無線點對點資訊分享網路中對搭便車資訊享用者之無滲透研究

ณጤᘉᄈᘉၦଊϸٵᆪၰϜᄈཧ߰ٚၦଊ

ٵң޲ϟณᅥഇःف

ुԉཐ ࢈ݾτᏱၦଊᆔ౪ق ༂ φ ࡰ ህϧτᏱၦଊᆔ౪ق

ᄣ्

GnutellaȃNapster ޠᘉᄈᘉȞPeer to Peer, P2Pȟၦଊϸٵقಜϟ໣ߩւулဏޠ ኈ៫ϟί८ᖞΠߩளᝓ२ޠࢆ᏾Ȃԫࢆ᏾֊࣐ၦଊϸٵϛᄈᆏޠୱᚡȄӶҐ፤НϜ஡ ϟᆏ࣐ȶၦଊ߰ٚ޲Ȟfree riderȟȷឋᚡȄᓎ຀ᘉᄈᘉณጤᓎཏᆪၰȞP2P wireless ad-hoc network, WP2PȟҼ࢑ڏԥା࡚ึ৥዗Ωޠስ஀ȂҐ፤Нӱԫණяᘉᄈᘉณጤᆪၰᕘძ ίᒌ໕ȶၦଊଔᝧ࡚ȷޠዂ࠯Ȃп၍؛ӶณጤᘉᄈᘉၦଊϸٵᆪၰϜၦଊ߰ٚ޲ޠୱ ᚡȂໍՅණٽᘉᄈᘉϸයԓᆪၰ࢝ᄻΚ՞ًԂޠၦଊϸٵஆᙄᕘძȄԫዂ࠯ᆏϟ࣐ȶ௒ ძϾၦଊଔᝧ࡚ዂ࠯ȷ(CICM)ȂӶᘉᄈᘉၦଊϸٵޠႇโϟϜȂڐՄኍژၦଊࠣ፵ȃ ၦଊቌঅȃਣਞܓȃϸٵ๋౲ȃ௒ძӱષȃᆪၰᓝቷІၸڏޠਞ૗๊ӱφȂٯйւң αख़ӱφȂп P-Grid ϸයԓᓾԇ࢝ᄻ࣐ஆᙄȂຠզঐᡞᄈܼԫၦଊϸٵᆪၰޠଔᝧ࡚Ȃ ٿ੒ଷӶᘉᄈᘉณጤᓎཏᆪၰၦଊϸٵϜၦଊ߰ٚ޲ޠୱᚡȂп෉ႁژစᔽਞ౦Іϵ ҂঩ࠍȂᗘռᘉᄈᘉᆪၰϜณਞ౦ȃϛᄈᆏޠၦଊϸٵȂܗ࢑ആԚঐᡞϟ໣ၦଊϸٵ ԚҐαณᒞޠ཭ѷȄ ᜱ ᗥ Ԇ Ȉ ᘉᄈᘉȃϸයԓ೏౪ȃၦଊϸٵȃၦଊଔᝧ࡚ȃ߰ٚၦଊٵң޲ȃ՘୞ᓎ ཏԓᆪȁ

A Study on Free-Rider Distributed Impermeability

for WP2P Content Sharing

Soe-Tsyr Yuan

MIS Department, National Chengchi University

Tzu-Heng Huang

IM Department, Fu-Jen University

Abstract

Gnutella and Napster, once most popular P2P file sharing systems, have encountered a serious problem for their significantly unbalanced content sharing between peers. Most peers in the sharing networks benefit from the sharing of others without contributing. This problem is so-called the free-rider problem. It is a very important issue to avoid the free-rider problem so that all of the peers in the networks can benefit substantially from collective actions so as to sustain the operations of the sharing networks. For the foreseeable wireless P2P contents sharing, this paper presents a correspondent solution model (utilizing the P-Grid storage model) named as “Contextualized Information Contribution Model, CICM”. This model enables peers in the network to share content fairly and efficiently by considering information contribution of peers and instituting peers the incentives to account for the global benefits of a private act.

Keyword: P2P, distributed computing, content sharing, information contribution, free rider, ad-hoc network

൧ȃᆲ፤

ߗٿȂȶᘉᄈᘉϸයԓᆪၰ࢝ᄻȷೞ Intel ൖ࣐ಒήхᆪၰ९ڽ[්྇സ, 2001] [A. Oram, 2001]ȄӶᆪርᆪၰȞInternetȟαޠԇӶ೩ӼᘉᄈᘉၦଊϸٵР਱ȂԄ Gnutellaȃ Napster ϟၦଊϸٵȄดՅڐϜ 90%пαޠঐᡞȞpeerȟ೾ள࢑ϛཽϸٵၦଊࠔѬ्ί ၸၦଊޠ [E. Adar et al., 2000]Ȃ೼ᆎၦଊϸٵϛᄈᆏޠୱᚡᝓ२ኈ៫ၦଊ݉ଡ଼ޠࡼε ܓȄӶҐН஡ԫࢆ᏾ಜᆏϟ࣐ȶၦଊ߰ٚ޲Ȟfree riderȟȷȄ

ѫѵȂᓎ຀ԥጤژณጤޠऌ׭ϟץഁᅌໍȂٻூґٿޠၦଊऌ׭౱ࠣ஡пಌ୞૗ Ω࣐ഷུޠຨؒȂณጤޠᔗң஡؂ຯߗঐ΢ҢࣁȂࣦՎᒋӬ࣐ҢࣁΚޠഌϸȄӶณጤ ᔗң݉ଡ଼ϜޠΚݓȂᎍ஀୧ଡ଼Ȟubiquitous commerceȟȂ೼ঐ྆܉஡ߵٻ՘୞୧ଡ଼Ȟmobile commerceȟᗛӪѫΚঐӓུޠძࣩȂٗял௄ԓȞclient/serverȟޠ࢝ᄻ [S. Schapp et al., 2002]Ȅၦଊᔗң݉ଡ଼ϛӕѬ࢑੒ຳܓ݉ଡ଼࣐лȂ஡ཽᙾӪঐ΢ϾޠၦଊᔗңȂٯ йױᘉᄈᘉณጤᆪၰޠ੬ܓાΤ೪ॏޠՄኍȂپԄȈՍשಣᙒȞself-organizationȟȃ୞ ᄙȞdynamicȟᆺӬȃ௒ძᄇញȞcontext awarenessȟ๊੬ܓ [N. Daswani et al., 2002]Ȅ ࢑ࢉȂณጤᘉᄈᘉᆪၰޠ࢝ᄻα஡Ҽ૗ᅌϾяᄊུޠၦଊᔗң݉ଡ଼ȂӶၦଊ݉ଡ଼೪ॏ ޠ࢝ᄻαη஡࢑ӓུޠࢆ᏾ȂپԄȈՍשಣᙒޠ૗Ωᡲٻң޲Ӷٻңၦଊ݉ଡ଼ਣ᐀ԥ ؂τޠኇܓȞscalabilityȟȂ୞ᄙᆺӬޠ૗Ωٻڐڏԥ֊ਣϾಣӬР਱Ȃ௒ძᄇញޠ૗ ΩࠍණٽΠഷᎍӬ࿌ίޠ݉ଡ଼ܗᔗԥޠЇᔗȄ ՅӶณጤᘉᄈᘉᆪၰϜȂၦଊණٽ޲ޠِՔཽԥܛ׾ᡑȂ༉ಜޠл௄࢝ᄻȂϑစ ࡈ֚Πၦଊ݉ଡ଼ණٽᇅٻңޠِՔ࢑ڿۢϛᡑޠȂၦଊණٽ޲ᇅٻң޲ϟ໣ޠِՔ࢑ ѠпϤඳȂߩڿۢϛᡑȂҼ֊ᆪၰϜޠؑঐ୥ᇅ޲എԥѠ૗ණٽ݉ଡ଼ܗٻң݉ଡ଼Ȃପ Ӭ୞ᄙޠณጤᆪၰᕘძȂᐍঐၦଊᆪၰ᠂ดםԚΚঐ୞ᄙҀ໲Ȃщᅗөᆎםԓޠၦଊ ݉ଡ଼ӶҀ໲αࢻ୞Ȃܛпᘉᄈᘉၦଊϸٵޠ୧ཿዂԓཽᡑூ؂ԥ፹ȃ؂ԥᡑϾܓȂѠ пึ৥؂ፓᚖޠ๋౲ᇅӼϰϾޠၦଊ݉ଡ଼Ȅ ดՅȂԄӤᆪርᆪၰޠᘉᄈᘉၦଊϸٵȂӶณጤᘉᄈᘉᆪၰᕘძϜȞپԄᗋޑϜ ЗϲޠᘉᄈᘉณጤᆪၰᕘძȟΚኻཽႅژၦଊϸٵϛᄈᆏޠ֩ძȂᖟԄȈ(1) ၦଊϸ ٵ໲ϜӶ੬ۢঐᡞȂϸٵϛᄈᆏޠ௒ძȄ(2) ᗋޑၦଊӱ࣐ၦଊ߰ٚ޲ณݳԥਞයո ژ୧ൠޠؑ՞ِဤȂณݳԥਞޠڗᐮ੒ຳȄ(3) ϸٵᆪၰޠҠ၍ܗ୅ᘜȄ ӱԫӶԄԫାഁᡑ୞ޠᘉᄈᘉณጤᆪၰᕘძϟίȂጃߴঐᡞϟ໣ၦଊϸٵᄈᆏޠ ᐡښ࢑२्ޠȄӱ࣐ጃߴၦଊϸٵᆪၰϜၦଊޠࢻ೾ܓ࢑ᘉᄈᘉၦଊϸٵᆪၰޠஆᙄ ౪፤Ȃη୳ԥጃߴᘉᄈᘉၦଊϸٵᆪၰϜၦଊޠࢻ೾ܓᇅঐᡞȞpeerȟޠϸٵ౦ȂϘ ૗ԥਞစᕋΚঐᘉᄈᘉၦଊϸٵޠ҂ѯȄดՅၦଊϸٵᄈᆏޠᐡښ࢑ሰ्Κঐڏറȶစ ᔽਞ౦ȷІȶϵ҂ࡋᢏȷϟၦଊଔᝧ࡚ዂ࠯೪ॏȄ

ҐНӱԫණяȶ௒ძϾၦଊଔᝧ࡚ዂ࠯ȞContextualized Information Contribution Model, CICMȟȷȂڐӶᘉᄈᘉၦଊϸٵޠႇโϟϜȂՄኍژၦଊࠣ፵ȃၦଊቌঅȃਣ ਞ ܓ ȃ ϸ ٵ ๋ ౲ ȃ ௒ ძ ӱ ષ Ȟ context factorȟȃ ᆪ ၰ ᓝ ቷ І ၸ ڏ Ȟ deviceȟ ޠ ਞ ૗

