificantimpactonthewayhowwegeneralizethese schemes. It can be easily seen in the case of the MW scheme [1] (or the refined scheme (2)), where the sets of unitarystrategiesarefinite. TheMWschemeprovidesuswithaquantummodel,wherethestrategysetsconsistof theidentityoperator1andthePaulioperatorσ . Underthisdescription,whatsubsetsofunitaryoperatorswould x besuitableforgeneraln mgames? Thecaseofa3-elementstrategysetcanbeidentifiedwithunitaryoperators 1 ,CandDactingonα0× +β1 +γ2 C3,where 3 | i | i | i∈ 1 0 = 0 , C0 = 2 , D0 = 1 , 3 1 |1i=|1i, C|1i=|1i, D|1i=|0i, (29) 3 1 |2i=|2i, C|2i=|0i, D|2i=|2i. 3 | i | i | i | i | i | i This construction can be found in [2] and [8]. Another way to generalize the MW scheme was presented in [9]. Having given a strategic-form game, we identify the players’ n strategies with n unitary operators V for k k=0,1,...,n 1. Theyactonstatesofthecomputationalbasis 0 , 1 ,..., n 1 asfollows: − {| i | i | − i} V i = i , 0 V |ii=|ii+1modn , 1 .|i | i (30) . . V i = i+n 1modn . n 1 − |i | − i BothwaystogeneralizetheMWschemeenableustoobtaintheclassicalgame. Soatthislevel,neither(29)nor (30) is questionable. If we seek otherproperties, we see that the MW scheme outputsthe classical game (or its isomorphiccounterpart)whentheinitialstateisoneofthecomputationalbasisstates. Given(29)and(30),only thelattercasesatisfiesthiscondition. FurtheranalysiswouldshowthattheMWschemeisinvariantwithrespect tostronglyisomorphicinputgames.Itturnsoutthatneither(29)or(30)satisfiestheisomorphismproperty. Example2 Letustakealookatthefollowing2 3bimatrixgames: × l m r l m r ′ ′ ′ t (4,8) (0,0) (8,8) t (0,0) (4,8) (8,8) Γ: , Γ : ′ . (31) b (0,4) (4,0) (8,0)! ′ b′ (4,0) (0,4) (8,0)! 5 ConsidertheMW-typeapproachesΓ andΓ togames(31)accordingtothefollowingassignement: Q ′Q l m r t P P P 00 01 02 , whereP = j j j j . (32) b P10 P11 P12! j1j2 | 1 2ih 1 2| Then Γ =(Ψ ,(D ,D ),(M ,M )), Γ =(Ψ ,(D ,D ),(M ,M )), (33) Q | i 1 2 1 2 ′Q | i ′1 ′2 1′ 2′ where M =4P +8P +4P +8P , M =4P +8P +4P +8P , 1 00 02 11 12 1′ 01 02 10 12 (34) M =8P +8P +4P , M =8P +8P +4P . 