୯ҥᆵεᏢႝᐒၗૻᏢଣၗૻπำᏢس ᅺγፕЎ
'HSDUWPHQW RI &RPSXWHU 6FLHQFH DQG ,QIRUPDWLRQ (QJLQHHULQJ
&ROOHJH RI (OHFWULFDO (QJLQHHULQJ DQG &RPSXWHU 6FLHQFH
1DWLRQDO 7DLZDQ 8QLYHUVLW\
0DVWHU 7KHVLV
ёԖਏӣᙟӷՍ࣬՟ࡋཛྷ൨ϐᄽᆉݤ
$ 4XHU\(IILFLHQW $OJRULWKP IRU 6WULQJ 6LPLODULW\ 6HDUFK
࠽യ
<RX6KHQJ +XDQJ
ࡰᏤ௲Ǻᇳڷपറγ
$GYLVRU .XQ0DR &KDR 3K'
ύ҇୯ ԃ Д
-XQH
$FNQRZOHGJPHQWV
7R EHJLQ ZLWK , ZRXOG OLNH WR H[SUHVV P\ JUDWLWXGH WR P\ DGYLVRU 3URI
.XQ0DR &KDR +LV JXLGDQFH SOD\V DQ LPSRUWDQW UROH LQ PH ILQLVKLQJ WKLV WKHVLV ,Q DGGLWLRQ WR WKH DFDGHPLF KHOS WKH HQFRXUDJHPHQW IURP KLP LV DQ LPSRUWDQW IDFWRU IRU PH LQ DFKLHYLQJ PDQ\ JRDOV , DOVR ZDQW WR WKDQN P\
H[DPLQDWLRQ FRPPLVVLRQHUV 3URI +XQJ/XQJ :DQJ DQG 3URI <DR7LQJ +XDQJ ZKR SURYLGHG PH ZLWK SUHFLRXV DGYLFH RQ P\ WKHVLV )XUWKHUPRUH
, DSSUHFLDWH DOO DVVLVWDQFH , KDYH UHFHLYHG IURP SDUWQHUV LQ $&% /DE $ERYH DOO , WKDQN P\ IDPLO\ IRU DOO WKHLU VXSSRUW
L
ᄔ ᄔा
ጓᒠຯᚆ HGLW GLVWDQFH ࢂঁቶݱӦҔܭෳໆӷՍϐ໔࣬՟ำࡋ
ޑࡋໆǴԶӷՍ࣬՟ࡋཛྷ൨VWULQJ VLPLODULW\ VHDUFK ߾ाפрӧۓޑ
ӷՍӝύک๏ϒޑ၌ӷՍ TXHU\ VWULQJ ࣬՟ޑӷՍǶаጓᒠຯᚆ
ࣁෳໆ٩ᏵޑӷՍ࣬՟ࡋཛྷ൨ǴεໆӦᔈҔӧӵኧᏵమGDWDEDVH
FOHDQLQJ ǵᒱᇤᔠ׳҅ (UURU GHWHFWLRQ DQG FRUUHFWLRQ ᆶၗᘏڗ GDWD UHWULHYDO ሦୱǶҞԜᜪୢᚒޑှ،БݤεӭӃஒё௨ନޑӷ ՍၸᘠࡕǴӆჹഭΠޑӷՍՉᔠᡍǶฅԶǴ೭٤Бݤӧ௨ନӷՍ ޑၸำύǴ൳Яӄ௦Ҕ࣬ӕޑၸᘠೕ߾Ǵ٬ளਏᒿӷՍӝ ޑׯᡂԶΠफ़ǶࣁΑլܺ೭ঁୢᚒǴךॺගрঁӝόӕၸᘠБݤ
ޑၗ่ᄬGDWD VWUXFWXUH ᡣӷՍޑၸᘠᆢᛙۓޑਏǶךॺ٠ගр
ჹᔈޑᄽᆉݤှ،ٿঁЬाޑӷՍ࣬՟ࡋཛྷ൨ୢᚒǺጄൎ၌UDQJH
TXHU\ ᆶ N ঁ၌ WRSk TXHU\ Ƕჴᡍ่݀ᡉҢߐᘖॶλܭۓޑ ำࡋਔǴךॺޑБݤӧጄൎ၌ڀԖᓬ౦ޑ܄ૈ߄Ƕ
ᜢᗖӷǺጓᒠຯᚆǵӷՍ࣬՟ࡋཛྷ൨ǵጄൎ၌ǵN ঁ၌
LL
$EVWUDFW
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
OHP ZH SURSRVH D GDWD VWUXFWXUH LQWHJUDWLQJ GLIIHUHQW ILOWHULQJ PHWKRGV DQG DGRSWLQJ D PRUH HIILFLHQW RQH RQ HDFK SKDVH :H DOVR JLYH FRUUHVSRQGLQJ DOJRULWKPV IRU WZR LPSRUWDQW TXHULHV RI VWULQJ VLPLODULW\ VHDUFK UDQJH TXHU\
DQG WRSk TXHU\ ([SHULPHQWDO UHVXOWV VKRZ WKDW RXU DSSURDFK LV FRPSHWLWLYH IRU UDQJH TXHU\ ZKHQ WKUHVKROGV DUH VPDOO HQRXJK
.H\ZRUGV HGLW GLVWDQFH VWULQJ VLPLODULW\ VHDUFK UDQJH TXHU\ WRSk TXHU\
LLL
&RQWHQWV
α
α၂ہቩۓਜ L
$FNQRZOHGJPHQWV L
ᄔा LL
$EVWUDFW LLL
&RQWHQWV LY
/LVW RI )LJXUHV YL
/LVW RI 7DEOHV YLL
,QWURGXFWLRQ
5HODWHG :RUN
3UHYLRXV :RUN
9HULI\LQJ $OJRULWKP
3UXQLQJ 7HFKQLTXH
2XU :RUN
3UHOLPLQDULHV
3UHOLPLQDU\ .QRZOHGJH
(GLW 'LVWDQFH
&RPSXWDWLRQV RI (GLW 'LVWDQFH
3UREOHP 'HILQLWLRQ
LY
5DQJH 4XHU\
7RSk 4XHU\
&RPSOH[7UHH %DVHG 6ROXWLRQ
)LOWHULQJ 0HWKRGV
6WULQJ /HQJWK
/HWWHU &RXQW
5HIHUHQFH 6WULQJ
,QGH[ &RQVWUXFWLRQ
6WULQJ /HQJWK
/HWWHU &RXQW
5HIHUHQFH 6WULQJ
4XHU\ $OJRULWKPV
5DQJH 4XHU\
7RSk 4XHU\
([SHULPHQWDO (YDOXDWLRQ
6HWXS
6SDFH &RQVXPSWLRQ
&RPSDULVRQ ZLWK 2WKHU $SSURDFKHV
5DQJH 4XHU\
7RSk 4XHU\
&RQFOXVLRQV DQG )XWXUH :RUN
&RQFOXVLRQV
)XWXUH :RUN
Y
/LVW RI )LJXUHV
6WUXFWXUH H[DPSOH
$YHUDJH UHVSRQVH WLPH IRU UDQJH TXHU\
$YHUDJH UHVSRQVH WLPH IRU WRSk TXHU\
YL
/LVW RI 7DEOHV
1RWDWLRQ
5XQQLQJ H[DPSOH
0DSSLQJ H[DPSOH
5HIHUHQFH VWULQJ FKRRVLQJ H[DPSOH
%RXQGDU\ UHVWULFWLRQ H[DPSOH ZLWK θ = 2
'DWDVHW VWDWLVWLFV
,QGH[ FRQVWUXFWLRQ WLPH LQ VHFRQG
6SDFH FRQVXPSWLRQ IRU LQGH[ FRQVWUXFWLRQ LQ 0%
YLL
&KDSWHU ,QWURGXFWLRQ
6WULQJ VLPLODULW\ VHDUFK LV DQ LPSRUWDQW RSHUDWLRQ XVHG LQ D ZLGH YDULHW\ RI DSSOLFDWLRQV
HJ HUURU FKHFNLQJ GDWD FOHDQLQJ GDWD LQWHJUDWLRQ DQG SDWWHUQ UHFRJQLWLRQ ,W GLVFRYHUV VWULQJV LQ GDWDVHWV WKDW DUH VLPLODU WR TXHU\ VWULQJV DFFRUGLQJ WR D JLYHQ GLVWDQFH PHDVXUH
)RU H[DPSOH LQ HUURU FKHFNLQJ ZKLFK ILQGV SRVVLEOH W\SRV LQ D JLYHQ GRFXPHQW VWULQJ VLPLODULW\ VHDUFK ILQGV ZRUGV QRW LQ D JLYHQ GLFWLRQDU\ DQG UHFRPPHQGV VLPLODU ZRUGV GHWHUPLQHG E\ HGLW GLVWDQFH
9DULRXV GLVWDQFH PHDVXUHV DUH XVHG LQ GLIIHUHQW ILHOGV EDVHG RQ WKHLU FKDUDFWHULVWLFV >
@ DQG HGLW GLVWDQFH LV RQH RI WKH PRVW ZLGHO\ DGRSWHG GLVWDQFH PHDVXUHV LQ UHOHYDQW WRSLFV
,W LV WKHUHIRUH DOVR HPSOR\HG LQ RXU ZRUN
%HFDXVH RI WKH QHHG RI D ERWK WLPH DQG VSDFH HIILFLHQW VROXWLRQ IRU DQVZHULQJ TXHULHV ZLWK VPDOO WKUHVKROGV ZKLFK LV WKH FDVH LQ PDQ\ DSSOLFDWLRQV HJ HUURU FKHFNLQJ GDWD FOHDQLQJ ZH DUH LQWHUHVWHG LQ KRZ WR UHVSRQG WR TXHULHV HIILFLHQWO\ XVLQJ UHDVRQDEOH VSDFH