Ȟperformanceȟ๊ӱφȂп෉ႁژڏစᔽਞ౦Іϵ҂ࡋᢏ੬ܓϟၦଊϸٵᄈᆏޠᐡښȄ ҐНпτ࠯ᗋޑϜЗȞshopping mallȟ࣐ᄃᡜ௒ძȞԄშ 1 ܛұȟȂӶᘉᄈᘉၦଊ ϸٵณጤᆪၰϟίȂ੒ຳ޲ܗ୧ঢ়Ѡпϸٵ֊ਣޠᗋޑၦଊȂپԄȈ୧ࠣၦଊȃኅ֚ ၦଊȃߵ᎜ࣁ୞ܗ׸ቌڕ๊Ȃй୧ۺϟ໣ҼѠᖟ՘๋౲ᖓ࿘ޠӬձࣁ୞ȞپԄȈၯ୧ ঢ়ޠᇰӤћȂᖓӫ੒ຳޠ׸ԛȟȄҐःف஡ؑΚঐ੒ຳ޲ְຝ࣐ঐᡞȞpeerȟȂঐᡞܛ ׹ᅌޠِՔ࢑ၦଊණٽ޲η࢑ၦଊίၸ޲Ȃ୧ࠣၦଊഇႇୣ஀ܓޠ provision servers ึଛȂӕҦؑ՞ၦଊঐᡞ׳ژҭࠊᗋޑϜЗϲ᎒ߗԇӶޠၦଊȞ݉ଡ଼ȟණٽ޲ໍ՘ၦ ଊһඳȄҐःفޠᄃᡜ҂ѯ࢝ᄻ឵ెӬԓᘉᄈᘉณጤᆪၰȄ ӶᗋޑϜЗޠᕘძϜȂ੒ຳၦଊϸٵ࢑Κᆎϛ໣ᘟޠࣁ୞Ȃέ੒ຳၦଊϸٵ҇໹ ૗୞ᄙޠಣᙒ໋᎐ޠঐᡞםԚϸٵᆪၰٿໍ՘ϸٵࣁ୞ȂՅ໲Ϝԓ࢝ᄻԥ൑Κᘉᡪᅾ ޠقಜॴᓏпІାଊਁྜྷ೾ᇅၦଊ೏౪૗ΩޠሰؒȂйᆪၰณݳՍשಣᙒȂ࢑ࢉҐः فпኇܓၷτޠϸයԓ࢝ᄻ୉࣐ԫᄃᡜᕘძޠஆᙄȄᙥҦଔᝧ࡚Ȟcontributionȟຠզ ঐᡞȂп੒ଷӶᘉᄈᘉᓎཏᆪၰၦଊϸٵϜၦଊ߰ٚ޲ޠୱᚡȂ෉గႁژစᔽਞ౦І ϵ҂঩ࠍޠȶ௒ძϾၦଊଔᝧ࡚ዂ࠯ȷȂᗘռᘉᄈᘉณጤᆪၰϜณਞ౦ȃϛᄈᆏޠၦଊ ϸٵȂܗ࢑ആԚঐᡞϟ໣ၦଊϸٵԚҐαณᒞޠ཭ѷȂٯණାΠᗋޑϜЗ၈ঐᡞϟ໣ ޠၦଊࢻ୞ܓȂй७մཿлп௱ԓኅ֚ആԚޠԚҐᇅณਞ౦ޠ௱ᙩ๗ݏȂܗ๞ϡ੒ຳ ޲ᎍӵϾޠ֊ਣ੒ຳၦଊȄ Ad-Hoc Network Other Group Shop 2 Free rider Peer A Shop 1 PDA Peer B WWW Provision Servers Fast Food Coffee Shop Restaurant შ 1.ᘉᄈᘉณጤᆪၰϜȂᗋޑϜЗޠ֊ਣၦଊϸٵ Ґ Н л ् ϸ ࣐ ή ഌ ӌ Ȅ ಒ Κ ഌ ӌ ᇴ ݃ ҭ ࠊ ԇ Ӷ ϟ ၦଊଔᝧ࡚ዂ࠯ᐡ ښ ๊ ࣻ ᜱ ः ف Ȅ ಒ Ρ ഌ ӌ ࠍ ֖ ౫ CICM ϟ ः ف Р ݳ ᇅ ࢝ ᄻ Ȅ ಒ ή ഌ ӌ ӗ я Π Ґ ः ف ϟ C IC M ق ಜ ҂ ѯ ᇨ ձ ᇅ ᄃ ᡜ ೪ ॏ Ȃ ٯ ᙥ Ҧ ᄃ ᡜ ኶ ᐄ ᡜ ຢ C IC M ϟ ቌ অ Ȅ ഷ ࡤ ࣐ Ґ Н ๗ ፤ Ȃ ٯ ණ я ࡤ ៊ ः ف ޠ ࡛ ឋ Ȅ

ະȃНᝧ௥ଇ

Ӷ Ґ ࿾ Ϝ ש উ ଭ ᄈ ᘉᄈᘉၦଊϸٵึ৥౫ݸІႇ џ ԥ ႇ ޠ ࣻ ᜱ ଔᝧ࡚ዂ࠯ ໍ ՘ ௥ ଇ Ȃ Ᏹ ಭ ႇ џ ޠ ޤ ᜌ ᇅ စ ᡜ ٯ ׳ я ႇ џ ः ف ϛ ٘ ϟ ೏ Ȅ

Websense Inc. (NASDAQ: WBSN) ܛձޠ።ࢦൣ֚ࡿяȂҦ Napster ழକޠᘉᄈᘉ ᔭ਱ϸٵᆪ๝Ȃܛһඳޠᔭ਱ϑစҦ MP3 ॲ዆ᘘ৥Վᗻቒȃࢻ՘Ⴌຝၾᔜ๊Ȃєᛴ࿳ ԥȄᡑϾӼᆓޠһඳϲৡȂϛ൑цུޠ P2P ᔗңโԓІᆪયԄߧࡤࢍ๐ૢСઊቩӼȂ Ҽ࣐ӓ౩ӍཿழٿΠөᆎᜱܼᓝቷȃݳࡢІߴԋޠୱᚡȄٲᄃαȂP2P ᔭ਱ϸٵᆪય ܼႇџ 12 ঐУϲቩߞႊ 300%ȂႁՎ 89,000 ঐϟӼȄᐄসϏαᆪᆔ౪(EIM)Р਱ӓ౩ ስᏳٽᔗ୧ Websense ࡿяȂ౫ਣϤᖓᆪαԥົႇ 130 ঐϛӤޠ P2P ᔗңโԓٽңЙ ٻңȂєࢃ KaZaaȃGrokster ๊Ȅःفᐡᄻ Yankee Group զॏџԒՍ P2P ᆪ๝ίၸޠ ॲ዆ᔭ਱ϑႊ 50 ቈঐȂՅၾᔜ໡ึ୧ Trymedia զॏџԒޠႬຝၾᔜίၸ኶ҭҼົႇ 500 ࿳ঐȄԫѵȂ៬ୱᐡᄻ Viant ҼզॏؑСޠႬኈᔭ਱ίၸ౦Ҽႁ 40 ࿳Վ 60 ࿳ঐȇ ՅؑСङԥ 300 ࿳ӫңЙܼ KaZaa ίၸ”Buffy the Vampire Slayer”๊ڨ᠎ߕޠႬຝഀ៊ ቒȄ

Ӷ [G. Kortuem, 2001] [N. Daswani et al., 2003] ޠःفࡿяȂᘉᄈᘉၦଊقಜ࢑Κ ঐՍݾޠϸයԓقಜȂؑ՞ঐᡞഎ࢑ᑀҴޠစᔽᡞ1Ȃ᐀ԥՍϐޠ؛๋Ȃйڐϸයԓޠ ࢝ᄻѠпϸٵঐᡞޠΚϹႬဟၦྜ2ȂՅᘉᄈᘉഷτޠ੬Ք൸࢑ϸයԓޠ೏౪ȂءԥϜ Ѷᆔ౪Ȟcentralized controlȟȂР߰ΠၦྜޠӔңȂկηആԚΠقಜਞ૗ຠզϛܿȃၦ ଊཫ൷ᇅၦଊقಜԋӓޠୱᚡȂڐϜᜱܼၦଊཫ൷ᇅၦଊԋӓޠଇ፤ᐍ౪ԄίȈ

Κȃཫ൷ᐡښȞsearch mechanismȟ

ΚঐԂޠཫ൷ᐡښȂӶᘉᄈᘉޠ࢝ᄻϜ׹ᅌΠ᡺ቁޠِՔȂӱ࣐ԥΠًԂޠཫ ൷ЖᔞϘԥᒳݳҔጃέԥਞ౦ޠ׳ژঐᡞܛሰژޠၦଊȄՅΚঐཫ൷ᐡښ஡ۢဏΠ ᘉᄈᘉᆪၰ೪ॏޠ࢝ᄻпІঐᡞ՘࣐ዂԓȂשউ௄ί८ήঐᆱ࡚ଇ፤Ȉ z ܦ⭚ȞtopologyȟȂঐᡞഀ௦ޠРԓȂ֊ᐍঐϸයԓᆪၰಣԚޠ࢝ᄻȂཽӱ࣐ ϛӤޠقಜԥϛӤޠ೪ॏȄ z ၦਠဋܺȞdata replacementȟȂӶϸයԓᆪၰϜȂঐᡞᄈܼၦਠᓾԇᇅᕖڦޠ РԓȄ z ଊਁၰҦȞmessage routingȟȂঐᡞᙾଛᆪၰࡍєޠၰҦϏձȂۢဏΠঐᡞޠ ՘࣐ȂᖟԄԄեࢦၛȃӲᔗܗ࢑༉ሏଊਁޠ೤ጓȄ 1ܛᒞޠစᔽᡞࡿޠ࢑ؑ՞ঐᡞཽп൷ؒՍٙւઊഷτϾ࣐яึᘉȂໍ՘ᘉᄈᘉၦଊϸٵܗ࢑һܿȂ ࢑ࢉϸٵᆪၰ࢑Κঐϊ࠯ޠစᔽ୽Ȅ 2ႬဟၦྜȂހࡿΚϹᔭ਱ȃᔗңโԓȃॏᆘ૗Ωȃᓾԇޫ໣ȃၰҦᙾԇ๊૗ΩȄ