2 00 02 10 2′ 01 02 11 Settheinitialstate Ψ = (1/2)00 +(√3/2)12 C2 C3 andassumefirstthat D = D = 1 ,σ andD = | i | i | i ∈ ⊗ 1 ′1 { 2 x} 2 aDn′2d=i={113,,C2,,aDn}dadsoiinn(g2s9i)m.iDlaertecramlciunliantgiotnrs(cid:16)(iUn1th⊗eUca2s)e|ΨoifhΨM|(Uw1†e⊗obUta2†i)nMi(cid:17)foreveryU1⊗U2 ∈{1,σx}⊗{13,C,D} i′ 1 C D 1 C D 3 3 1 (7,2) (2,5) (6,0) 1 (6,0) (5,2) (7,2) Γ : 2 , Γ : 2 . (35) Q σx (6,7) (5,6) (7,6)! ′Q σx (7,6) (2,0) (6,7)! Ontheotherhand,replacing(29)by(30)givesD = D = 1 ,V ,V ,where 2 ′2 { 3 1 2} 1 0 0 0 0 1 0 1 0 13 = 0 1 0 , V1 = 1 0 0 , V2 = 0 0 1 . (36)  0 0 1   0 1 0   1 0 0  Then,wehave 1 V V 1 V V 3 1 2 3 1 2 1 (7,2) (0,3) (5,2) 1 (6,0) (4,2) (2,5) Γ : 2 , Γ : 2 . (37) Q σx (6,7) (4,6) (2,0)! ′Q σx (7,6) (0,1) (5,6)! ThereisnopureNashequilibriuminthefirstgameof(35)and(37),whereastherearetwoNashequilibriainthe secondgames.Asaresult,eachpairofthegamesdonotdetermineastrongisomorphism. Example2showsthatplayers’strategysetsdefinedby(29)and(36)needtoberevisedinordertohaveageneral- izedMWschemeinvariantwithrespecttotheisomorphism. Weshallstickforthemomenttoconsideringgames (31). Let A ,A ,A ,A ,A ,A beplayer2’sstrategysetdefinedtobe 012 102 021 120 201 210 { } A 0 = 0 A 0 = 1 A 0 = 0 A 0 = 1 A 0 = 2 A 0 = 2 , 012 102 021 120 201 210 A |1i=|1i A |1i=|0i A |1i=|2i A |1i=|2i A |1i=|0i A |1i=|1i, (38) 012 102 021 120 201 210 A |2i=|2i A |2i=|2i A |2i=|1i A |2i=|0i A |2i=|1i A |2i=|0i. 012 102 021 120 201 210 | i | i | i | i | i | i | i | i | i | i | i | i Each A isapermutationmatrixthatcorrespondstoa specificpermutationπ = (0 j ,1 j ,2 j )of j1j2j3 → 1 → 2 → 3 the set 0,1,2 . Note also that operators(29) and (36) are included in (38). Hence, the MW scheme with (38) { } implies,inparticular,theclassicalgame. Wenowcheckifitoutputstheisomorphicgames. Sincetherearenow sixoperatorsavailableforplayer2,theresultinggamemaybewrittenasa2 6bimatrixgamewithentries × tr (U U )Ψ Ψ(UT UT)M (39) 1⊗ 2 | ih | 1 ⊗ 2 i (cid:16) (cid:17) forU 1 ,σ andU A : π permutationsof 0,1,2 .Asaresult,weobtain 1 2 x 2 π ∈{ } ∈{ − { }} A A A A A A 012 102 021 120 201 210 1 (7,2) (6,0) (4,2) (0,3) (5,2) (2,5) Γ : 2 (40) Q σ (6,7) (7,6) (0,1) (4,6) (2,0) (5,6) x   and A A A A A A 012 102 021 120 201 210 1 (6,0) (7,2) (0,3) (4,2) (2,5) (5,2) Γ : 2 . (41) ′Q σ (7,6) (6,7) (4,6) (0,1) (5,6) (2,0) x   6 Thegamesdeterminetheisomorphism f˜=(id ,ϕ˜ ,ϕ˜ ),where N 1 2 ϕ˜ =(1 1 ,σ σ ), 1 2 2 x x → → (42) ϕ˜ =(A A ,A A ,A A ,A A ,A A ,A A ). 