'XH WR WKH FRVWO\ FRPSOH[LW\ O(|s|θ) RI YHULI\LQJ ZKHWKHU WZR VWULQJV ZLWK OHQJWK s KDYH HGLW GLVWDQFH VPDOOHU WKDQ RU HTXDO WR D JLYHQ WKUHVKROG θ RXU DSSURDFK OLNH PRVW RWKHU H[LVWLQJ RQHV >±@ WULHV WR ILOWHU RXW DV PDQ\ VWULQJV LQ GDWDVHWV DV SRVVLEOH WR PLQLPL]H WKH YHULILFDWLRQ QHHGHG
,Q WKLV WKHVLV D TXHU\HIILFLHQW DOJRULWKP XVLQJ DFFHSWDEOH VSDFH LV SURSRVHG *LYHQ WKDW YDULRXV ILOWHULQJ PHWKRGV RXWSHUIRUP RWKHUV LQ FHUWDLQ GDWDVHWV >@ GLIIHUHQW ILOWHULQJ PHWKRGV DUH LQWHJUDWHG LQ RXU DSSURDFK DQG DGRSWHG RQ GLIIHUHQW SKDVHV UHVSHFWLYHO\ ([
SHULPHQWV RQ UHDO GDWDVHWV VKRZ WKDW RXU DSSURDFK RXWSHUIRUPV RWKHUV ZKHQ WKH WKUHVKROGV
DUH VPDOO HQRXJK
5HODWHG :RUN
6HFWLRQ LQWURGXFHV UHFHQW VWXGLHV UHODWHG WR WKH WRSLF RI VWULQJ VLPLODULW\ VHDUFK
7KH DOJRULWKPV DQG WHFKQLTXHV H[SORLWHG LQ RXU ZRUN DUH H[SODLQHG LQ WKH IROORZLQJ VXE
VHFWLRQV
3UHYLRXV :RUN
(GLW GLVWDQFH ZKLFK LV DOVR UHIHUUHG WR DV /HYHQVKWHLQ GLVWDQFH LV RQH RI WKH UHVHDUFK WRSLFV WKDW KDYH EHHQ VWXGLHG IRU D ORQJ SHULRG RI WLPH $ QXPEHU RI DOJRULWKPV IRU FRP
SXWLQJ HGLW GLVWDQFH KDYH EHHQ SURSRVHG RYHU WKH SDVW IHZ GHFDGHV DQG WKHUH LV XVXDOO\ D WUDGHRII EHWZHHQ WLPH DQG VSDFH FRPSOH[LWLHV $W WKH PRPHQW WR WKH EHVW RI RXU NQRZO
HGJH WKH IDVWHVW DOJRULWKP KDV WLPH FRPSOH[LW\ O(|s|2/log |s|) >@ RQ WKH RWKHU KDQG
WKH PRVW VSDFHHIILFLHQW DOJRULWKP KDV VSDFH FRPSOH[LW\ O(max {|s1|, |s2|}) >@
7KH UHVHDUFK RI VWULQJ VLPLODULW\ VHDUFK LV DQRWKHU WRSLF WKDW KDV JRWWHQ D ORW RI DW
WHQWLRQ ,Q FRQWUDVW ZLWK HGLW GLVWDQFH YDULRXV DOJRULWKPV IRU HIILFLHQWO\ VROYLQJ VWULQJ VLPLODULW\ VHDUFK SUREOHPV DUH VWLOO EHLQJ SURSRVHG DW WKH PRPHQW 0RVW RI WKH DOJRULWKPV GR WKHLU EHVW WR ILOWHU RXW DV PDQ\ VWULQJV LQ WKH GDWDVHWV DV SRVVLEOH VR WKDW WKH YHULILFDWLRQV QHHGHG ZKLFK FRQVXPH PRVW RI WKH WLPH DUH PLQLPL]HG 7KH ILOWHULQJ PHWKRGV XVHG FDQ EH URXJKO\ FDWHJRUL]HG LQWR IROORZLQJ WZR JURXSV
1JUDP %DVHG ,QYHUWHG ,QGH[ >±@ ,Q WKLV DSSURDFK D QXPEHU RI LQYHUWHG OLVWV DUH PDLQWDLQHG IRU DQVZHULQJ VWULQJV LQ GDWDVHWV VKDULQJ FHUWDLQ QXPEHU RI QJUDPV ZLWK WKH TXHU\ 2EYLRXVO\ WKH SHUIRUPDQFH RI PHWKRGV XVLQJ WKLV DSSURDFK PDLQO\ GHSHQGV RQ WKH DOJRULWKPV WKH\ XVH WR PHUJH WKH LQYHUWHG OLVWV $Q HIILFLHQW PHUJLQJ DOJRULWKP LV SURSRVHG LQ >@ $OWKRXJK LW LV HIILFLHQW ZLWK D VPDOO WKUHVKROG LWV HIILFLHQF\ GHJUDGHV ZKHQ TXHU\
VWULQJV RU VWULQJV LQ GDWDVHWV DUH VKRUW $OVR LW VXIIHUV IURP VHYHUH VSDFH RYHUKHDG
%+WUHH >@ ,W LV ILUVW DGYDQFHG E\ =KDQJ HW DO >@ $Q RUGHU RI VWULQJV LQ GDWDVHWV LV GHILQHG EHIRUHKDQG ZKLFK HQDEOHV WKH LQGH[HV WR EH VWRUHG LQ D %+WUHH 7KH %+WUHH
LV WKHQ XVHG E\ GLIIHUHQW DOJRULWKPV WR DQVZHU TXHULHV $QRWKHU HIILFLHQW VWULQJ RUGHU DQG DOJRULWKP LV SURSRVHG E\ /X HW DO >@ 7KHVH DSSURDFKHV FDQ XVXDOO\ KDYH JRRG SHUIRU
PDQFH KRZHYHU LW VXIIHUV IURP VHDUFK RYHUKHDG ZKHQ LQDSSURSULDWH VWULQJ RUGHUV DUH FKRVHQ ZKLFK YDU\ IURP RQH GDWDVHW WR DQRWKHU
9HULI\LQJ $OJRULWKP
9HULILFDWLRQ ZKLFK GHWHUPLQHV ZKHWKHU WZR VWULQJV KDYH HGLW GLVWDQFH VPDOOHU WKDQ RU HTXDO WR D JLYHQ WKUHVKROG LV DQ LQWHQVHO\ XVHG RSHUDWLRQ LQ VWULQJ VLPLODULW\ VHDUFK ,Q
VWHDG RI FRPSXWLQJ WKH DFWXDO HGLW GLVWDQFH EHWZHHQ WZR VWULQJV D PRUH HIILFLHQW DSSURDFK
$OJRULWKP 9HULI\(' LV SURSRVHG LQ >@
$OJRULWKP 9HULI\(' VWULQJ s1 VWULQJ s2 GLVWDQFH WKUHVKROG θ LI ||s1| − |s2|| > θ WKHQ
UHWXUQ )$/6(
HQG LI
&RQVWUXFW D WDEOH T RI2 URZV DQG |s2| + 1 FROXPQV IRU j = 1 WR min{|s2| + 1, 1 + θ} GR
T[1][j] = j − 1 HQG IRU
6HW m= θ + 1
IRU i = 2 WR |s1| + 1 GR
IRU j min{1, i − θ} WR min{|s2| + 1, i + θ` GR d1 = (j < i + θ) ? T [1][j] + 1 : θ + 1 d2 = (j > 1) ? T [2][j − 1] + 1 : θ + 1
d3 = (j > 1) ? T [1][j − 1] + (s1[i − 1] = s2[j − 1]) ? 0 : 1) : θ + 1 T[2][j] = min{d1, d2, d3}
m = min{m, T [2][j]}
HQG IRU LI m > θ WKHQ
UHWXUQ )$/6(
HQG LI
IRU j = 0 WR |s2| + 1 GR T[1][j] = T [2][j]
HQG IRU HQG IRU UHWXUQ 758(