ΡȃԋӓȞSecurityȟ

ᄈܼඍཏঐᡞȞmalicious peerȟޠ߮ΤܗખᚾȂѠ૗ኈ៫ژϸයԓقಜޠᛨୋ ࡚Ȃܛпԋӓޠឋᚡϛৡ܈ຝȄՅᄈܼԋӓޠ೪ॏѠՄኍޠӱφԄίȈ z ѠூܓȞavailabilityȟȂؑঐᘉᄈᘉᆪၰϜޠঐᡞȂ҇໹૗௦ڨڐуঐᡞޠଊਁȂ ٯ й ϸ ٵ ᆪ ၰ ܗ ࢑ Ⴌ ဟ ၦ ྜ Ȃ կ Ѡ ૗ ೞ ඍ ཏ ঐ ᡞ ᕁ ң Ȃ Ԅ ึ ୞ ߣ ᘟ ԓ ׿ ᔟ Ȟdenial-of-service, DoSȟȄ z ᔭ਱ᇰᜍȞfile authenticityȟȂᄈܼঐᡞܛࢦၛޠᔭ਱ၦਠȂٯሰ૗ձژ઎Ϲܓ ޠᡜᜍȂጃۢዀޠޑԇӶܼᘉᄈᘉၦଊϸٵᆪၰϟϜȄ z ୢӫȞanonymityȟȂᄈܼၦଊϸٵ޲ܗ࢑፝ؒ޲҇໹૗ձژߴៗঐᡞঐ΢ၦଊ ޠԋӓ௢ښȄ z ԇڦ௢ښȞaccess controlȟȂᄈܼقಜၦྜܗ࢑᠍४ޠᆔ౪ȂпІහኌଓ౱᠍ޠ ୱᚡȄ ҭࠊᘉᄈᘉၦଊϸٵܛ᎐ႅژၷτޠୱᚡ࢑ၦଊ߰ٚ޲Ȟfree riderȟᇅ߭ӉȞtrustȟ ޠឋᚡȄӶၦଊ߰ٚ޲ޠഌϸȂEytan Adar ๊΢ޠःف [E. Adar et al., 2000] Ϝп Gnutella ࣐پࡿяȂӶ Gnutella ၦଊϸٵقಜϜȂ஡ߗԼϸϟΝΫޠٻң޲࢑ϛϸٵ ၦਠޠȂՅй஡ߗԼϸϟϥΫޠӲᔗଊਁ࢑ٿՍԼϸϟΚޠٻң޲Ȃη൸࢑ᇴංоΟ Ԛޠٻң޲ᄈ೼ঐၦଊᆪၰ࢑ءԥଔᝧޠȄ೼ঐኈ៫ޣ௦ആԚΠၦଊϸٵᆪၰޠ୅ᅘ ܗ࢑஬၍Ȃ࢑ҭࠊᘉᄈᘉၦଊϸٵᆪၰࡩሰ၍؛ޠឋᚡϟΚȄӶ߭ӉޠഌϸȂKarl Aberer [K. Aberer et al., 2001] ࡿя߭Ӊ࢑ᘉᄈᘉၦଊϸٵ२्ޠΚᕘ(Ԅშ 2 ܛұ) Ȅ ӱ࣐ӶᘉᄈᘉၦଊϸٵᆪၰϜȂΚૢՅّȂٻң޲ܛ௦ដژޠঐᡞഎ࢑ϛᇰᜌޠঐᡞȂ ܛпȂঐᡞ໣߭Ӊޠᆔ౪஡ኈ៫ᘉᄈᘉقಜޠԚంȂՅᆔ౪߭ӉޠРݳѠпւңঐᡞ ޠ߭៘Ȟreputationȟٿᒌ໕߭Ӊ࡚Ȅ

ҭࠊڏхߓܓޠଔᝧ࡚ዂ࠯ٿ၍؛ၦଊ߰ٚ޲ޠРݳ൸࢑ Sepandar D. Kamvar ๊ ΢ණяп߭៘Ȟreputationȟ࣐ஆᙄޠ߭Ӊ࡚ᆔ౪ [S. D. Kamvar et al., 2003] ᇅ C. Courcoubetis ණяޠҀൠԚҐዂԓ࣐ஆᙄޠຠ໕قಜ [C. Courcoubetis et al., 2002]ȄՅ ณ፤߭៘ޠᒌ໕ྦࠍܗ࢑ᅌᆘݳԄեᡑϾᒌ໕ȂڐनࡤԥΚঐࣻӤޠ౪፤ஆᙄȂ٦൸ ࢑စᔽҀൠȂؑ՞ःف޲എຝᘉᄈᘉၦଊϸٵᆪၰޠၦଊϸٵһ࣐ܿΚ໷စᔽҀൠα ޠһܿȂйؑঐঐᡞഎ࢑ΚঐစᔽޠঐᡞȂпഷτϾঐᡞւઊ࣐؛๋һܿޠ঩ࠍȂԥ Π х ቌ ޠ ྆ ܉ Ȃ ᖟ δ ؑ Ԫ һ ܿ എ ԥ Κ ঐ р ຳ ޠ ྆ ܉ Ȅ ௄ ߭ ៘ Ȟ reputationȟȃ ਞ ࡚ ȞutilityȟȃԚҐȞcostȟяึȂٯԥΚૢစᔽҀൠޠᝰލȂӶԫዂԓίႁژစᔽ҂ᒌȂ пᕖூᐍᡞစᔽҀൠഷτޠਞઊȄӱԫȂҀൠዂԓѠձ࣐ᘉᄈᘉၦଊϸٵΚঐ२्ޠ ஆᙄᕘძȂ҂ಌαख़ޠ྆܉Ȃस࢑૗ႁژ׈ӓਞ౦ޠစᔽҀൠȂ٦ቅҼ૗໣௦၍؛ഌ ϸޠၦଊϸٵϛᄈᆏޠឋᚡȄ ϛႇҭࠊ೼ٳଔᝧ࡚ዂ࠯എᗚ࢑࡛ҴӶл௄ԓޠ࢝ᄻϟίȂณ፤࢑߭Ӊ࡚ޠᆔ౪ ܗ࢑рຳᐡښॏᆘഎሰ्ഇႇϜѶޠժ݉Ꮳٿཫ൷ঐᡞϟ໣ޠၦଊпໍ՘ၦଊϸٵࣁ ୞Ȅѫѵ೼ٳଔᝧ࡚ዂ࠯܈౲Π௒ძ឵ܓȞcontextualized attributeȟ೼ঐӱφȄӱԫȂ ᗷด૗ॏᆘяၦଊϸٵҐٙ࢑֐૗࣐ሰؒ޲ܗထᡞȞgroupȟ഻ആяւઊȂໍՅ؛ۢ࢑ ֐ၦଊϸٵȂկ࢑ࠔณݳॏᆘяӱ࣐௒ძ឵ܓȃٻң޲ঢຬȞuser profileȟܛആԚޠ ၦଊϸٵԚҐȂпІւңϸයԓॏᆘ࣐ஆᙄޠᓻᘉᇅ੬ܓȂӱԫঅூђп׾๢Ȅ ՅҐःفණяޠ CICM ዂ࠯ࠍ࢑ᒶᐆпϸයԓ࣐ஆᙄ࢝ᄻȂঐᡞϟ໣ޠၦଊһඳ ᇅଔᝧ࡚ޠॏᆘ࢑ϸයԓйӶ୞ᄙޠᕘძϟίܛಣӬՅԚȄՅڎᆎ࢝ᄻഷτޠৰ౵Ӷ ܼȈ z ϸයԓޠ࢝ᄻءԥ໲Ϝԓ൑ᘉᡪᅾᏳयᆪၰ୅ᘜޠୱᚡȄ z ϸයԓѠп؂୞ᄙйኇܓޠҦ໋᎐ޠঐᡞٿ؛ۢϸٵޠထᡞȂဏ֊୞ᄙᆺӬȞslice of momentȟϟ૗ΩȄ z ϸයԓѠпϸයၦଊ೏౪ޠၦྜᇅᆪၰࢻ໕Ȃᗘռ໲ϜӶϜѶȂՅሰ्τ࠯лᐡ ٿ೏౪ᛂτޠྜྷ೾ᇅڐуၦଊ೏౪ޠԚҐȄ ࢑ࢉȂϸයԓй୞ᄙᆺӬঐᡞ࢑ᇅ౫ԇޠःفഷτޠϛӤϟ೏Ȃ࢑Ґःفޠଔᝧ ܛӶɯණٽޠΚঐུϸٵຠզ࢝ᄻȄ

୥ȃःفРݳ

࣐ᗘռᘉᄈᘉณጤᆪၰϜณਞ౦ȃϛᄈᆏޠၦଊϸٵȂܗ࢑ആԚঐᡞϟ໣ၦଊϸ ٵԚҐαณᒞޠ཭ѷȂҐ፤НණяΚঐȶ௒ძϾၦଊଔᝧ࡚ዂ࠯ȷ(CICM)ȄӶᘉᄈᘉ ၦଊϸٵޠႇโϟϜȂՄኍژၦଊࠣ፵ȃၦଊቌঅȃਣਞܓȃϸٵ๋౲ȃ௒ძӱષȃ ᆪၰᓝቷІၸڏޠਞ૗๊ӱφȂٯйւңαख़ӱφȂпϸයԓॏᆘ࣐ஆᙄȂຠզঐᡞ ᄈܼԫၦଊϸٵᆪၰޠଔᝧ࡚Ȃٯւңԫၦଊଔᝧ࡚੒ଷӶᘉᄈᘉᓎཏᆪၰၦଊϸٵ Ϝၦଊ߰ٚ޲ޠୱᚡȂп෉ႁژစᔽਞ౦Іϵ҂঩ࠍȄ ѫѵȂҦܼณጤᆪၰޠ୞ᄙܓȂࢉ CICM ණٽޠٯߩ࢑ΚঐՄኍᐍᡞޠଔᝧ࡚ Ȟglobal contributionȟዂ࠯ȂՅ࢑Κঐ୞ᄙᆺӬȞslice of momentȟԓޠຠզዂ࠯ٿଷ џၦଊϸٵᆪၰϜၦଊ߰ٚ޲Ȟfree riderȟޠୱᚡȄԫѵȂCICM ၦଊϸٵ҂ѯଅᓄ ୞ᄙᆺӬ໣ঐᡞޠһܿစᡜȂٯй҂ಌစᡜژίԪһܿࣁ୞Ȃпණ݈ၦଊϸٵ݉ଡ଼ޠ ঐ΢Ͼโ࡚Ȅ ӶҐ࿾ϜȂשউ஡Ӓඣख़ CICM ၦଊϸٵ҂ѯϟ࢝ᄻІ࢑ԄեႁژႲ෉ҭዀȂٯ йӶഛ៊ޠϊ࿾ϜၐಡޠϮಞ CICM ϲഌөϰӈޠѓ૗ᇅړ኶ዂ࠯ޠۢဏІᅌᆘݳȄ

ΚȃCICM ၦଊϸٵ҂ѯϟ࢝ᄻ

ܼ CICM ၦଊϸٵ҂ѯ࢝ᄻ (ܼშ 3 ܛұ)ϜȂᘉᄈᘉၦଊϸٵຠզՄ໕Πঐᡞϟ ϸٵቌঅȞsharing valueȟȃঐᡞϟܣ๙౦Іঐᡞϟೞܣ๙౦ήঐӪ໕пॏᆘяၦଊଔ ᝧ࡚Ȟinformation contributionȟȄٯйᙥҦၦଊଔᝧ࡚ޠॏᆘໍՅᗘռณጤᘉᄈᘉၦଊ ϸٵᆪၰϜၦଊ߰ٚ޲ޠୱᚡȄӶຠզၦଊଔᝧ࡚ޠӤਣȂؑ՞ٻң޲᐀ԥՍϐޠၦ ଊϸٵ๋౲ȂՅၦଊϸٵ๋౲ޠםԚࠍ࢑ഇႇ CICM ҂ѯϜ Strategy Evaluator ޠϰӈ ٿ౱ҢȂStrategy Evaluator ཽ਴ᐄٻң޲ Profile Ϝޠ឵ܓܗ࢑ٻң޲Ⴒ೪ޠ੬ۢ௒ძ ϸٵ๋౲ȂпІ Context Agent ܛ୏ขژޠᕘძӱφȂԄᆪၰᓝቷȃࢻ໕ȃഀጤ኶ȃಥ ᆓ೪റ೏౪૗ΩȂःᔤяഷᎍӬٻң޲࿌ί఩ӈޠϸٵ๋౲ȂӱԫȂ࿌ໍ՘ၦଊϸٵ ޠਣ঑ȂCICM ҂ѯཽՄኍژ࿌ί໋᎐ޠᕘძޒᄙȃᆪၰޒᄙпІٻң޲ᄈϛӤൠӬ ਣᐡޠϸٵዂԓȄ