2 012 102 102 012 021 120 120 021 201 210 210 201 → → → → → → UsingpermutationmatricesleadsustoformulateanothergeneralizedMWscheme.Forsimplicity,weconfine attentionto(n+1) (m+1)bimatrixgames. × LetS bethesetofallpermutationsπof 0,1,...,n . WitheachπthereisassociatedapermutationmatrixA , n π { } A i = π(i) fori=0,1,...,n. (43) π |i | i WeletB denotethepermutationmatrixassociatedwithapermutationσ S . Given(n+1) (m+1)bimatrix σ m gameΓwedefine ∈ × Γ =(Ψ ,(D ,D ),(M ,M )), (44) Q 1 2 1 2 | i where n m Ψ = α j j Cn+1 Cm+1, D = A : π S , D = B : σ S , | i j1j2| 1 2i∈ ⊗ 1 { π ∈ n} 2 { σ ∈ m} Xj1=0Xj2=0 (45) n m (M ,M )= (a ,b )P . 1 2 j1j2 j1j2 j1j2 Xj1=0Xj2=0 BeforestatingthemainresultofthissectionwestartwiththeobservationthattheMWschemeremainsinvariant tonumberingoftheplayers.Considertwoisomorphicbimatrixgames: t t t 0 1 m ··· s (a ,b ) (a ,b ) (a ,b ) 0 00 00 01 01 0m 0m ··· s (a ,b ) (a ,b ) (a ,b ) Γ: s...n1 (an100,...bn100) (an111,...bn111) ··.··.··. (an1mm,...bn1mm) (46) and s s s ′0 ′1 ··· ′n t (b ,a ) (b ,a ) (b ,a ) 0′ 00 00 10 10 ··· n0 n0 t  (b ,a ) (b ,a ) (b ,a )  Γ′: tm...1′′ (b00m1,...a00m1) (b11m1,...a11m1) ··.··.··. (bnnm1,...annm1). (47) Clearly,theisomorphismisdefinedbyagamemapping f = π,ϕ ,ϕ ,where 1 2 { } π=(1 2,2 1), ϕ (s )= s , ϕ (t )=t (48) → → 1 j1 ′j1 2 j2 ′j2 for j = 0,1,...,n, j = 0,1,...,m. The generalMW schemefor(46)is simplygivenby(44). ForΓ, we can 1 2 ′ write Γ =(Ψ ,(D ,D ),(M ,M )), (49) ′Q | ′i ′1 ′2 1′ 2′ where n m Ψ = α j j Cm+1 Cn+1, | ′i j1j2| 2 1i∈ ⊗ Xj1=0Xj2=0 D = B ,σ S ,D = A ,π S , (50) ′1 { σ ∈ m} ′2 { π ∈ n} n m n m M = b j j j j ,M = a j j j j . 1′ j1j2| 2 1ih 2 1| 2′ j1j2| 2 1ih 2 1| Xj1=0Xj2=0 Xj1=0Xj2=0 Gamesdeterminedby(44)and(49)arethenisomorphic.Toprovethis,let f˜=(π,ϕ˜ ,ϕ˜ )beagamemappingsuch 1 2 that π=(1 2,2 1), ϕ˜ : D D , ϕ˜ (A )= A , ϕ˜ : D D , ϕ˜ (B )= B . (51) → → 1 1 → ′2 1 π π 2 2 → ′1 2 σ σ 7 OnaccountofDefinition3wehave f˜(A B )=ϕ (B ) ϕ (A )= B A . (52) π σ 2 σ 1 π σ π ⊗ ⊗ ⊗ Asaresult, u (f˜(A B ))=u (f˜(A B )) ′π(1) π⊗ σ ′2 π⊗ σ =tr f˜(A B )Ψ Ψ f˜(A B )TM π⊗ σ | ′ih ′| π⊗ σ 2′ (cid:16) (cid:17) =tr (ϕ (B ) ϕ (A ))Ψ Ψ (ϕ (B )T ϕ (A )T)M 2 σ ⊗ 1 π | ′ih ′| 2 σ ⊗ 1 π 2′ (cid:16) (cid:17) n m =tr (B A )Ψ Ψ (BT AT) a j j j j  σ⊗ π | ′ih ′| σ⊗ π j1j2| 2 1ih 2 1|  nXj1=0mXj2=0  =tr (A B )Ψ Ψ(AT BT) a j j j j  π⊗ σ | ih | π ⊗ σ j1j2| 1 2ih 1 2| =u(A B ). Xj1=0Xj2=0  (53) 1 π σ ⊗ Byasimilarargument,wecanshowthatu (f˜(A B )) = u (A B ). Wecannowformulatethefollowing ′π(2) π⊗ σ 2 π⊗ σ proposition: Proposition1 AssumethatΓandΓ arestronglyisomorphicbimatrixgamesandΓ andΓ arethecorresponding ′ Q ′Q