7KH LGHD RI $OJRULWKP LV WKDW RQO\ WKH HQWULHV RI WKH G\QDPLF SURJUDPPLQJ WDEOH RQ WKH GLDJRQDO ZLWK RIIVHW QR ODUJHU WKDQ WKH WKUHVKROG QHHG WHVWLQJ ,W LV HDV\ WR REVHUYH WKDW HQWULHV QRW WHVWHG OHDGV WR D ODUJHU HGLW GLVWDQFH WKDQ θ WKHUHIRUH WKH\ DUH VLPSO\ VNLSSHG
%\ WKLV DOJRULWKP D YHULILFDWLRQ RSHUDWLRQ FDQ EH ILQLVKHG LQ O(|s|θ) ZKHUH s LV WKH VKRUWHU VWULQJ EHLQJ YHULILHG
3UXQLQJ 7HFKQLTXH
$ SDUWLWLRQ EDVHG SUXQLQJ VROXWLRQ LV SURSRVHG LQ >@ 6WULQJV LQ GDWDVHWV DUH VSOLW LQWR VHYHUDO SDUWLWLRQV (DFK SDUWLWLRQ Pi LV FKDUDFWHUL]HG E\ D UHSUHVHQWDWLYH UHIHUHQFH VWULQJ oi 7KH PLQLPXP HGLW GLVWDQFH Pi.l DQG PD[LPXP HGLW GLVWDQFH Pi.u EHWZHHQ oi DQG DQ\ VWULQJV LQ Pi LV PDLQWDLQHG )RUPDOO\ Pi.l = PLQ{ed(s, oi)|s ∈ Pi} Pi.u = PD[{ed(s, oi)|s ∈ Pi} ZKHUH ed(s, oi) LV WKH HGLW GLVWDQFH EHWZHHQ s DQG oi ,Q WKH UHVW RI WKLV WKHVH ZH XVH ed(s1, s2) WR GHQRWH WKH HGLW GLVWDQFH EHWZHHQ s1 DQG s2 :KHQ D TXHU\
LV LVVXHG SDUWLWLRQV DUH SUXQHG DFFRUGLQJ WR WKH IROORZLQJ UXOHV
*LYHQ D UDQJH TXHU\ ZLWK WKUHVKROG θ DQG TXHU\ VWULQJ q D SDUWLWLRQ P LV D SUXQDEOH SDUWLWLRQ LI DQG RQO\ LI RQH RI WKH IROORZLQJ FRQGLWLRQV KROGV ed(o, q) − P.u > θ RU P.l− ed(o, q) > θ
3UXQLQJ 5XOH ,I SDUWLWLRQ P LV D SUXQDEOH SDUWLWLRQ WKHQ∀s ∈ P, ed(s, q) > θ 7KXV
VWULQJV WKDW OLH LQ P GR QRW KDYH WR EH YHULILHG DQG FDQ EH SUXQHG GLUHFWO\
*LYHQ D UDQJH TXHU\ ZLWK WKUHVKROG θ DQG TXHU\ VWULQJ q D SDUWLWLRQ P LV D FDQGLGDWH SDUWLWLRQ LI DQG RQO\ LI RQH RI WKH IROORZLQJ FRQGLWLRQV KROGV P.l ≤ ed(o, q) ≤ P.u
RU P.l− θ ≤ ed(o, q) ≤ P.l RU P.u ≤ ed(o, q) ≤ P.u + θ
3UXQLQJ 5XOH 6XSSRVH P LV D FDQGLGDWH SDUWLWLRQ 7KHQ∀s ∈ P s QHHGV WR EH YHULILHG LI DQG RQO\ LI
lb≤ ed(o, s) ≤ ub
ZKHUH lb = max{ed(o, q) − θ, P.l} ub = min{ed(o, q) + θ, P.u} :H UHIHU WR UDQJH [lb, ub] DV FDQGLGDWH UHJLRQ
*LYHQ D UDQJH TXHU\ ZLWK WKUHVKROG θ DQG TXHU\ VWULQJ q D SDUWLWLRQ P LV D VHOHFWDEOH SDUWLWLRQ LI DQG RQO\ LI WKH IROORZLQJ FRQGLWLRQ KROGV ed(o, q) + P.u ≤ θ
3UXQLQJ 5XOH 6XSSRVH WKDW SDUWLWLRQ P LV D VHOHFWDEOH SDUWLWLRQ 7KHQ ∀s ∈ P ed(s, q) ≤ θ DQG s LV UHSRUWHG DV D UHVXOW :H UHIHU WR UDQJH [P.l, P.u] DV VHOHFWDEOH UHJLRQ
,I D SDUWLWLRQ LV SUXQDEOH RU VHOHFWDEOH QR IXUWKHU YHULILFDWLRQ RQ WKH VWULQJV LQ WKH SDUWLWLRQ LV UHTXLUHG ,I WKH SDUWLWLRQ LV D FDQGLGDWH SDUWLWLRQ WKHQ RQO\ VWULQJV WKDW OLH LQ WKH FDQGLGDWH UHJLRQ QHHG WR EH YHULILHG
2XU :RUN
,Q WKLV WKHVLV ZH LQWURGXFH D QHZ DSSURDFK WR VWULQJ VLPLODULW\ VHDUFK 2XU DSSURDFK LV EDVHG RQ WKH IDFW WKDW GLIIHUHQW ILOWHULQJ PHWKRGV RXWSHUIRUP RWKHUV LQ FHUWDLQ GDWDVHWV
5RXJKHU ILOWHULQJ PHWKRGV DUH XVHG ILUVW ZKLFK JUHDWO\ UHGXFHV WKH FDQGLGDWH VWULQJV LQ IHZ VWHSV 7KH PRUH SUHFLVH ILOWHULQJ PHWKRGV DUH WKHQ DGRSWHG WR GLPLQLVK WKH FDQGLGDWH VWULQJV IXUWKHU
,Q WKLV WKHVLV ZH PDNH IROORZLQJ FRQWULEXWLRQV
:H SURSRVH DQ LQGH[LQJ VWUXFWXUH WKDW LQWHJUDWHV VHYHUDO ILOWHULQJ PHWKRGV
:H SURSRVH DOJRULWKPV IRU VWULQJ VLPLODULW\ VHDUFK EDVHG RQ WKH VWUXFWXUH
$ WUDGHRII EHWZHHQ TXHU\ HIILFLHQF\ DQG PHPRU\ FRQVXPSWLRQ LV SURYLGHG ZKLFK PDNHV LW IHDVLEOH WR GHDO ZLWK WUHPHQGRXV GDWDVHWV
:H FRQGXFW H[SHULPHQWV WR HYDOXDWH WKH SHUIRUPDQFH RI RXU DSSURDFK DQG FRPSDUH LW ZLWK RWKHUV
7KH UHVW RI WKLV WKHVLV LV VWUXFWXUHG DV IROORZV &KDSWHU JLYHV WKH EDFNJURXQG DQG IRU
PDO GHILQLWLRQV RI VWULQJ VLPLODULW\ VHDUFK SUREOHP &KDSWHU JLYHV GHWDLOV RI WKH DSSURDFK ZH SURSRVH &KDSWHU SUHVHQWV WKH H[SHULPHQWV ZH FRQGXFWHG DQG &KDSWHU FRQFOXGHV WKLV WKHVLV
&KDSWHU
3UHOLPLQDULHV
,Q WKLV FKDSWHU VRPH SUHOLPLQDU\ NQRZOHGJH DQG SUREOHP GHILQLWLRQV DUH SUHVHQWHG
,Q WKH UHVW RI WKLV WKHVLV ZH DVVXPHΣ D ILQLWH DOSKDEHW )RU VWULQJ s WKH OHQJWK RI LW LV GHQRWHG E\|s| DQG WKH iWK OHWWHU LV GHQRWHG E\ s[i] ZKHUH s[i] ∈ Σ DQG 1 ≤ i ≤ |s| 7KH QRWDWLRQ XVHG LQ WKH UHVW RI WKLV WKHVLV DUH VXPPDUL]HG LQ 7DEOH
7DEOH 1RWDWLRQ
1RWDWLRQ ([SODQDWLRQ Σ WKH DOSKDEHW
s D VWULQJ FRQVLVWV RI OHWWHUV LQΣ