ԫѵȂӶၦଊϸٵޠႇโϟϜȂCICM ҂ѯཽ҂ಌؑԪၦଊϸٵһܿޠစᡜژί Ԫၦଊϸٵࣁ୞ٙαȂٸᐄٻң޲ޠһܿࣁ୞ȂCICM ഇႇ Preference Learning Agent ȞPLAȟٿ።ᐍࣁ୞᜹ր Ontology [N. Guarino, 1997] [R. Jasper et al., 1999] ޠ᜹ր᠍ २ȂҼ֊Ᏹಭٻң޲ޠһܿစᡜȂᏱಭٻң޲Ӷᒶᐆၦଊᔭ਱ޠ᜹ր୒Ԃᇅ੬ՔȂٯ ӶίԪၦଊϸٵࣁ୞౱ҢਣȂՅࡤ਴ᐄٻң޲ؑԪၦଊϸٵһܿᄈዀޠၦଊޠᒶᐆȂ ණٽᑣᒶၦଊఽ൑ਣ๞ϡᎍ࿌ޠ።ᐍȂپԄ஡ٻң޲ൊԂޠ᜹րၦଊࢆᒶяٿٯဋܼ ఽ൑ࠊӗȂ࢑ࢉȂ૗ႁژ҂ಌؑԪһܿစᡜޠҭޠȄ

˘ˡ˩˜˥ˢˡˠ˘ˡ˧ ʻ˔˷ˀ˛̂˶ʳˡ˸̇̊̂̅˾ʼ ˣ˥ˢ˙˜˟˘ ˖˜˖ˠʳˣ˟˔˧˙ˢ˥ˠ ˖ˢˡ˧˘˫˧ʳ˔˚˘ˡ˧ ˣˀ˚˥˜˗ ˠ˔ˡ˜ˣ˨˟˔˧ˢ˥ ˦˧˥˔˧˘˚ˬ ˘˩˔˟˨˔˧ˢ˥ ˣ˥˘˙˘˥˘ˡ˖˘ ˟˘˔˥ˡ˜ˡ˚ ˔˚˘ˡ˧ ˖ˢˡ˧˥˜˕˨˧˜ˢˡ ˘˩˔˟˨˔˧ˢ˥ ˗ˢ˪ˡ˟ˢ˔˗˜ˡ˚ ˔˚˘ˡ˧ ˦˛˔˥˜ˡ˚ ˔˚˘ˡ˧ ˙˘˘˗˕˔˖˞ ˘ˡ˩˜˥ˢˡˠ˘ˡ˧ ʻ˔˷ˀ˛̂˶ʳˡ˸̇̊̂̅˾ʼ ˣ˥ˢ˙˜˟˘ ˖˜˖ˠʳˣ˟˔˧˙ˢ˥ˠ ˖ˢˡ˧˘˫˧ʳ˔˚˘ˡ˧ ˣˀ˚˥˜˗ ˠ˔ˡ˜ˣ˨˟˔˧ˢ˥ ˦˧˥˔˧˘˚ˬ ˘˩˔˟˨˔˧ˢ˥ ˣ˥˘˙˘˥˘ˡ˖˘ ˟˘˔˥ˡ˜ˡ˚ ˔˚˘ˡ˧ ˖ˢˡ˧˥˜˕˨˧˜ˢˡ ˘˩˔˟˨˔˧ˢ˥ ˗ˢ˪ˡ˟ˢ˔˗˜ˡ˚ ˔˚˘ˡ˧ ˦˛˔˥˜ˡ˚ ˔˚˘ˡ˧ ˙˘˘˗˕˔˖˞ შ 3.CICM ၦଊϸٵ҂ѯ࢝ᄻშ ૮஡ CICM ၦଊϸٵ҂ѯ࢝ᄻϲഌঐրϰӈȞcomponentȟޠѓ૗ඣख़ᇅ྆܉Ԅ ίȈ z ProfileȈखᓄٻң޲ޠঐ΢ၦଊȂᖟδঐ΢ၦਠȃ୒Ԃȃ᜹ր᠍२ Ontologyȃၦ ଊϸٵစᡜȃѠϸٵၦྜȃϸٵ๋౲๊ၦଊഎ᠅ࢃڐϜȂٻң޲ѠпՍॐՍשޠ ProfileȂٯйѠп؛ۢӶϨቅਣ໣ܗ࢑ӵᘉȂ஡೼ٳঐ΢ޠၦଊϸٵяџȂпٽ ڐуঐᡞၽңȄProfile ޠ೪ॏѠпᡲٻң޲Ѡпூژ؂ຯߗঐ΢ޠၦଊ݉ଡ଼Ȃӱ ࣐Ӷᘉᄈᘉᆪၰԇڦၦଊ݉ଡ଼ਣȂѠഇႇ Profile ױԥᜱঐ΢ޠၦଊȂԄȈ୒Ԃȃ ಭᄜȃཽসவဵȃஞጇ…๊ۢဏ੬ੇޠϸٵ๋౲Ȃᡲၦଊ݉ଡ଼ණٽ޲Ѡпණٽ؂ ຯߗٻң޲ሰ्ޠၦଊ݉ଡ଼Ȃٻң޲ڏԥ௢ښ Profile ၈ঐ΢ၦਠޠ᠍ΩȂՅй೼ ٳؾஞޠঐ΢ၦਠѬԇܺӶঐ΢ޠಥᆓ೪റٙαȄ

z Context AgentȞCAȟȈCICM ҂ѯϜޠ Context AgentȞCAȟ஠ߟ॓ೱᇕ໲ϸٵᕘ ძϜᇅၦଊϸٵࣻᜱޠ௒ძӱφȞcontext factorȟȂپԄȈܛӶޠਣ໣ȃܛӶӵᘉȃ ੬ੇ௒ძࣁ୞ȃᆪၰᓝቷȃၸڏਞ૗ȃίၸഀጤ኶๊ӱφȂ೼ٳစҦϸٵᆪၰα ڦூޠᕘძᡑ኶Ѡпձ࣐ Preference Learning AgentȞPLAȟᏱಭᇅ Contribution EvaluatorȞCEȟຠզޠ୥኶অȂᡲᏱಭڸຠզޠ๗ݏ૗ٸᐄϸٵᆪၰϜᕘძӱφ ޠᡑϾᇅึ৥Ȃ୉яᎍ࿌ޠցᘟᇅӲᔗȄܛпȂ௒ძϾȞcontextualizedȟޠၦଊ ݉ଡ଼ණٽΠ؂ӼޠᕘძၦଊȂٻூၦଊϸٵܗ࢑ණٽၦଊ݉ଡ଼ޠႇโϜȂ૗؂ᆡ ྦΠ၍࿌ίٻң޲ޠሰؒᇅցᘟᕘძޠᡑϾȄ

z P-Grid ManipulatorȞPGMȟȈҦܼҐःفٻңΠ P-Grid ᘉᄈᘉϸයԓޠၦਠᓾԇ ࢝ᄻȂ࿌ԇڦؑ՞ঐᡞޠၦਠਣȂሰ्ഇႇؑ՞ঐᡞ CICM ҂ѯαޠ P-Grid Manipulator (PGM)ٿཫ൷ІԇڦዀޠၦਠȂٯй PGM ཽ௄ Profile Ϝڦூٻң޲ ࣁ୞᜹ր Ontology ޠ᠍२Ȃٸᐄ೼ٳ᠍२ӕᑣᒶяӬᎍޠၦଊఽ൑๞ٻң޲ձᒶ ᐆȄӶ P-Grid ޠᓾԇ࢝ᄻίȂؑ՞ঐᡞԇԥഌϸޠཫ൷ᐚȞp-treeȟȂཫ൷ᐚϜޠ ࿾ᘉଅᓄΠؑ՞ঐᡞޠᜌրጇȞidentifierȟпІ၏࿾ᘉ᐀ԥޠၦਠ໷ҭȂ࿌ P-Grid ManipulatorȞPGMȟӶঐᡞҐٙޠཫ൷ᐚα׳ϛژዀޠޑਣȂPGM ൸஡ଊਁᙾଛ Ȟforwardȟяџ๞ڐуޠঐᡞޠ P-Grid ManipulatorȞPGMȟڟֆཫ൷ዀޠޑȂޣ ژഷࡤึ౫ዀޠޑܗӲ༉ณݳཫ൷ژ၏ዀޠޑȂՅᙾଛޠᄈຬࠍ࢑਴ᐄ࿌ਣঐᡞ ၰҦߓȞrouting tableȟ࿌Ϝܛᓾԇޠ࿾ᘉ࣐ᄈຬȂйؑ՞ঐᡞޠၰҦߓϜՎЎཽ

ᓾԇΚঐ೾܂ѫΚᗼཫ൷ᐚ࿾ᘉޠၰ৸ȂпІӤΚᗼཫ൷ᐚίޠ࿾ᘉၰ৸Ȅ࢑ࢉȂ ւң P-Grid ޠϸයԓᓾԇ࢝ᄻȂ૗пϸයԓޠԇڦԥਞޠ௄ٲᘉᄈᘉޠၦଊϸ ٵȂӤਣηআֆڐуঐᡞޠಥᆓ೪റ೏౪૗Ωٿཫ൷᐀ԥዀޠޑޠঐᡞȂܛпȂ ҼႁژΠ Grid Computing ޠძࣩȂ֊ϸයԓޠᘉᄈᘉᆪၰϜঐᡞϟ໣Ӕٵقಜ೏ ౪ၦྜޠዂԓȄ

z Preference Learning AgentȞPLAȟȈCICM ҂ѯϜޠ Preference Learning Agent ཽ਴ ᐄٻң޲ؑԪၦଊᒶᐆޠһܿ๗ݏȂٿ።ᐍ Profile Ϝࣁ୞᜹ր Ontology ޠ᜹ր᠍ २Ȃٯ஡።ᐍᏱಭࡤޠ๗ݏӕӲ㔴ژ Profile ϟϜȂໍՅᏱಭٻң޲ޠၦଊϸٵ୒ ԂȃᇅٻңಭᄜȄഇႇ Preference Learning Agent Ᏹಭޠ೼ٳһܿစᡜȂ૗ӶίԪ ፝ؒίၸޠཫ൷๗ݏఽ൑Ϝ҂ಌϑԇӶޠၦଊϸٵစᡜȂഷࡤܼఽ൑Ϝ୉яಓӬ ٻң޲ಭܓІൊԂޠ።ᐍȄپԄȈस࢑ࣁ୞᜹րޠ Ontology ϸ࣐ॶȃՙȃ՟ȃ՘ȃ يȃ዆ϥঐᆱ࡚Ȃ࿌ٻң޲Ӷॶޠᆱ࡚ϜᒶᐆΠȶਣۧȷޠ᜹ր੬ܓȂࠍѠпഇ ႇ Preference Learning Agent ޠᏱಭ።ᐍ஡ȶਣۧȷ᜹րޠ᠍२ණାȂٯйӶίԪ ሰ्ڐуίၸᆱ࡚ޠၦଊ݉ଡ଼ਣȂ஡ȶਣۧȷ᜹րІڐу᠍२ၷା᜹րޠࣻᜱޠ ၦଊᑣᒶяٿȂпٽٻң޲ᒶᐆȂႁژ҂ಌһܿစᡜޠҭޠȄӱԫȂCICM ҂ѯ ཽತᑗٻң޲ޠၦଊϸٵစᡜȂٯණٽঐ΢Ͼޠၦଊϸٵ݉ଡ଼ٽٻң޲ٻңȄ z Downloading AgentȞDAȟȈ࿌ٻң޲௄ Preference Learning Agent Ᏹಭᑣᒶႇࡤޠ