quantumgamesdefinedby(44). ThenΓ andΓ arestronglyisomorphic. Q ′Q Proof Let Γ and Γ be bimatrix games of dimension n m and let f: Γ Γ, f = (η,ϕ ,ϕ ) be the strong ′ ′ 1 2 × → isomorphism. Since the MW scheme is invariant to numbering of the players, there is no loss of generality in assumingη=id . Now,itfollowsfromDefinition3thatgamesΓandΓ differintheorderofplayers’strategies. N ′ Let us identify players’ strategies (i.e., rows and columns) in Γ with sequences (0,1,...,n) and (0,1,...,m), respectively. Then,we denotebyπ andσ thepermutationsofthesets 0,1,...,n and 0,1,...,m associated ∗ ∗ with the order of strategies in game Γ. A trivial verification shows tha{t the payoff} oper{ator M in}Γ may be ′ i′ ′Q writtenas M =(A B )M (A B )T, (54) i′ π∗ ⊗ σ∗ i π∗ ⊗ σ∗ where A and B are the permutation matrices corresponding to π and σ . Define a game mapping f˜ = π σ ∗ ∗ ∗ ∗ (id ,ϕ˜ ,ϕ˜ ),where N 1 2 ϕ˜ : A : π S A : π S , ϕ (A )= A A ; 1 π n π n 1 π π π { ∈ }→{ ∈ } ∗ (55) ϕ˜ : B : σ S B : σ S , ϕ (B )= B B . 2 σ m σ m 2 σ σ σ { ∈ }→{ ∈ } ∗ Hence, f˜mapsA B toA A B B . Thus,weobtain π σ π π σ σ ⊗ ∗ ⊗ ∗ u(f˜(A B ))=tr f˜(A B )Ψ Ψ f˜(A B ) T M ′i π⊗ σ (cid:18) π⊗ σ | ih |(cid:16) π⊗ σ (cid:17) i′(cid:19) =tr (A A B B )Ψ Ψ(A A B B )T(A B )M (A B )T π π σ σ π π σ σ π σ i π σ ∗ ⊗ ∗ | ih | ∗ ⊗ ∗ ∗ ⊗ ∗ ∗ ⊗ ∗ (cid:16) (cid:17) =tr (A B )T(A A B B )Ψ Ψ(A A B B )T(A B )M π σ π π σ σ π π σ σ π σ i ∗ ⊗ ∗ ∗ ⊗ ∗ | ih | ∗ ⊗ ∗ ∗ ⊗ ∗ (cid:16) (cid:17) =tr AT A A BT B B Ψ Ψ ATAT A BTBT B M π∗ π∗ π⊗ σ∗ σ∗ σ | ih | π π∗ π∗ ⊗ σ σ∗ σ∗ i (cid:16)(cid:16) (cid:17) (cid:16) (cid:17) (cid:17) =tr (A B ) Ψ Ψ (A B )T M =u(A B ). (56) π σ π σ i i π σ ⊗ | ih | ⊗ ⊗ (cid:16) (cid:17) Thisfinishestheproof. (cid:4) Note that operators (43) come down to 1 and σ for n = 1. Therefore, the original MW scheme preserves the x isomorphism.ThesameconclusioncanbedrawnfortherefinedMWscheme(2). Corollary1 If Γ and Γ are strongly isomorphic bimatrix games and Γ and Γ are the corresponding games ′ Q ′Q definedby(2). ThenΓ andΓ arestronglyisomorphic. Q ′Q Proof Let Γ and Γ be strongly isomorphic 2 2 bimatrix games. By Proposition 1 there exists a strong iso- ′ morphism f˜ = (id ,ϕ˜ ,ϕ˜ ) between the game×sΓ and Γ playedaccordingto (44)-(45). Given the quantum 1,2 1 2 Q ′Q approach(2)toΓa{nd}Γ wedefineg˜ =(id ,ξ˜ ,ξ˜ ),where ′ 1,2 1 2 { } ξ˜ : S S , ξ˜ P(1) U(3) = P(1) ϕ˜ U(3) , 1 1 → 1 1(cid:16) i ⊗ j (cid:17) i ⊗ 1(cid:16) j (cid:17) (57) ξ˜ : S S , ξ˜ P(2) U(4) = P(2) ϕ˜ U(4) . 