|s| WKH OHQJWK RI VWULQJ s s[i] WKH iWK OHWWHU RI VWULQJ s
s[i, j] WKH VXEVWULQJ RI s IURP WKH iWK OHWWHU WR WKH jWK OHWWHU q D TXHU\ VWULQJ
θ D WKUHVKROG RI HGLW GLVWDQFH ed(q, s) WKH HGLW GLVWDQFH EHWZHHQ q DQG s
c(s, l) WKH QXPEHU RI RFFXUUHQFHV RI OHWWHU l LQ s h(n) WKH KHLJKW RI WKH WUHH FRQWDLQLQJ QRGH n d(n) WKH OHYHO RI QRGH n
v(n) WKH YDOXH RI QRGH n
l(n) WKH OHWWHU LPSOLHG E\ QRGHV ZLWK OHYHO HTXDO WR n
p(n, k) WKH DQFHVWRU RI n ZLWK OHYHO HTXDO WR k
3UHOLPLQDU\ .QRZOHGJH
(GLW 'LVWDQFH
7KHUH DUH WKUHH HGLW RSHUDWLRQV RQ VWULQJ s LQVHUWLRQ GHOHWLRQ DQG VXEVWLWXWLRQ $Q LQVHUWLRQ RSHUDWLRQ LQVHUWV D FKDUDFWHU x∈ Σ LQWR s ZKLFK IRUPV D QHZ VWULQJ s s[1, i−1]
x s[i, |s|] ZLWK OHQJWK |s|+1 $ GHOHWLRQ RSHUDWLRQ UHPRYHV D OHWWHU s[i] IURP s ZKLFK IRUPV D QHZ VWULQJ s s[1, i−1] s[i+1, |s|] ZLWK OHQJWK |s|−1 $ VXEVWLWXWLRQ RSHUDWLRQ UHSODFHV D OHWWHU s[i] ZLWK D FKDUDFWHU x ∈ Σ ZKLFK IRUPV D QHZ VWULQJ s s[1, i − 1] x s[i + 1, |s|]
ZLWK OHQJWK|s| 7KH HGLW GLVWDQFH LV GHILQHG DV IROORZ
'HILQLWLRQ (GLW 'LVWDQFH *LYHQ WZR VWULQJV si DQG sj WKH HGLW GLVWDQFH EHWZHHQ si DQG sj GHQRWHG E\ ed(si, sj) LV GHILQHG DV WKH PLQLPXP QXPEHU RI HGLW RSHUDWLRQV QHHGHG WR WUDQVIRUP si WR sj
&RPSXWDWLRQV RI (GLW 'LVWDQFH
%HVLGHV 9HULILFDWLRQ RI ZKLFK $OJRULWKP 9HULI\(' JLYHV DQ HIILFLHQW VROXWLRQ WKH FRPSXWDWLRQ RI H[DFW HGLW GLVWDQFHV LV DOVR QHHGHG LQ RXU DSSURDFK IRU ERWK FRQVWUXFWLQJ LQGH[LQJ VWUXFWXUHV DQG DQVZHULQJ TXHULHV ,W FDQ EH FRPSXWHG XVLQJ G\QDPLF SURJUDP
PLQJ $OJRULWKP JLYHV GHWDLOV RI WKH FRPSXWDWLRQ XVLQJ G\QDPLF SURJUDPPLQJ
$OJRULWKP (' VWULQJ s1 VWULQJ s2 LQW lenBs1 LQW lenBs2 LI lenBs1 == 0 WKHQ
UHWXUQ lenBs2 HQG LI
LI lenBs2 == 0 WKHQ UHWXUQ lenBs1 HQG LI
LI s1[lenBs1− 1] == s2[lenBs2− 1] WKHQ cost= 0
HOVH
cost= 1 HQG LI
UHWXUQ PLQLPXP ('s1 s2 lenBs1− 1 lenBs2+1
('s1 s2 lenBs1 lenBs2− 1+1
('s1 s2 lenBs1− 1 lenBs2− 1+cost
3UREOHP 'HILQLWLRQ
:H FDQ WKHQ JLYH WKH GHILQLWLRQV RI VWULQJ VLPLODULW\ TXHULHV EDVHG RQ HGLW GLVWDQFH
6HFWLRQ DQG JLYH IRUPDO GHILQLWLRQV RI UDQJH TXHU\ DQG WRSk TXHU\ ZKLFK DUH WKH PRVW IUHTXHQWO\ FRQVLGHUHG TXHULHV UHVSHFWLYHO\
5DQJH 4XHU\
5DQJH TXHU\ LV RQH RI WKH PRVW LPSRUWDQW TXHU\ W\SHV RI VWULQJ VLPLODULW\ ,Q PDQ\
FDVHV LW UHWXUQV PRUH PHDQLQJIXO UHVXOWV WKDQ WRSk TXHU\ ZKLFK VRPHWLPHV UHWXUQV UHVXOWV WKDW GHYLDWH IURP TXHU\ VWULQJV WRR PXFK
'HILQLWLRQ 5DQJH 4XHU\ *LYHQ D TXHU\ VWULQJ q DQG D VWULQJ VHW S = {s1, s2, ..., s|S|}
D UDQJH TXHU\ UHWXUQV D VXEVHW S RI S ZLWK HGLW GLVWDQFH WR q QR ODUJHU WKDQ θ LH
S = {si ∈ S|ed(si, q) ≤ θ}
7RSk 4XHU\
7RSk TXHU\ DQRWKHU LPSRUWDQW TXHU\ W\SH RI VWULQJ VLPLODULW\ UHWXUQV k VWULQJV WKDW DUH WKH PRVW VLPLODU WR TXHU\ VWULQJV
'HILQLWLRQ 7RSk 4XHU\ *LYHQ D TXHU\ VWULQJ q DQG D VWULQJ VHW S = {s1, s2, ..., s|S|}
D 7RSk TXHU\ UHWXUQV D VXEVHW S RI k VWULQJV LQ S ZLWK HGLW GLVWDQFH WR q QR ODUJHU WKDQ WKDW RI DQ\ RWKHU VWULQJ LQ S− S
&KDSWHU
&RPSOH[7UHH %DVHG 6ROXWLRQ
7KH LGHD RI RXU DSSURDFK LV WR DGRSW GLIIHUHQW ILOWHULQJ PHWKRGV RQ HDFK SKDVH ZKLFK RSWLPL]HV WKH SRZHU RI SUXQLQJ VWULQJV :H PDNH HIIRUWV RQ FKRRVLQJ WKH ILOWHULQJ PHWKRGV XVHG RQ HDFK SKDVH DQG KRZ WR LQWHJUDWH WKHP LQWR RXU VWUXFWXUH 6HFWLRQ JLYHV WKH SURSHUWLHV RI ILOWHULQJ PHWKRGV XVHG :H WKHQ GHPRQVWUDWH KRZ WR EXLOG D FRPSOH[WUHH LQWHJUDWLQJ WKHVH ILOWHULQJ PHWKRGV LQ 6HFWLRQ )LQDOO\ ZH SUHVHQW HIILFLHQW DOJRULWKPV IRU DQVZHULQJ ERWK UDQJH DQG WRSk TXHULHV XVLQJ FRPSOH[WUHH LQ 6HFWLRQ
)LOWHULQJ 0HWKRGV
:H SUHVHQW WKUHH ILOWHULQJ PHWKRGV LQ WKLV VHFWLRQ ZKLFK DUH VWULQJ OHQJWK OHWWHU FRXQW DQG UHIHUHQFH VWULQJ UHVSHFWLYHO\
7DEOH 5XQQLQJ H[DPSOH
6WULQJ 6WULQJ &RQWHQW |s| g1 g2 g3
q /L =RQJ\R
s1 /L =RQJ\RQJ
s2 /L =RX
s3 /LX =RQJWLDQ
s4 /LX =RQJ\X
s5 ;L =RQJ\XH
s6 ;L =ROH\XH
s7 ;LQJ =RX[O
6WULQJ /HQJWK
6WULQJ OHQJWK LV RQH RI WKH PRVW LQWXLWLYH ZD\V WR ILOHU RXW VWULQJV 3URSHUW\ JLYHV DQ DFFXUDWH ZD\ WR GR LW