ၦଊఽ൑Ϝᒶᐆޠᔭ਱ዀޠޑࡤȂDownloading Agent ཽ॓ೱึଛ፝ؒϸٵޠଊਁ ๞ၦଊණٽ޲Ȃٯй॓ೱڸၦଊණٽ޲ޠ Sharing AgentȞSAȟձၦଊίၸޠፚց ྜྷ೾Ȃഷࡤ Downloading Agent ཽӲ᙮๞ٻң޲ҭࠊၦଊණٽ޲ᔭ਱ޠίၸޒᄙȂ пІᆔ౪ೞ೩ѠޠίၸഀጤȄ

z Strategy EvaluatorȞSEȟȈ਴ᐄٻң޲ Profile ϲޠঐ΢ၦଊܗ࢑Ⴒ೪ޠϸٵዂԓȂ ഇႇ Strategy Evaluator ᔤяٻң޲ঐ΢ޠၦଊϸٵ๋౲ȂپԄȈϸٵ๋౲Ѡდϸ Ԛ׈ӓ໡ܺȃഌϸ໡ܺȃԂЅϸٵȃ੬ੇӵᘉϸٵȃ४ഁ४໕๊ዂԓȂดՅၦଊ ϸٵ๋౲஡ᓎ຀ Profile ೪ۢпІਣ໣ӵᘉܗ࢑ϸٵᄈຬ׾ᡑՅᅌϾяϛӤޠϸٵ ๋ ౲ ๞ ϛ Ӥ ޠ ௒ ძ ٻ ң Ȃ ࿌ ᒶ ۢ Π ϸ ٵ ޠ ๋ ౲ ࡤ Ȃ ӕ ױ ๋ ౲ ዂ ԓ ༉ ଛ ๞ Contribution EvaluatorȞCEȟڸ Sharing AgentȞSAȟձϸٵᔭ਱ਣޠࡿዀȄԄԫȂ ࿌ঐᡞϸٵၦଊਣȂ֊Ѡٸᐄঐ΢ಭܓڸཏ᜺ϸٵၦਠᔭ਱Ȅ

z Contribution EvaluatorȞCEȟȈӶ௦Ԟژ፝ؒ޲्ؒϸٵޠଊਁᇅጃۢΠၦଊϸٵ ๋౲ࡤȂCICM ҂ѯޠ Contribution Evaluator ஡໡ۗຠզ፝ؒ޲ޠၦੀȂໍՅ؛ ۢ࢑֐ϸٵၦଊ࿌਱๞፝ؒ޲ȄڐϜȂၦଊϸٵޠዀྦࠍ࢑਴ᐄ፝ؒ޲ᄈϸٵᆪ ၰޠၦଊଔᝧ࡚Յ؛ۢȂՅଔᝧ࡚ޠॏᆘࠍ࢑਴ᐄٻң޲ҐٙޠһܿစᡜІᆪၰ Ϝࣻծঐᡞޠһܿစᡜٿຠզ፝ؒ޲ȄҦܼᘉᄈᘉณጤᆪၰޠ୞ᄙܓȂܛпᆪၰ ϜޠԚসٯߩڿۢޠঐᡞйءԥڿۢޠᆪၰᜌրȂ࢑ࢉȂԫଔᝧ࡚ޠຠզ๗ݏѠ пᇴ࢑Κঐ୞ᄙᆺӬȞslice of momentȟޠຠզ๗ݏȄέӱ࣐ڨ४ܼณጤᆪၰޠ୞ ᄙܓȂঐᡞࢻ୞౦ାȂܛпณݳຠզঐᡞޠᐍᡞଔᝧ࡚Ȟglobal contributionȟȂկ ഇႇԫ୞ᄙᆺӬޠຠզȂѠпூژഷڏхߓܓȃಓӬ࿌ί௒ძޠຠզ๗ݏȄ

z Sharing AgentȞSAȟȈ࿌ၦଊණٽ޲ޠ Sharing Agent ௦Ԟژၦଊ፝ؒ޲्ؒϸٵ ޠଊਁࡤȂһҦ Contribution Evaluator ຠզ፝ؒ޲ޠᆪၰၦଊଔᝧ࡚ȂᙥҦԫၦ ଊଔᝧ࡚ޠ๗ݏڸၦଊ፝ؒ޲ޠ Downloading Agent ձၦଊίၸޠፚցྜྷ೾Ȃഷ ࡤ Sharing Agent ཽӲ᙮๞ٻң޲ҭࠊၦଊ፝ؒ޲ޠίၸޒᄙȂпІᆔ౪ȃϸٵಓ Ӭၦଊଔᝧ࡚ޠίၸഀጤȄ

ΡȃP-Grid ᓾԇ๗ᄻ

ӶၦਠԇڦȞdata accessȟޠഌϸȂҐःف௵ң๗ᄻϾޠϸයԓᓾԇၦਠ๗ᄻ ȞP-GridȟȄP-Grid п๗ᄻϾޠРԓ୞ᄙ࡛Ҵཫ൷ᐚ (P-Grid Tree)ȂؑঐঐᡞഎѬԇԥ ഌϸޠཫ൷ᐚȄॷӒȂؑঐঐᡞޠၰҦߓȞRouting tableȟϜՎЎԇԥΚಣژѫΚᗼᐚ ޠ࿾ᘉȞnodeȟȂ࿌ঐᡞ्ඊ൷׳࢛Κ੬ۢޠၦਠਣȂпࢦၛȞqueryȟޠРԓڦூၦ ਠ Ȃ й ࿌ ࢦ ၛ ޠ л ᗥ Ȟ IDȟ ϛ ឵ ܼ Ԟ ژ ଊ ਁ ޠ ࿾ ᘉ ਣ Ȃ Ԟ ژ ଊ ਁ ޠ ࿾ ᘉ ཽ Ӷ ᙾ ଛ ȞforwardȟяџȂ๞ഷᎭߗҭዀ࿾ᘉޠ࿾ᘉȂԫؐ᡾Ї᙮உ՘ޣژ׳ژҭዀ࿾ᘉ࣐ЦȄ ڐԪȂԥུޠၦਠ्ђΤԫཫ൷ᐚܗ࢑ঔ׾ၦਠਣȂࠍ࢑ϸրпུቩȞinsertȟڸঔ׾ Ȟmodifyȟޠړ኶׈ԚȄშ 4 შᇅ 5 ࠍ࣐შ၍ޠᇴ݃Ȅᖟپᇴ݃Ȃၦଊঐᡞ 6 ԇԥસ Ж 00 ޠၦଊ໷ҭȂ࿌ᆪၰϜڐуޠၦଊঐᡞȞԄȈၦଊঐᡞ 2ȟདྷ्ཫ൷સЖ࢑ 00 ޠၦଊ໷ҭȂԫਣၦଊঐᡞ 2 ൸ཽࢦࣽڐၰҦߓᇰᜌޠၦଊঐᡞϜ࢑֐ԥᓾԇસЖ࣐ 00 ޠၦଊঐᡞȂпშ 5 ٿࣽȂၦଊঐᡞ 2 ϜޠၰҦߓϜԥၦଊঐᡞ 4ȃ6ȂڐϜၦଊঐ ᡞ 6 ࣐સЖ 00 ޠঐᡞȂ࢑ࢉၦଊঐᡞ 2 ѠпࢦၛژસЖ 00 ޠၦଊȄӕѫΚঐپφȂ स࢑ၦଊঐᡞ 2 དྷࢦၛޠસЖ࣐ 11 ໡ᓟޠၦଊ໷ҭȂέӶၦଊঐᡞ 2 ءԥᇰᜌસЖ 11 ޠঐᡞȂԫਣၦଊঐᡞ 2 ཽ஡ࢦၛ༉๞ѻܛᇰᜌޠঐᡞȂࣽࣽ೼ٳၦଊঐᡞ࢑֐ԥ ᇰᜌસЖ 11ȂԄԫ೤ࠍпԫ᜹௱ȂშپϜഷࡤၦଊঐᡞ 2 ࠍཽഇႇၦଊঐᡞ 4 ࢦၛژ ၦଊঐᡞ 5 ޠ໷ҭȄ ඳّϟȂԥΠ P-Grid ᐚޒޠ࢝ᄻࡤȂ௦ίٿഇႇؑ՞ၦଊঐᡞٙαޠၰҦߓၦଊ ༉ሏࢦၛޠଊਁȂӶ P-Grid ཫ൷ᐚϜޠӉեΚঐ࿾ᘉഎѠпึяࢦၛȂؑ՞ঐᡞཽԥ ՍϐܛᇰᜌޠၦଊঐᡞȂԄშ 4 Ϝঐᡞϟ໣ഀ๗ጤ֊࢑ȂӱԫםԚΚᆎϸයԓȃሏಌ ޠၦଊᆪၰȄP-Grid ࢑឵ܼΚঐ׈ӓᘉᄈᘉȞpure P2Pȟޠقಜ࢝ᄻȂϛሰ्໲Ϝԓ ޠ࢝ᄻٿ೏౪೾ଊȄ ௵ң P-Grid л्঩ӱࠍ࢑࣐Πᗘռ჌ Gnutella ೼ᆎϸයԓϸٵ҂ѯᄈᆪၰആԚࢺ ࢻȞfloodingȟпІ༊૗ᄈঐᡞޠ᎒۩ձԥ४ጓ൝ޠኅክ๊ୱᚡȂйԥࣻᜱःفൣ֚ޠ ᄃᡜ኶ᐄࡿя P-Grid ᄈܼঐᡞޠཫ൷ၷ૗ྦጃޠජඬйӶၦଊϸٵޠႇโϜਞ౦ӼΠ [K. Aberer et al., 2003]Ȅ

შ 4.P-Grid ๗ᄻᇴ݃შ[K. Aberer et al., 2003]