2 2 → 2 2 k ⊗ l k ⊗ 2 l (cid:16) (cid:17) (cid:16) (cid:17) 8 Now,wehave u g˜ P(1) P(2) U(3) U(4) ′i i ⊗ k ⊗ j ⊗ l (cid:16) (cid:16) (cid:17)(cid:17) =tr g˜ P(1) P(2) U(3) U(4) Hg˜ P(1) P(2) U(3) U(4) T M (cid:18) (cid:16) i ⊗ k ⊗ j ⊗ l (cid:17) (cid:16) i ⊗ k ⊗ j ⊗ l (cid:17) 1′(cid:19) =tr P(1) P(2) ϕ˜ U(3) ϕ˜ U(4) HP(1) P(2) ϕ˜ U(3) T ϕ˜ U(4) T M (cid:18) i ⊗ k ⊗ 1(cid:16) j (cid:17)⊗ 2(cid:16) l (cid:17) i ⊗ k ⊗ 1(cid:16) j (cid:17) ⊗ 2(cid:16) l (cid:17) i′(cid:19) =tr P(1) P(2) f˜ U(3) U(4) HP(1) P(2) f˜ U(3) U(4) T M . (58) (cid:18) i ⊗ k ⊗ (cid:16) j ⊗ l (cid:17) i ⊗ k ⊗ (cid:16) j ⊗ l (cid:17) i′(cid:19) ForfixedP P ,wecanwritetherightsideof(58)intheform i k ⊗ tr(cid:18)|ikihik|⊗ f˜(cid:16)U(j3)⊗Ul(4)(cid:17)ρf˜(cid:16)U(j3)⊗Ul(4)(cid:17)T Mi′(cid:19), ρ=||Ψ00iihhΨ0|0,|, ((ii,, jj))=,((11,,11));. (59) Byreasoningsimilarto(56)weconcludethat  u g˜ P(1) P(2) U(3) U(4) ′i i ⊗ k ⊗ j ⊗ l (cid:16) (cid:16) (cid:17)(cid:17) =tr ik ik f˜ U(3) U(4) ρf˜ U(3) U(4) T M (cid:18)| ih |⊗ (cid:16) j ⊗ l (cid:17) (cid:16) j ⊗ l (cid:17) i′(cid:19) =tr ik ik U(3) U(4)ρ U(3) U(4) T M (cid:18)| ih |⊗ j ⊗ l (cid:16) j ⊗ l (cid:17) i′(cid:19) =u P(1) P(2) U(3) U(4) . (60) i i ⊗ k ⊗ j ⊗ l (cid:16) (cid:17) WehavethusprovedthatΓ andΓ areisomorphic. (cid:4) Q ′Q Itisworthnotingthattheconversemaynotbetrue.GivenisomorphicgamesΓ andΓ ,theinputgamesΓandΓ Q ′Q ′ maynotdeterminethestrongisomorphism.Indeed,bimatrixgames l r l r ′ t (3,1) (0,0) t (4,0) (0,0) and ′ (61) b (0,0) (1,3)! b′ (0,0) (0,4)! are notstronglyisomorphic. Howeverthe MW approach(with the initialstate (00 + 11 )/√2) to each one of | i | i (61)impliesthesameoutputgamegivenby 1 σ x 1 (3,1) (0,0) (62) σx (0,0) (1,3)! 5 Conclusions The theory of quantum games does not provide us with clear definitions of how a quantum game should look like. In fact, only one condition is taken into consideration. A quantum game scheme is merely required to generalize the classical game. As a result, this allows us to define a quantum game scheme in many different ways. However, a wide variety of techniques to describe a game in the quantum domain can imply different quantumgameresults. Therefore,itwouldbeconvenienttospecifythatsomequantumschemesworkundersome furtherrestrictions. Wehavebeenworkingundertheassumptionthataquantumschemeisinvariantwithrespect