3URSHUW\ 6WULQJ /HQJWK 3UXQLQJ *LYHQ D UDQJH TXHU\ ZLWK TXHU\ VWULQJ q DQG WKUHVK
ROG θ VWULQJV LQ WKH GDWDVHW ZLWK OHQJWK ODUJHU WKDQ|q|+θ RU OHVV WKDQ |q|−θ DUH SUXQDEOH
&RQVLGHU WKH VDPSOH TXHU\ VWULQJ DQG GDWDVHW VKRZQ LQ 7DEOH *LYHQ WKDW WKH WKUHVK
ROG LV ERWK VWULQJ s2 DQG s3 DUH SUXQDEOH GXH WR WKH H[FHVV RI|q| + θ RU WKH LQVXIIL
FLHQF\ RI|q| − θ LQ OHQJWK
/HWWHU &RXQW
/HWWHU FRXQW LV DQRWKHU ILOWHULQJ PHWKRG WKDW LV LQWXLWLYH EXW HIILFLHQW 3URSHUW\ JLYHV D VLPSOH VROXWLRQ KRZHYHU LW LV UHILQHG IXUWKHU ZKHQ ZH LQWHJUDWH LW LQWR RXU VWUXFWXUH
3URSHUW\ /HWWHU &RXQW 3UXQLQJ *LYHQ D UDQJH TXHU\ ZLWK TXHU\ VWULQJ q DQG WKUHVK
ROG θ )RU VWULQJ s OHW bl EH WKH VXP RI HYHU\ GLIIHUHQFH EHWZHHQ c(s, ai) DQG c(q, ai)
ZKHUH ai LV HDFK OHWWHU LQ s VDWLVI\LQJ WKDW c(s, ai) > c(q, ai) LH bl = c(s, ai) − c(q, ai)|ai ∈ s, c(s, ai) > c(q, ai) DQG OHW br EH WKH VXP RI HYHU\ GLIIHUHQFH EHWZHHQ c(q, aj) DQG c(s, aj) ZKHUH aj LV HDFK OHWWHU LQ q VDWLVI\LQJ WKDW c(q, aj) > c(s, aj) LH
br =c(q, aj) − c(s, aj)|aj ∈ q, c(q, aj) > c(s, aj) ,I bl> θRU br > θ s LV SUXQDEOH
&RQVLGHU WKH VDPSOH TXHU\ VWULQJ DQG GDWDVHW VKRZQ LQ 7DEOH *LYHQ WKDW WKH WKUHVK
ROG LV VWULQJ s2 ZLWK br = 4 s3ZLWK bl = 5 s5 ZLWK bl = 3 s6 ZLWK bl = 5 DQG s7 ZLWK bl = 4 DUH SUXQDEOH
5HIHUHQFH 6WULQJ
6LPLODU WR WKH WHFKQLTXH PHQWLRQHG LQ 6HFWLRQ ZH FDQ ILOWHU VWULQJV RXW EDVHG RQ WKHLU HGLW GLVWDQFH WR DQRWKHU VSHFLILF UHIHUHQFH VWULQJ
3URSHUW\ 5HIHUHQFH 6WULQJ 3UXQLQJ *LYHQ D UHIHUHQFH VWULQJ r DQG D UDQJH TXHU\
ZLWK TXHU\ VWULQJ q DQG WKUHVKROG θ LI VWULQJ s VDWLVILHV WKDW|ed(s, r) − ed(r, q)| > θ s LV SUXQDEOH
&RQVLGHU WKH VDPSOH TXHU\ VWULQJ DQG GDWDVHW VKRZQ LQ 7DEOH *LYHQ WKDW WKH WKUHVK
ROG LV DQG WKH UHIHUHQFH VWULQJ LV s4 ERWK VWULQJ s4DQG s7DUH SUXQDEOH
,QGH[ &RQVWUXFWLRQ
2QFH WKH ILOWHULQJ PHWKRGV XVHG DUH GHWHUPLQHG ZH FDQ EXLOG FRPSOH[WUHHV EDVHG RQ WKHVH ILOWHULQJ PHWKRGV 8QOLNH JHQHUDO VHDUFK WUHHV D FRPSOH[WUHH KDV GLIIHUHQW FULWHULD IRU WUDYHUVLQJ RQ GLIIHUHQW OHYHOV ZKLFK DUH VWULQJ OHQJWK OHWWHU FRXQW DQG WKH HGLW GLV
WDQFH WR D VSHFLILF UHIHUHQFH VWULQJ UHVSHFWLYHO\ $OJRULWKP DQG $OJRULWKP GHVFULEH WKH GHWDLOV RI FRQVWUXFWLRQ 7R FKRRVH WKH GHVLUHG UHIHUHQFH VWULQJ $OJRULWKP VKRXOG EH IROORZHG E\ $OJRULWKP
$OJRULWKP &RQVWUXFWLRQ VWULQJ s &RPSOH[WUHH QRGH N DOSKDEHW VL]H t LI d(N) == 0 WKHQ
LI N KDV QR FKLOG nc VXFK WKDW v(nc) = |s| WKHQ
$GG QRGH nc ZLWK YDOXH= |s| LQWR N¶V FKLOGUHQ
HQG LI
&RQVWUXFWLRQs nc HQG LI
LI 1 ≤ d(N) ≤ t − 1 WKHQ
LI N KDV QR FKLOG nc VXFK WKDW v(nc) = c(s, l(d(N))) WKHQ
$GG QRGH nc ZLWK YDOXH= c(s, l(d(N))) LQWR N¶V FKLOGUHQ
HQG LI
&RQVWUXFWLRQs nc HQG LI
LI d(N) == t WKHQ
$GG QRGH nc ZLWK YDOXH= s LQWR N¶V FKLOGUHQ
HQG LI
6WULQJ /HQJWK
6WULQJV LQ GDWDVHWV DUH ILUVW FDWHJRUL]HG E\ WKHLU OHQJWK 7KDW LV IRU D FRPSOH[WUHH LI QRGH nc VDWLVILHV WKDW d(nc) = 1 WKH VWULQJV LQGH[HG E\ GHVFHQGDQWV RI d(nc) KDYH OHQJWK HTXDO WR v(nc)
$OJRULWKP &KRRVLQJ5HIHUHQFH &RPSOH[WUHH QRGH N DOSKDEHW VL]H t LI d(N)! = t WKHQ
IRU HDFK nc ∈ FKLOGUHQ RI N GR
&KRRVLQJ5HIHUHQFHnc HQG IRU
HOVH
)LQG WKH FKLOG RI N nc WKDW v(nc) PD[LPL]HV WKH VWDQGDUG GHYLDWLRQ RI WKH HGLW GLVWDQFH EHWZHHQ LW DQG DOO RWKHU VLEOLQJV
IRU HDFK nd ∈ FKLOGUHQ RI N GR
LI nc KDV QR FKLOG neVXFK WKDW v(ne) = ed(nc, nd) WKHQ
$GG QRGH neZLWK YDOXH= ed(nc, nd) LQWR nc¶V FKLOGUHQ
HQG LI
$GG D QRGH ZLWK YDOXH= ndLQWR ne¶V FKLOGUHQ
LI nd! = ncWKHQ 'HOHWH QRGH nd HQG LI
HQG IRU HQG LI
/HWWHU &RXQW
6WULQJV DUH WKHQ FDWHJRUL]HG E\ HDFK DPRXQW RI GLIIHUHQW OHWWHUV WKH\ FRQWDLQV 7KDW LV
IRU D FRPSOH[WUHH LI QRGH nc VDWLVILHV WKDW2 ≤ d(nc) ≤ h(n) − 3 OHWWHU l(d(nc)) DSSHDUV v(nc) WLPHV LQ HDFK VWULQJ LQGH[HG E\ GHVFHQGDQWV RI nc
:H PD\ LQWXLWLYHO\ HQXPHUDWH DOO OHWWHUV XVHG LQ GDWDVHWV ZKHQ FRQVWUXFWLQJ FRPSOH[
WUHHV +RZHYHU LW VXIIHUV IURP VHYHUH PHPRU\ FRQVXPSWLRQ ZKLFK LV WRR H[SHQVLYH IRU PDQ\ PDFKLQHV 7R HDVH WKLV VLWXDWLRQ ZH XVH D KDVK IXQFWLRQ WR PDS WKH DOSKDEHW LQWR D VPDOOHU XQLYHUVH ZKLFK LV YDULDEOH LQ VL]H WR ILW LQ GLIIHUHQW PHPRU\ FDSDFLW\