შ 5.ᄈᔗშ 4 ϟ P-Grid Routing Table Exchange [K. Aberer et al., 2003] P-Grid ϸයԓޠၦਠᓾԇӶᘉᄈᘉޠᕘძϜȂϑစᅛᅛӵኅހೞٻңȂڐϜ P-Grid л्ޠ੬ՔѠᘫાԄίȈ z ׈ӓϸයԓޠᕘძ࢝ᄻȂءԥ໲Ϝޠժ݉Ꮳ೏౪ӉեϸٵႇโϜܛሰ्ޠၦਠଊ ਁȄ z ᓎᐡϾޠཫ൷ᅌᆘݳȂ࿌࡛ᄻၦਠԇڦ๗ᄻਣȂءԥ੬ۢޠ໲ӬᄈຬȞpeer setȟȂ ԫРݳѠпණାၦਠཫ൷ޠጓ൝пІၦਠཫ൷ޠڽϜ౦ȂٯйѠпණЁقಜᛨۢ ܓȞrobustȟȂӱ࣐࿌ၦଊᆪၰϜޠ࢛ٳၦଊঐᡞѷంȞfailureȟܗ࢑ᚕ໡ณݳණ ٽ݉ଡ଼ਣȂཫ൷ޠᅌᆘݳϛཽೞڿۢӶ࢛ٳ੬ۢঐᡞϟαȄ z ҦܼαΚᘉޠ੬ܓȂٻூԫϸයԓޠၦਠ๗ᄻోᇑΩȞscalabilityȟ஽Ȃ૗ԥਞޠ ཫ൷ᐍঐၦଊᆪၰȄ

ήȃContribution Evaluator

Ґःف஡ၦଊϸٵᆪၰϜঐᡞȞpeerȟؑԪޠၦଊϸٵစᡜഎຝ࣐ Instance-Based LearningȞIBLȟ[T. Mitchell, 1997]ϜޠΚঐኻҐȞsample instanceȟȂٯйᓾԇӶөր ঐᡞޠಥᆓ೪റٙαȂ๊ژၦଊ፝ؒ޲ึя्ؒϸٵޠଊਁȂӶҦၦଊණٽ޲ഇႇ

Contribution EvaluatorȞCEȟຠզၦଊ፝ؒ޲ޠၦଊଔᝧ࡚Ȃϸᒲяᄈϸٵᆪၰԥଔ ᝧޠঐᡞᇅءԥଔᝧޠၦଊ߰ٚ޲Ȟfree riderȟȄ࿌ໍ՘ၦଊଔᝧ࡚ޠຠզਣȂሰ्ഇ ႇ P-Grid ManipulatorȞPGMȟٿཫ൷ P-treeȂᐍঐ P-Grid ࢑ഇႇϸයԓᓾԇ࢝ᄻȂԇ ڦ ᘉ ᄈ ᘉ ᆪ ၰ Ϝ ڐ у ঐ ᡞ ᇅ ၦ ଊ ፝ ؒ ޲ ෇ စ ึ Ң ႇ ޠ ၦ ଊ ϸ ٵ စ ᡜ Ȃ Ҽ ֊ Ҿ ਞ Instance-Based LearningȞIBLȟޠᏱಭዂԓίȂഇႇ P-Grid ٿԇڦኻҐȂໍ՘ၦଊଔ ᝧ࡚ޠຠզȄഷࡤȂसຠզяٿޠၦଊଔᝧ࡚ႁژقಜႲ೪ޠዀྦ໕Ȟthresholdȟࠍ ᇅ၏፝ؒ޲ໍ՘ၦଊϸٵȂٯଅᓄί၏ԪၦଊϸٵစᡜȂЇϟࠍܣ๙፝ؒ޲ޠ्ؒȂ ٯखᓄܣ๙ᇅ၏ঐᡞһܿȄ ᖟΚঐᄃር਱پᇴ݃(Ԅშ 6 ܛұ)Ȃᇴ݃࿌ঐᡞ A Ӷ௦ڨژঐᡞ C ޠ፝ؒଊਁਣȂ Ԅեഇႇ P-Grid ManipulatorȞPGMȟϸයԓᓾԇ๗ᄻȞP-GridȟȂԇڦڐуঐᡞޠၦ ଊϸٵစᡜȞsharing experienceȟȂٯӶຠզঐᡞ C ޠၦଊଔᝧ࡚ࡤȂ؛ۢ࢑֐ᇅঐᡞ C ໍ՘ၦଊϸٵһܿȄॷӒȂঐᡞ C ஡ᕖூڐуঐᡞޠၦଊఽ൑Ȟinformation listȟȂ ԫၦଊఽ൑Ϝखᓄ຀᐀ԥঐᡞ C ܛሰၦਠᔭ਱ޠঐᡞӫ൑ȂӶঐᡞ C ᒶۢΠڐϜޠΚ ঐϸٵঐᡞࡤȂঐᡞ C ᒶᐆΠঐᡞ AȂ࢑ࢉঐᡞ C ޠ Downloading AgentȞDAȟཽึ яၦଊϸٵޠ፝ؒଊਁ๞ঐᡞ A ޠ Sharing AgentȞSAȟȂӶঐᡞ A ௦Ԟژঐᡞ C ޠ፝ ؒଊਁࡤȂঐᡞ A ཽӒః୞ Context AgentȞCAȟᇕ໲ԥᜱ௒ძӱφޠၦଊȂپԄȈ ܛӶޠਣ໣ȃܛӶӵᘉȃ੬ੇ௒ძࣁ୞ȃᆪၰᓝቷȃၸڏਞ૗ȃίၸഀጤ኶๊ӱφȄ ดࡤȂ਴ᐄҭࠊޠᕘძޒᄙᇅٻң޲ޠϸٵൊԂಭᄜȂҦ Strategy EvaluatorȞSEȟᔤ яᎍӬ࿌ίᇅಓӬٻң޲ޠၦଊϸٵ๋౲ȂپԄȈϸٵ๋౲ѠდϸԚ׈ӓ໡ܺȃഌϸ ໡ܺȃԂЅϸٵȃ੬ੇӵᘉϸٵȃ४ഁ४໕๊ዂԓȄ PGM PGM PGM PGM ˔ ˗ ˕ ˙ ˖ ˘ ˚ ˃ ˃˃ ˃˄ ˄ ˄˃ ˄˄ ˦̇̂̅˴˺˸ ˦˻˴̅˼́˺ʳ˘̋̃˸̅˼˸́˶˸ʳ˗ ˦˻˴̅˼́˺ʳ˘̋̃˸̅˼˸́˶˸ʳ˕ ˦˻˴̅˼́˺ʳ˘̋̃˸̅˼˸́˶˸ʳ˖ ʻʳ̊˼̇˻ʳ˾˸̌ʳ̃̅˸˹˼̋ʳ˃˃ʳʼ ˦̇̂̅˴˺˸ ˦˻˴̅˼́˺ʳ˘̋̃˸̅˼˸́˶˸ʳ˖ ˦˻˴̅˼́˺ʳ˘̋̃˸̅˼˸́˶˸ʳ˗ ˦˻˴̅˼́˺ʳ˘̋̃˸̅˼˸́˶˸ʳ˖˄ ʻʳ̊˼̇˻ʳ˾˸̌ʳ̃̅˸˹˼̋ʳ˄˃ʳʼ ˦̇̂̅˴˺˸ ˦ˁ˘ˁʳ˖ ˦ˁ˘ˁʳ˔ ʻʳ˃˄ʳʼ ˦̇̂̅˴˺˸ ˦ˁ˘ˁʳ˕ ˦ˁ˘ˁʳ˙ ʻʳ˄˄ʳʼ შ 6.ၦଊଔᝧ࡚ຠզࢻโұཏშ ӶጃᇰϸٵዂԓࡤȂঐᡞ A ൸ໍΤΠຠզঐᡞ C ၦଊଔᝧ࡚ޠ໧ࢳȂঐᡞ A ሰ ्ၦଊϸٵᆪၰϜڐуঐᡞᇅঐᡞ C ޠၦଊϸٵစᡜȂܗ࢑Ґٙᇅঐᡞ C ෇စԥႇޠ

ၦଊϸٵစᡜȄӶԫΚ໧ࢳϜȂঐᡞ A ཽഇႇ P-Grid ManipulatorȞPGMȟϸයԓᓾ ԇ๗ᄻȞP-GridȟȂԇڦڐуঐᡞޠၦଊϸٵစᡜȞsharing experienceȟȂՅԇڦޠРԓ ࠍ࢑ᙥҦؑঐঐᡞٙαޠၰҦߓȞrouting tableȟȂؑ஼ၰҦߓଷΠखᓄژڐуঐᡞޠ ၰ৸ϟѵȂᗚཽଅᓄঐրޠঐᡞᓾԇΠ঻ٳၦଊϸٵစᡜȞԄშ 4 ᇅშ 5 ϟᇴ݃ȟȄԄ ԫޠ೪ॏԥֆܼؑ՞ঐᡞ૗ഇႇ P-Grid ManipulatorȞPGMȟ൷׳੬ۢঐᡞϟ໣ޠၦଊ ϸٵစᡜȂໍՅձ࣐Սٙຠզڐуঐᡞၦଊଔᝧ࡚ޠٸᐄȄԥΠڐуঐᡞޠၦଊϸٵ စᡜࡤȂঐᡞ A ֊Ѡః୞ Contribution EvaluatorȞCEȟໍ՘ຠզঐᡞ C ޠၦଊଔᝧ࡚Ȃ ഷࡤ؛ۢᇅঐᡞ C ϸٵၦଊܗ࢑ܣ๙ϸٵһܿޠ፝ؒȄ Ҧܼᇕ໲ڐуঐᡞϸٵစᡜޠႇโ࢑൷׳ҭࠊϸٵᆪၰϜԇӶޠঐᡞȂٯйпϸ යԓޠ࢝ᄻԇڦϸٵစᡜȂܛпȂҐःفዂ࠯ூژޠၦଊଔᝧ࡚࢑Κঐ୞ᄙᆺӬȞslice of momentȟޠຠզ๗ݏȂϛӤܼڿۢԓޠᆪၰ࢝ᄻܛຠզޠ๗ݏȂϸයԓ࢝ᄻ૗஋ණ ٽ֊ਣᓎӵޠຠզȂຠզޠႇโϛሰ्ۢဋޠ೪റٿ೏౪ԫຠզ๗ݏȂණٽΠ؂ڏ୞ ᄙܓᇅಌ୞ܓޠ݉ଡ଼ȄՅ Contribution EvaluatorȞCEȟϰӈۢဏᇅ Contribution Evaluator ȞCEȟᅌᆘݳ஡ӗߓٯၐಡᇴ݃ԄίȈ

ߓ 1. Contribution Evaluator Algorithm ᡑ኶ᇴ݃

Requester X : a peer who requests content sharing in the p2p ad-hoc network. Peeri: peers who currently are alive in the p2p ad-hoc network.

DefinitionΚ

IcΰRequester X, PeeriαΚthe information contribution of Requester X evaluated by Peeri.