to isomorphic transformationsof an input game. We have shown that this requirementmay be essential tool in definingaquantumscheme.Theprotocolthatreplicatesclassicalcorrelatedequilibriaisanexamplethatdoesnot satisfyourcriterion.Therefineddefinitionforaquantumgameschememayalsobeusefultogeneralizeprotocols. OurworkhasshownthatdependenceoflocalunitaryoperatorsintheMWschemeonthenumberofstrategiesin a classical gameis notlinear. Infact, the generalizedapproachto n m bimatrixgamecanbe identifiedwith a × gameofdimensionn! m!. × Acknowledges ThisworkwassupportedbytheMinistryofScienceandHigherEducationinPolandundertheProjectIuventus PlusIP2014010973intheyears2015–2017. 9 References [1] L.MarinattoandT.Weber,Aquantumapproachtostaticgamesofcompleteinformation,Phys.Lett.A272 291(2000) [2] A.IqbalandA.H.Toor,StabilityofmixedNashequilibriainsymmetricquantumgames,Commun.Theor. Phys.42335(2004) [3] A. Nawaz and A. H. Toor, Evolutionarily stable strategies in quantum Hawk-Dove game, Chinese Phys. Lett.27,050303(2010) [4] P.Fra¸ckiewicz,Quantuminformationapproachtotheultimatumgame,Int.J.Theor.Phys.533248(2014) [5] A.IqbalandA.H.Toor,Backwards-inductionoutcomeinaquantumgame,Phys.Rev.A65,052328(2002) [6] S.Khan,M.RamzanandM.K.Khan,QuantummodelofBertrandduopoly,Chin.Phys.Lett.27,080302 (2010) [7] P.Fra¸ckiewicz,AnewmodelforquantumgamesbasedontheMarinatto–Weberapproach,J.Phys.AMath. Theor.46,275301(2013) [8] A.IqbalandA.H.Toor,QuantummechanicsgivesstabilitytoaNashequilibrium,Phys.Rev.A65,022306 (2002) [9] P.Fra¸ckiewicz,AcommentonthegeneralizationoftheMarinatto-Weberquantumgamescheme,ActaPhys. Pol.B4429(2013) [10] P.Frackiewicz,Anewquantumschemefornormalformgames,QuantumInfProcess141809(2015) [11] J.Gabarró,A.GarcíaandM.Serna,Thecomplexityofgameisomorphism,Theor.Comput.Sci.4126675 (2011) [12] J.Nash,Non-cooperativegames,AnnMath54286(1951) [13] B.Peleg,J.RosenmüllerandP.Sudhölter,TheCanonicalExtensiveFormofaGameForm-PartI-Sym- metriesInCurrentTrendsinEconomics,AdvancementofStudiesinEconomics367(1999) [14] P. Sudhölter, J. Rosenmüller and B. Peleg, The Canonical Extensive Form of a Game Form - Part II - RepresentationJMathEcon33299(2000) [15] P.Fra¸ckiewicz,Onquantumgameapproachtocorrelatedequilibrium,ProceedingsinGlobalVirtualCon- ference,Slovakia,(2016) 10

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.