&RQVLGHU WKH VDPSOH GDWDVHW VKRZQ LQ 7DEOH DQG WKH PDSSLQJ IXQFWLRQ VKRZQ LQ 7DEOH *LYHQ WKDW l(2) = g1 l(3) = g2DQG l(4) = g3 EHFDXVH s1FRQVLVWV RI WZR OHWWHUV PDSSHG WR g1 ILYH OHWWHUV PDSSHG WR g2DQG ILYH OHWWHUV PDSSHG WR g3 WKH DQFHVWRUV RI WKH QRGH LQGH[LQJ s1 RQ OHYHO DQG VKRXOG KDYH YDOXHV DQG UHVSHFWLYHO\
+RZHYHU WKHUH LV QR QHHG WR UHFRUG WKH OHWWHU FRXQW RI WKH ODVW JURXS ZKLFK LV LQIHUDEOH E\ GHGXFWLQJ DOO OHWWHU FRXQWV RI RWKHU JURXSV IURP WKH VWULQJ OHQJWK $V ZH FDQ VHH LQ WKH VWUXFWXUH H[DPSOH VKRZQ LQ )LJXUH ZKLFK LV FRQVWUXFWHG DFFRUGLQJ WR WKH GDWDVHW VKRZQ LQ 7DEOH DQG WKH PDSSLQJ IXQFWLRQ VKRZQ LQ 7DEOH WKH OHWWHU FRXQW RI g3 ZKLFK FDQ EH FRPSXWHG E\ GHGXFWLQJ WKH YDOXHV RI FRUUHVSRQGLQJ QRGHV RQ OHYHO DQG
7DEOH 0DSSLQJ H[DPSOH
*URXS /HWWHU
g1
L W X [
\
;
g2
D J Q VSDFH
g3
H O R ] /
=
)LJXUH 6WUXFWXUH H[DPSOH
OHYHO IURP WKH YDOXH RI FRUUHVSRQGLQJ QRGH RQ OHYHO LV QRW UHFRUGHG
5HIHUHQFH 6WULQJ
7DEOH 5HIHUHQFH VWULQJ FKRRVLQJ H[DPSOH
6WULQJ ed(x, s4) ed(x, s5) ed(x, s7) VWDQGDUG GHYLDWLRQ
s4
s5
s7
$IWHU DERYH FDWHJRUL]DWLRQV DUH ILQLVKHG ZH WKHQ FKRRVH D VWULQJ LQ HDFK VHW DV UHIHU
HQFH VWULQJ 7KH VWULQJ WKDW PDNHV WKH HGLW GLVWDQFHV EHWZHHQ LW DQG DOO RWKHU VWULQJV KDYH WKH ODUJHVW VWDQGDUG GHYLDWLRQ LV FKRVHQ
&RQVLGHU WKH VWULQJV FRPSXWHG LQ 7DEOH 7KH HGLW GLVWDQFH EHWZHHQ HDFK VWULQJ DQG WKH RWKHUV DUH FRPSXWHG ILUVW DQG WKH VWDQGDUG GHYLDWLRQ RI WKHP DUH WKHQ FDOFXODWHG 7KH VWULQJ PDNLQJ WKH ODUJHVW VWDQGDUG GHYLDWLRQ ZKLFK LV s7 LQ WKH H[DPSOH LV FKRVHQ DV WKH UHIHUHQFH VWULQJ
2QFH WKH UHIHUHQFH VWULQJ LV GHFLGHG DOO VWULQJV DUH WKHQ FDWHJRUL]HG E\ WKH HGLW GLVWDQFH EHWZHHQ LW DQG WKH UHIHUHQFH VWULQJ
$V ZH FDQ VHH LQ )LJXUH RQFH s7 LV FKRVHQ DV WKH UHIHUHQFH VWULQJ s4 DQG s5 DUH WKHQ LQGH[HG E\ LWV FKLOG ZLWK YDOXH7 $OVR s7 LV LQGH[HG E\ LWV FKLOG ZLWK YDOXH0
4XHU\ $OJRULWKPV
2QFH WKH VWUXFWXUH LV FRQVWUXFWHG ZH FDQ HIILFLHQWO\ UHVSRQG WR TXHULHV ZLWK WKH VWUXF
WXUH
5DQJH 4XHU\
$OJRULWKP JLYHV WKH GHWDLOV RI KRZ WR DQVZHU UDQJH TXHULHV ZLWK D FRPSOH[WUHH $V WKH SURFHGXUH RI EXLOGLQJ WKH LQGH[ HYHU\ TXHU\ LV SURFHVVHG ZLWK WKUHH SKDVHV ZKLFK DUH VWULQJ OHQJWK OHWWHU FRXQW DQG UHIHUHQFH VWULQJ UHVSHFWLYHO\
$OJRULWKP 5DQJH4XHU\ VWULQJ q WKUHVKROG θ &RPSOH[WUHH QRGH N LQW bl LQW br LI d(N) == 0 WKHQ
IRU HDFK nc ∈ FKLOGUHQ RI N GR LI |q| − θ ≤ v(nc) ≤ |q| + θ WKHQ
5DQJH4XHU\q θ ncmin{θ, θ − (v(nc) − |q|)} min{θ, θ − (|q| − v(nc))}
HQG LI HQG IRU HQG LI
LI 1 ≤ d(N) ≤ h(N) − 3 WKHQ IRU HDFK nc ∈ FKLOGUHQ RI N GR
LI c(q, l(d(N))) − bl≤ v(nc) ≤ c(q, l(d(N))) + br WKHQ
5DQJH4XHU\q θ ncmin{bl, bl− (c(q, l(d(n))) − v(nc))}
min{br, br− (v(nc) − c(q, l(d(n))))}
HQG LI HQG IRU HQG LI
LI d(N) == h(n) − 2 WKHQ
IRU HDFK nc ∈ FKLOGUHQ RI N GR
LI ED(q, v(N), |q|, |v(n)|) − θ ≤ v(nc) ≤ ED(q, v(N), |q|, |v(n)|) + θ WKHQ IRU HDFK nd∈ FKLOGUHQ RI nc GR
LI 9HULI\('q, v(nd), θ WKHQ
$GG v(nd) LQWR TXHU\ UHVXOW HQG LI
HQG IRU HQG LI HQG IRU HQG LI
)RU TXHU\ VWULQJ q ZH ILUVW WUDYHUVH WKH VXEWUHHV ZLWK VWULQJV LQGH[HG KDYLQJ OHQJWK EHWZHHQ|q| − θ DQG |q| + θ EDVHG RQ SURSHUW\
+RZHYHU ZH FDQ UHVWULFW WKH QXPEHU RI VXEWUHH WUDYHUVHG IXUWKHU E\ GHQRWLQJ WKH OHIW DQG ULJKW ERXQGDU\ RI VXEWUHHV WR WUDYHUVH LQ HDFK LWHUDWLRQ
3URSHUW\ %RXQGDU\ 5HVWULFWLRQ *LYHQ D UDQJH TXHU\ ZLWK TXHU\ VWULQJ q DQG WKUHVK
ROG θ LI VWULQJ s LV D UHVXOW VWULQJ s ODFNV DW PRVWmin{θ, θ − (|s| − |q|)} OHWWHUV LQ q DQG KDV DW PRVWmin{θ, θ − (|q| − |s|)} DGGLWLRQDO OHWWHUV EHVLGHV WKRVH LQ q
*LYHQ WKH TXHU\ ZLWK TXHU\ VWULQJ q /HH /LH DQG θ :LWKRXW ORVV RI JHQHUDOLW\
FRQVLGHU WKH OHWWHU FRXQW RI OHWWHU µH¶ 7DEOH OLVWV DOO SRVVLEOH FRPELQDWLRQ RI|s| DQG c(s,H) :H FDQ ILQG WKDW WKHUH LV QR UHVXOW VWULQJ s FRQWDLQLQJ OHVV WKDQ min{2, 2−(|s|−7)}
H RU PRUH WKDQmin{2, 2 − (7 − |s|)} H