FwiΚthe weights of features that constitute the information contribution construct, e.g. value of

sharing, rate of rejection and rate of refusal.

Contextual attributes

EnΰRequester Xα: detected location, device and time from the environment. NtΰRequester Xα: detected bandwidth, connection and speed from the network. StΰPeeriα: the selected sharing strategy from the strategy pool by Strategy Evaluator.

VsΰRequester Xα: Value of sharing delivered by X who triggers the computation of En(), Nt(),and

St().

Ӻΰyα: Sigmoid function, the output is a continuous function of its input which is between 0 and 1. i.e., ₁ 1 1 ) ( 0 ≤ + = ≤ _{− y} e y σ

RrjΰRequester X, Peeriα: rate of rejection representing the ratio of content-sharing requests (from

other peers) rejected by X (the records of these rejections are retained at the side of the requestors – Peeri).

RrfΰRequester X, Peeriα: rate of refusal denoting the ratio of peer X’s requests of content sharing

refused by other peers – Peeri ( the records of these refusals are retained at the side of the providers – Peeri).

Contribution Evaluator Algorithm:

Ӷ Contribution EvaluatorȞCEȟॏᆘၦଊଔᝧ࡚Ȟinformation contributionȟޠᅌ ᆘݳϜ:

¦

= n i 1 Tt(X,Peeri) i) Peer (X,

Trj _{(Formula 1 ) ܛॏᆘޠ࢑ܣ๙౦ȞRate of Rejectionȟ࢑ࡿ࿌ঐᡞ i ึ}

я ၦ ଊ ϸ ٵ ޠ ፝ ؒ ࡤ Ȃ ዀ ޠ ঐ ᡞ Ȟ target peerȟ X ܣ ๙ ϸ ٵ ၦ ਠ ޠ Щ ౦ Ȃ Յ ԓ φ

¦

= n i 1 Tt(X,Peeri) i) Peer (X,

Trf _{(Formula 2) ܛॏᆘޠ࢑ೞܣ๙౦ȞRate of RefusalȟȂࠍ࢑ࡿؑԪঐᡞ} X ึяၦଊϸٵޠ፝ؒࡤȂ᎐ڨڐуঐᡞ i ܣ๙ϸٵޠЩ౦Ȅ࢑ࢉȂFormula 1 ࢑ңٿ ᒌ໕ዀޠঐᡞ X ϸٵཏ᜺ޠࡿዀȂңٿॏᆘዀޠঐᡞ X ϛ᜺ཏϸٵၦਠޠЩ౦Ȃ࿌Щ

Function IcΰRequest X, Peeriα returns the information contribution of peer X evaluated by

Peeri in a slice of moment.

InputΚX, a peer who is requesting for information

sub function VsΰRequester Xαreturns the value of sharing delivered by X

1. Start the Context Agent to detect the contextual values between Requester X and Peeri in the p2p be.ipd network.

2. Start the Strategy Evaluator to choose a strategy of content sharing based on the contextual values.

3. Choosing a model of content sharing based on the strategy. 4. Return ıΰEn(Requester X)+Nt(Requester X)α

sub function RrjΰRequester X, Peeriαreturns the ratio of content-sharing requests (from

other peers- Peeri ) rejected by X

1. Start the P-Grid Manipulator to search the transaction experience about X in the p2p network 2. Return

_¦

= n i 1 Tt(X,Peeri) i) Peer (X, Trj

sub function RrfΰRequester X, Peeriαreturns ratio of peer X’s requests of content sharing

refused by other peers – Peeri.

1. Start the P-Grid Manipulator to search the transaction experience about X in the p2p network 2.Return

_¦

= n i 1 Tt(X,Peeri) i) Peer (X, Trf

While receiving the request MSG. from Requester X 1. Vs = Call function VsΰRequester Xα 2.Rrj = Call function RrjΰRequester Xα

౦བାȂߓұዀޠঐᡞ X ၦଊϸٵޠଔᝧ࡚བৰȄࣻᄈՅّȂFormula 2 ࠍ࢑Κঐᒌ ໕ঐᡞ X ΢੊ࡿ኶ޠࡿዀȂңٿॏᆘঐᡞ X ፝ؒϸٵࠔ᎐ڨܣ๙ޠЩ౦Ȃ࿌Щ౦བାȂ ߓұ X ၦଊϸٵޠᐤѭһܿଅᓄЋৰȂᏳयଔᝧ࡚մဤȂܛпؑؑ᎐ڨܣ๙ȄᙥҦ Formula 1&2ȂҐःفѠп௱፤ΚӈٲᄃȂ֊࿌࢛Κঐᡞޠܣ๙౦ȞRate of Rejectionȟ ᇅೞܣ๙౦ȞRate of RefusalȟഎົႇقಜޠႲ೪অȞthresholdȟਣ঑Ȃࠍߓұԫঐᡞ ࣐ၦଊϸٵᆪၰϜޠΚӫၦଊ߰ٚ޲Ȟfree riderȟȄԄԫȂ߰૗؂ԥਞ౦ӵ׳яᄈၦଊ ϸٵᆪၰءԥଔᝧޠၦଊ߰ٚ޲Ȃ၍؛ᘉᄈᘉᆪၰϜȂၦଊϸٵณਞ౦ᇅϛᄈᆏޠୱ ᚡȄԫѵȂӶᙾඳϸٵቌঅȞVs, value of sharingȟਣȂҐःفၽң s םړ኶Ȟsigmoid functionȟ y e y ₋ + = 1 1 ) ( σ ! (Formula 3)ᙾඳ௄ᕘძϜ୏ขژޠ௒ძӱφІᆪၰᡑ኶Ȃᕘძ ᡑ኶ᙾඳژϮܼႮژΚϟ໣ޠள኶3ȄഷࡤȂ๗Ӭϸٵቌঅȃܣ๙౦ᇅೞܣ๙౦Ȃॏᆘ ၦଊଔᝧ࡚Ȃٯйٸᐄ௒ძӱφᇅ User Profile ౱Ңϸٵ๋౲ȂپԄȈ࿌ٻң޲Ӷ੬ۢ ޠऌ׭৥៕ཽൠ୥ᢏਣȂѠп೪ۢၦଊϸٵѬᇅϵѨყ໦ԚসϸٵȂԫਣ௒ძӱφཽ ୏ขҭࠊޠᕘძӵᘉȂ๗Ӭٻң޲ഇႇ Strategy Evaluator ܛ୉ޠ Profile ޠ೪ۢȞi.e. ԂЅϸٵዂԓȟȂ൸Ѡп౱Ң឵ܼٻң޲ޠၦଊϸٵ๋౲ዂԓȂ឵ܼԂЅޠԚস Stw অ ཽೞ೪Ԛ 1ȂЇϟȂߩԂЅޠٻң޲ Stw অཽೞ೪Ԛ 0Ȅ

ѳȃStrategy Evaluator

ਥᏵ٬Ҕޣ Profile ϣޑঁΓၗૻ܈ࢂႣ೛ޑϩ٦ኳԄǴ೸ၸ Strategy Evaluator ȐSEȑᔕр٬ҔޣঁΓޑၗૻϩ٦฼ౣǴٯӵǺϩ٦฼ౣёჄϩԋֹӄ໒ܫǵ೽ϩ໒ ܫǵӳ϶ϩ٦ǵ੝ਸӦᗺϩ٦ǵज़ೲज़ໆ฻ኳԄǴฅԶၗૻϩ٦฼ౣஒᒿ๱ Profile ೛ ۓаϷਔ໔Ӧᗺ܈ࢂϩ٦ჹຝׯᡂԶᄽϯрόӕޑϩ٦฼ౣ๏όӕޑ௃ნ٬ҔǴ྽ᒧ ۓΑϩ٦ޑ฼ౣࡕǴӆע฼ౣኳԄ໺ଌ๏ Contribution EvaluatorȐCEȑک Sharing Agent ȐSAȑբϩ٦ᔞਢਔޑࡰ኱Ǵբࣁᒧ᏷ၗૻϩ٦ޑኳԄޑ٩ᏵǶӵԜǴ྽ঁᡏϩ٦ၗ ૻਔǴջё٩ᏵঁΓಞ܄کཀᜫϩ٦ၗ਑ᔞਢǶԶϩ٦฼ౣኳԄޑ៾ख़߄ӵΠǺ

ߓ 2.ϸٵ๋౲ዂԓ᠍२ߓ ϩ٦฼ౣኳԄ ៾ख़ॶ ֹӄ໒ܫ 1 ೽ϩ໒ܫ 0.8 ό໒ܫϩ٦ 0 ӳ϶ϩ٦ {YesǺ1, NoǺ0} ੝ਸӦᗺϩ٦ {YesǺ1, NoǺ0} ज़ೲज़ໆ 1, ज़ڋࢬໆᆶελ ჹ฻ϩ٦ {YesǺ1, NoǺ0} ϩ٦฼ౣኳԄ ៾ख़ॶ ֹӄ໒ܫ 1 ೽ϩ໒ܫ 0.8 ό໒ܫϩ٦ 0 ӳ϶ϩ٦ {YesǺ1, NoǺ0} ੝ਸӦᗺϩ٦ {YesǺ1, NoǺ0} ज़ೲज़ໆ 1, ज़ڋࢬໆᆶελ ჹ฻ϩ٦ {YesǺ1, NoǺ0} ϩ٦฼ౣኳԄ ϩ٦฼ౣኳԄ ៾ख़ॶ៾ख़ॶ ֹӄ໒ܫ ֹӄ໒ܫ 11 ೽ϩ໒ܫ ೽ϩ໒ܫ 0.80.8 ό໒ܫϩ٦ ό໒ܫϩ٦ 00 ӳ϶ϩ٦

ӳ϶ϩ٦ {YesǺ1, NoǺ0}{YesǺ1, NoǺ0} ੝ਸӦᗺϩ٦

੝ਸӦᗺϩ٦ {YesǺ1, NoǺ0}{YesǺ1, NoǺ0} ज़ೲज़ໆ

ज़ೲज़ໆ 1, ज़ڋࢬໆᆶελ1, ज़ڋࢬໆᆶελ ჹ฻ϩ٦

ჹ฻ϩ٦ {YesǺ1, NoǺ0}{YesǺ1, NoǺ0}

࿌ॏᆘၦଊଔᝧ࡚ਣȂContribution EvaluatorȞCEȟ֊Ѡٸᐄٸᐄ Strategy Evaluator ȞSEȟ౱Ңޠၦଊϸٵ๋౲Ȃᒶᐆαߓޠ᠍२᠍२অȂ୉ഷࡤၦଊଔᝧ࡚ޠ೏౪Ȃߓ 2 Ϝޠ᠍२֊ԓφStw∗ȞFwv*Vs+Fwj*Rrj+Fwf*RrfȟȞFormula 4ȟϜޠ StwȄ