7KH DOJRULWKP WKHQ GHWHUPLQHV WKH VXEWUHHV WR WUDYHUVH EDVHG RQ HDFK OHWWHU FRXQW RI
7DEOH %RXQGDU\ UHVWULFWLRQ H[DPSOH ZLWK θ = 2
|s| c(s,H) s
/HH /LH TXHU\ VWULQJ
/H /L / /LH
/HH / /H /H
/HH H HH /H
/H /LD /H /LE
/HH /D /HH /E
/HH /H /HH $H
/HH HH HH /HH
/HD /LD /DH /LD
/HH /DD /HH /DE
/HH /LH /HH /DH
/HH /HH /HH DHH
/HH HHH /HHH/HH
/HH /LDD /HH /LDE
/HH /LHD /HH /DHD
/HH /HHD /HH /HHD
/HH /HHH /HHH/LHH
/HH /LHDD /HH /LDHD
/HH /LHHD /HH /LHDH
/HH /HHHD /HHH /LHH
qDQG WKH SDUDPHWHU bl DQG br ZKLFK GHQRWHV WKH OHIW ERXQG DQG ULJKW ERXQG UHVSHFWLYHO\
(YHU\ FKLOGUHQ nc RI WKH SURFHVVHG QRGH N VDWLVI\LQJ WKDW c(q, l(d(N))) − bl ≤ v(nc) ≤ c(q, l(d(N))) + br LV WUDYHUVHG )RU HDFK LWHUDWLRQ LI DQ\ TXRWD RI bl RU br LV XVHG LW LV GHGXFWHG LQ WKH IROORZLQJ LWHUDWLRQ
:H WKHQ WUDYHUVH WKH VXEWUHHV LQGH[LQJ VWULQJV KDYLQJ HGLW GLVWDQFH WR UHIHUHQFH VWULQJV EHWZHHQ ed(q, v(N)) − θ DQG ed(q, v(N)) + θ $OO VWULQJV LQGH[HG E\ QRGHV LQ WKHVH VXEWUHHV DUH FRQVLGHUHG FDQGLGDWH VWULQJV
)LQDOO\ IRU HDFK FDQGLGDWH VWULQJV ZH XVH $OJRULWKP WR YHULI\ ZKHWKHU LW LV D UHVXOW VWULQJ RU QRW ,I LW LV ZH DGG LW LQWR D UHVXOW VHW
7RSk 4XHU\
$OJRULWKP 7RSN4XHU\ VWULQJ q LQW k &RPSOH[WUHH QRGH N QRQUHSHDWLQJ VHW R= ∅
WKUHVKROG θ = 0 ZKLOH|R| < k GR
5DQJH4XHU\q θ N θ θ DGG WKH UHVXOW LQWR R θ = θ + 1
HQG ZKLOH
UHWXUQ WKH k VWULQJV LQ R LQVHUWHG ILUVW
$OJRULWKP JLYHV WKH GHWDLOV RI DQVZHULQJ 7RSk 4XHULHV 7R DQVZHU WRSk TXHULHV
ZH PDNH D VHULHV RI UDQJH TXHU\ DQG LQFUHDVH WKH WKUHVKROG JUDGXDOO\ ,Q HDFK LWHUDWLRQ
ZH DGG WKH UHVXOW LQWR D QRQ UHSHDWLQJ VHW R WR PDNH VXUH WKDW UHVXOW VWULQJV KDYLQJ VPDOOHU HGLW GLVWDQFH WR WKH TXHU\ VWULQJ DUH DGGHG ILUVW
&KDSWHU
([SHULPHQWDO (YDOXDWLRQ
,Q WKLV FKDSWHU ZH JLYH WKH UHVXOWV RI H[SHULPHQWV ZH KDYH FRQGXFWHG %RWK UDQJH DQG WRSk TXHULHV DUH HYDOXDWHG
6HWXS
,Q WKH H[SHULPHQWV UHDO GDWDVHWV DUH XVHG :H XVH WKH GDWD RI $XWKRU IURP '%/3 7KH GDWD ZKLFK PD\ EHFRPH ODUJHU ZLWK WLPH LV GUDZQ RQ -DQXDU\ )RU HDFK TXHU\ ZH DYHUDJH WKH UHVSRQVH WLPH IRU TXHU\ VWULQJV ZKLFK DUH JHQHUDWHG E\ NHHSLQJ UDQGRP VDPSOLQJ XQWLO D VHW ZLWK GLIIHUHQFH EHWZHHQ WKH DYHUDJH DQG YDULDQFH RI LW DQG WKH GDWDVHW LV OHVV WKDQ 7DEOH VKRZV WKH LQIRUPDWLRQ RI WKH GDWDVHW DQG WKH TXHU\ VWULQJ ZH JHQHUDWH
7DEOH 'DWDVHW VWDWLVWLFV
'DWDVHW &DUGLQDOLW\ 0LQ 0D[ $YJ 9DU
$XWKRU
4XHU\
,Q WKH H[SHULPHQWV ZH FRPSDUH RXU DSSURDFK GHQRWHG E\ &7 WR %edWUHH >@ ZLWK WZR GLIIHUHQW VWULQJ RUGHUV LW SURSRVHV GHQRWHG E\ %' DQG %*& UHVSHFWLYHO\
KWWSGEOSXQLWULHUGHGE
$OO SURJUDPV DUH FRPSLOHG E\ * ([SHULPHQWV DUH FRQGXFWHG RQ D PDFKLQH UXQQLQJ /LQX[ DQG HTXLSSHG ZLWK ;HRQ 3URFHVVRU ( DQG *% PDLQ PHPRU\
7KH LQGH[ FRQVWUXFWLRQ WLPH VSHQW E\ WKH PHWKRGV LV OLVWHG LQ 7DEOH 7KH LQGH[
VKRXOG EH FRQVWUXFWHG RQO\ RQFH IRU HDFK GDWDVHW
7DEOH ,QGH[ FRQVWUXFWLRQ WLPH LQ VHFRQG
&7 %' %*&
$XWKRU
6SDFH &RQVXPSWLRQ
:H ILUVW WHVW WKH VSDFH XVHG IRU EXLOGLQJ WKH LQGH[LQJ VWUXFWXUH 7KH VSDFH XVHG RQ HYHU\ OD\HU DUH OLVWHG LQ 7DEOH
7DEOH 6SDFH FRQVXPSWLRQ IRU LQGH[ FRQVWUXFWLRQ LQ 0%
6WULQJ /HQJWK /HWWHU &RXQW 5HIHUHQFH 6WULQJ 7RWDO
&RPSDULVRQ ZLWK 2WKHU $SSURDFKHV
,Q WKLV VHFWLRQ ZH FRPSDUH WKH UHVXOWV RI RXU DSSURDFK ZLWK RWKHUV 5DQJH DQG WRSk TXHULHV DUH HYDOXDWHG LQ VXEVHFWLRQV DQG UHVSHFWLYHO\
5DQJH 4XHU\
,Q )LJXUH ZH FRPSDUH WKH DYHUDJH UHVSRQVH WLPH IRU UDQJH TXHULHV ZLWK GLIIHUHQW WKUHVKROGV RQ WKH DXWKRU GDWD IURP '%/3 7KH UHVXOWV VKRZ WKDW RXU DSSURDFK RXWSHU
IRUPV RWKHUV VLJQLILFDQWO\ IRU VPDOO WKUHVKROGV DQG LV PRUH HIILFLHQW ZKHQ WKUHVKROGV DUH VPDOOHU WKDQ &RPSOH[WUHH LV PRUH HIILFLHQW WKDQ RWKHU DSSURDFKHV IRU VPDOO WKUHVKROG EHFDXVH LWV SUXQLQJ SRZHU GRHV QRW GHJUDGH ZLWK WKH FKDQJH RI OHYHO :KHQ WKUHVKROG LV
1 2 3 4 5 6 10−2
10−1 100 101 102 103 104
WKUHVKROG θ
UHVSRQVHWLPHPV
&7
%'
%*&
)LJXUH $YHUDJH UHVSRQVH WLPH IRU UDQJH TXHU\
VPDOO DQG &RPSOH[WUHH FDQ DOPRVW UHVSRQVH TXHULHV LPPHGLDWHO\ $OWKRXJK WKH VWULQJV WR YHULI\ LQ &RPSOH[WUHH H[SDQG IDVWHU WKDQ WKDW LQ RWKHU DSSURDFKHV GR LW NHHSV LWV DGYDQWDJH IRU D ORQJ SHULRG
7RSk 4XHU\
1 2 4 8
10−2 10−1 100 101 102 103 104
k
UHVSRQVHWLPHPV
&7
%'
%*&
)LJXUH $YHUDJH UHVSRQVH WLPH IRU WRSk TXHU\
)LJXUH VKRZV WKH DYHUDJH UHVSRQVH WLPH IRU WRSk TXHULHV ZLWK GLIIHUHQW WKUHVKROGV RQ WKH DXWKRU GDWD IURP '%/3 )URP WKH UHVXOW ZH FDQ ILQG WKDW &RPSOH[WUHH EHVWV RWKHU
DSSURDFKHV RQO\ ZKHQ k LV OHVV WKDQ +RZHYHU ZH DUJXH WKDW LW LV PDLQO\ FDXVHG E\ WKH UHVXOW VWULQJV WKDW KDYH ODUJH HGLW GLVWDQFH WR TXHU\ VWULQJV ZKLFK DUH XVXDOO\ PHDQLQJOHVV LQ PDQ\ FDVHV
&KDSWHU
&RQFOXVLRQV DQG )XWXUH :RUN
,Q WKLV FKDSWHU ZH FRQFOXGH ZKDW ZH DFKLHYH LQ WKLV WKHVLV LQ 6HFWLRQ :H DOVR VXJJHVW VRPH GLUHFWLRQV IRU IXWXUH ZRUN ZKLFK PDNH WKH DSSURDFK PRUH SRZHUIXO DQG SUDFWLFDO
&RQFOXVLRQV
,Q WKLV WKHVLV ZH SURSRVH DQ LQGH[LQJ VWUXFWXUH LQWHJUDWLQJ GLIIHUHQW ILOWHULQJ PHWKRGV DQG DOJRULWKPV XVLQJ WKLV VWUXFWXUH WR DQVZHU UDQJH TXHULHV DQG WRSk TXHULHV 7KH H[SHU
LPHQWDO UHVXOWV VKRZ WKDW RXU DSSURDFK RXWSHUIRUPV RWKHUV ZKHQ WKH WKUHVKROGV DUH VPDOO HQRXJK
)XWXUH :RUN
2QH GLUHFWLRQ WR HQKDQFH RXU DSSURDFK ZRXOG EH WR LQWHJUDWH PRUH SRZHUIXO ILOWHULQJ PHWKRGV LQWR WKH LQGH[LQJ VWUXFWXUH ,W FRXOG EH HLWKHU FRPSDWLEOH WR WKH SUHVHQW VWUXFWXUH RU PRUH SRZHUIXO WKDQ DQ ROG ILOWHULQJ PHWKRG XVHG
$QRWKHU SURJUHVV WR PDNH LV WR JHQHUDOL]H WKH KDVK IXQFWLRQ PDSSLQJ WKH DOSKDEHW LQWR DQRWKHU XQLYHUVH 7KH KDVK IXQFWLRQ LQ RXU LPSOHPHQW VXSSRUWV RQO\ $6&,, QRZ 0DNLQJ LW VXSSRUW 8QLFRGH HQDEOHV WKH DOJRULWKP WR GHDO ZLWK PRUH GDWDVHWV DQG EH PRUH SUDFWLFDO
%LEOLRJUDSK\
>@ $ &KDQGHO 2 +DVVDQ]DGHK 1 .RXGDV 0 6DGRJKL DQG ' 6ULYDVWDYD %HQFKPDUN
LQJ GHFODUDWLYH DSSUR[LPDWH VHOHFWLRQ SUHGLFDWHV ,Q 6,*02' SS
>@ 6 &KDXGKXUL %& &KHQ 9 *DQWL DQG 5 .DXVKLN ([DPSOHGULYHQ GHVLJQ RI HIIL
FLHQW UHFRUG PDWFKLQJ TXHULHV ,Q 9/'% SS
>@ / *UDYDQR 3 * ,SHLURWLV + 9 -DJDGLVK 1 .RXGDV 6 0XWKXNULVKQDQ DQG '
6ULYDVWDYD $SSUR[LPDWH VWULQJ MRLQV LQ D GDWDEDVH DOPRVW IRU IUHH ,Q 9/'% SS
>@ & /L % :DQJ DQG ; <DQJ 9JUDP ,PSURYLQJ SHUIRUPDQFH RI DSSUR[LPDWH TXHULHV RQ VWULQJ FROOHFWLRQV XVLQJ YDULDEOHOHQJWK JUDPV ,Q 9/'% SS
>@ & /L - /X DQG < /X (IILFLHQW PHUJLQJ DQG ILOWHULQJ DOJRULWKPV IRU DSSUR[LPDWH VWULQJ VHDUFKHV ,Q ,&'( SS
>@ 5 9HUQLFDDQG DQG & /L (IILFLHQW WRSk DOJRULWKPV IRU IX]]\ VHDUFK LQ VWULQJ FROOHF
WLRQV ,Q .(<6 ȷ 3URFHHGLQJV RI WKH )LUVW ,QWHUQDWLRQDO :RUNVKRS RQ .H\ZRUG 6HDUFK RQ 6WUXFWXUHG 'DWD SS
>@ - 4LQ : :DQJ < /X & ;LDR DQG ; /LQ (IILFLHQW H[DFW HGLW VLPLODULW\ TXHU\
SURFHVVLQJ ZLWK WKH DV\PPHWULF VLJQDWXUH VFKHPH ,Q 6,*02' &RQIHUHQFH SS
>@ - :DQJ * /L DQG - )HQJ &DQ ZH EHDW WKH SUHIL[ ILOWHULQJ" DQ DGDSWLYH IUDPHZRUN IRU VLPLODULW\ MRLQ DQG VHDUFK ,Q 6,*02' &RQIHUHQFH SS
>@ $ %HKP 6 -L & /L DQG - /X 6SDFHFRQVWUDLQHG JUDPEDVHG LQGH[LQJ IRU HIILFLHQW DSSUR[LPDWH VWULQJ VHDUFK ,Q ,&'( SS
>@ $ %HKP & /L DQG 0 - &DUH\ $QVZHULQJ DSSUR[LPDWH VWULQJ TXHULHV RQ ODUJH GDWD VHWV XVLQJ H[WHUQDO PHPRU\ ,Q ,&'( SS
>@ 6 &KDXGKXUL 9 *DQWL DQG 5 .DXVKLN $ SULPLWLYH RSHUDWRU IRU VLPLODULW\ MRLQV LQ GDWD FOHDQLQJ ,Q ,&'( SS
>@ $ $UDVX 9 *DQWL DQG 5 .DXVKLN (IILFLHQW H[DFW VHWVLPLODULW\ MRLQV ,Q 9/'%
SS
>@ * /L ' 'HQJ - :DQJ DQG - )HQJ 3DVVMRLQ $ SDUWLWLRQEDVHG PHWKRG IRU VLPL
ODULW\ MRLQV ,Q 39/'% YRO QR SS
>@ = =KDQJ 0 +DGMLHOHIWKHULRX % & 2RL DQG ' 6ULYDVWDYD %HGWUHH DQ DOOSXUSRVH LQGH[ VWUXFWXUH IRU VWULQJ VLPLODULW\ VHDUFK EDVHG RQ HGLW GLVWDQFH ,Q 6,*02' SS
>@ : /X ; 'X 0 +DGMLHOHIWKHULRX DQG % & 2RL (IILFLHQWO\ VXSSRUWLQJ HGLW GLVWDQFH EDVHG VWULQJ VLPLODULW\ VHDUFK XVLQJ %+WUHHV ,Q ,((( 7UDQVDFWLRQV RQ .QRZOHGJH DQG 'DWD (QJLQHHULQJ YRO QR SS
>@ : - 0DVHN DQG 0 3DWHUVRQ $ IDVWHU DOJRULWKP FRPSXWLQJ VWULQJ HGLW GLVWDQFHV ,Q -RXUQDO RI &RPSXWHU DQG 6\VWHP 6FLHQFHV
>@ 7 + &RUPHQ & ( /HLVHUVRQ 5 / 5LYHVW DQG & 6WHLQ ,QWURGXFWLRQ WR $OJRULWKPV
,Q 0,7 3UHVV DQG 0F*UDZ+LOO