ϥȃPreference Learning Agent

Preference Learning AgentȞPLAȟӶؑԪٻң޲׈ԚၦଊᒶᐆޠһܿϟࡤȂཽ਴ ᐄࣁ୞᜹ր Ontology ޠһܿ๗ݏᙾඳяᄈᔗޠ᜹րӪ໕ઑଳȂϟࡤӕ਴ᐄԫӪ໕ઑଳ џ።ᐍ҂ѯα Profile Ϝ᜹ր᠍२Ȟcategory weightȟȂٻң޲୒Ԃޠ᜹ր᠍२஡ཽᅛᅛ ೞණାȂӱԫȂഇႇԫዂԓѠпᏱಭژٻң޲Ӷ௄ٲၦଊϸٵਣၦଊᒶᐆޠ୒Ԃȃᇅ ٻңಭᄜȄ࢑ࢉȂҐःفւң Preference Learning AgentȞPLAȟᏱಭІತॏ೼ٳၦଊ ϸٵޠһܿစᡜȂٯйȂӶίԪණٽٻң޲፝ؒίၸޠཫ൷๗ݏఽ൑ਣȂٸᐄॏᆘя ޠ᜹ր᠍२Ȃᄈϸٵޠၦଊఽ൑ΡԪᑣᒶȂႁژ҂ಌҭࠊϑԇӶޠၦଊϸٵһܿစᡜ ޠҭޠȂҼ֊ٸᐄࣁ୞᜹ր Ontology ܛॏᆘяٿޠ᜹ր᠍२ȂණٽಓӬٻң޲ಭܓІ ൊԂޠၦଊϸٵཫ൷๗ݏఽ൑Ȅ

ଈ٣ΔPreference Learning AgentΰPLAαᄎ൷࠹ࠐ۞ P-Grid ManipulatorΰPGMα א֗ Context AgentΰCAαࠟଡցٙᙁԵऱ೶ᑇଖΖڇ P-Grid ManipulatorΰPGMαऱ ຝ։Δല൷گࠩࠌشृኙ࣍ᇷಛ堚໢ႈؾऱᙇᖗ࿨࣠Δڇ Context AgentΰCAαຝ։Δ ঞਢ൷گࠩጻሁፖൣቼڂ՗ऱ೶ᑇΔءઔߒঞࠉᖕࠌشृऱᙇᖗ࿨࣠א֗ᛩቼ᧢ᑇଖ ࠐᓳᖞᣊܑᦞૹΔۖᓳᖞऱֱڤঞਢലຍࠄ೶ᑇଖ᠏ངګᣊܑᙇᖗٻၦፖᣊܑٻၦఢ ೄΰcategory vector matrix, CVMαΔຘመڼᣊܑٻၦఢೄૠጩנᄅऱᦞૹࠀᓳᖞᣊܑᦞ ૹΔ່৵٦ڃ墅ࠩࠌشृऱ Profile հխΖۖᣊܑᙇᖗٻၦऱ᥆ࢤፖᣊܑٻၦఢೄऱࡳ ᆠڕ। 3Κ

ߓ 3.᜹րᒶᐆӪ໕឵ܓߓ ᜪձឦ܄ ឦ܄ॶ ᜪ ࠠ ਔۘǵࣽמǵൺђǵ΋૓ǵၮ୏ǵߥ଼ǵۚৎ ࢲ ୏ ࠠ ᄊ ୏ᄊǵᓉᄊ ᜪձឦ܄ ឦ܄ॶ ᜪ ࠠ ਔۘǵࣽמǵൺђǵ΋૓ǵၮ୏ǵߥ଼ǵۚৎ ࢲ ୏ ࠠ ᄊ ୏ᄊǵᓉᄊ ᜪձឦ܄ ᜪձឦ܄ ឦ܄ॶឦ܄ॶ ᜪ ࠠ ᜪ ࠠ ਔۘǵࣽמǵൺђǵ΋૓ǵၮ୏ǵߥ଼ǵۚৎਔۘǵࣽמǵൺђǵ΋૓ǵၮ୏ǵߥ଼ǵۚৎ ࢲ ୏ ࠠ ᄊ ࢲ ୏ ࠠ ᄊ ୏ᄊǵᓉᄊ୏ᄊǵᓉᄊ ˄ˁʳʻ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ˁˁˁʼ ˅ˁʳʻ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ˁˁˁʼ ˆˁʳʻ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ˁˁˁʼ ࢲ୏ᜪձ Poupmphz १ Պ ু኷ ϱᓓ Πϱૡ ఁᓓ ΢Պ ᑿη ଛҹ ˄ ˅ ̀ ˄ ̀ ˄ ˅ ̀ ˄ ́ ਓၯ ຎ᠋ ᎙᠐ ΞΞΞ ΞΞΞ ΞΞΞ ΞΞΞ ˅ ˅ 2/ђڂދᅧ༜ 3/ᐊηπ֝ 4/඲๮ଚ۫ ࢲ୏໨Ҟ 2/঺း 3/QPMPၮ୏׼ !; l/Ǿ ࢲ୏໨Ҟ 2/Ꮬ΍ 3/޸ጪ 4/໚ܴξ ࢲ୏໨Ҟ 2/ႝቹ 3/ॣ኷ 4/ᄽࠩ཮ ࢲ୏໨Ҟ ᜪձӛໆંତ ᜪձӛໆંତ ᜪձӛໆંତ ˄ˁʳʻ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ˁˁˁʼ ˅ˁʳʻ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ˁˁˁʼ ˆˁʳʻ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ˁˁˁʼ ˄ˁʳʻ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ˁˁˁʼ ˅ˁʳʻ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ʿʳ˃ˁ˄ˁˁˁʼ ˍ ˾ˁʳΞ შ 7.ࣁ୞᜹ր Ontology Ґःفዂ࠯Ϝޠ᜹ր᠍२࢑Κঐತᑗޠ๗ݏȂProfile खᓄٻң޲঩ԥޠ᜹ր᠍ २ȂसԫঐᡞӶၦଊϸٵᆪၰϜءԥ௄ٲႇၦଊϸٵޠһܿࣁ୞ਣȂࠍ๞ϡႲ೪অ 0.1 ޠߒۗ᠍२অȂҼ֊஡ܛԥ᜹րޠࣁ୞᜹րຝ࣐Ӥ๊२्ȂԄშ 7ȄᅮࡤȂ࿌ԥུޠ ϸٵһܿࣁ୞౱ҢਣȂစҦᙾඳԚ᜹րᒶᐆӪ໕ᇅ᜹րӪ໕ઑଳȂٯйٸᐄԫུࣁ୞ ॏᆘяུޠ᜹ր᠍२Ȃഷࡤӕᇅ঩ԥޠ᜹ր᠍२ձ҂ྥְ኶Ȃഇႇ҂ྥࡿ኶ڦڎ޲ޠ ҂ྥ҂ְঅձ࣐ঐᡞུޠ᜹ր᠍२Ȃ҂ྥְ኶Ӷ೼၈ޠҭޠ࢑ңٿ҂ྥӱ࣐੬ੇ௒ݸ ܛആԚޠ྄ᆓঅϟኈ៫ȄܛпȂҐःفዂ࠯Ϝޠ᜹ր᠍२Ȟcategory weightȟ࢑Κঐሏ ಌޠತໍ๗ݏȂڐϜє֥пखᓄ᜹ր᠍२অޠРԓЇ࢏ٻң޲ᒶᐆၦଊਣ୒ԂІಭᄜ ޠᐤѭखᓄȄ ܼშ 7 ϜȂ஡ࣁ୞᜹րޠ Ontology ϸ࣐ॶȃՙȃ৐዆ήঐᆱ࡚Ȃ࿌ٻң޲Ӷॶޠ ᆱ࡚Ϝޠၦଊ໷ҭϜڐ֥ԥȶਣۧȷޠ᜹ր੬ܓȂԫਣࠍѠпഇႇ Preference Learning AgentȞPLAȟޠᏱಭ።ᐍ஡ȶਣۧȷ੬ኊঅޠ᠍२ණାȂٯйӶίԪሰ्ڐуίၸᆱ ࡚ޠၦଊ݉ଡ଼ਣȂ஡ȶਣۧȷ᜹րІڐу᠍२ၷା᜹րޠࣻᜱޠၦଊᑣᒶяٿȂпٽ ٻң޲ᒶᐆȂႁژഇႇ PLA ҂ಌႇџॶޠၦଊ໷ҭܛᏱژޠһܿစᡜޠҭޠȄӱԫȂ

CICM ҂ѯཽತᑗٻң޲ޠၦଊϸٵစᡜȂٯණٽঐ΢Ͼޠၦଊϸٵ݉ଡ଼ٽٻң޲ٻ ңȄՅ Preference Learning AgentȞPLAȟ᠍२ᏱಭпІ҂ಌᐤѭစᡜޠϰષۢဏڸᅌ ᆘݳԄίȈ

ߓ 4.Preference Learning Agent Algorithm ᡑ኶ᇴ݃

Պ ኻԄ ΞΞΞ ˅ ˅ ຎ᠋ ΞΞΞ Պ ႝቹ ΞΞΞ ˅ ࢲ୏ᜪձ Poupmphz १ Պ ু኷ ϱᓓ Πϱૡ ఁᓓ ΢Պ ᑿη ଛҹ ˄ ̀ ˄ ̀ ˄ ˅ ̀ ˄ ́ ਓၯ ຎ᠋ ᎙᠐ ΞΞΞ ΞΞΞ ΞΞΞ ΞΞΞ ˅ ˅ ˄ ୠТ ˄ ෝԄ ᓙ࠼Һ୍ ᔞਢᄔा ࣽמǵਔۘǵҶ໕ǵႝቹ ຉǵ࠻ϣǵኻऍǵᓉᄊ Juwᙯඤ ᜪձᒧ᏷ӛໆ )!! 1-2-1-2-1-1-1-2-1-1-1-1-1-1 * შ 8.ࣁ୞᜹ր Ontology შپᇴ݃

CVLΰSrt, CfαΚCategory vector learning machine that calculates

the category weights based on the behavior of peerX.

Srt is a set of selections (rendered by peeri) from the option menu. Cf is the contextualized factor created by Context Agent.

CVMȐȑΚCategory Vector Matrix storing the category vector for peerX. CVMȐSrt, Cfȑ= def ∃n ,,m k isȐC1T1I1, C1T1I2, …., C1T1Ik,

C1T2I1, C1T2I2, .…, C1TmIk, C2T1I1, C2T1I2, …., CnTmIkȑ, n,m,k ∈N .

CnTmIk =def is x1,x2,x3,...,or x14∈CVM(), x1,x2,x3,...,x14∈{0,1}.

ItvΰsαΚItem to vector (transforming the selected options and the contextualized factors

into a category vector matrix). i.e., basing on the file feature set match with the feature of category selected vector.

CWoldΚAn old category weight stored in the user profile.

CWnewΚAn new category weight calculated with user’s file sharing behavior. αΚLearning rate, α∈[0,1], α∈ℜ.

ߓ 4 ۢဏΠ᜹րӪ໕ઑଳޠಣԚϰષȂڐϜ᜹րᒶᐆӪ໕ CnTmIk ۢဏΠӶಒ n ঐ᜹րϜಒ m ঐᆎ᜹ޠಒ k ঐၦଊϸٵ໷ҭϟӪ໕অȄՅؑԪٻң޲ᒶᐆΠၦଊϸٵ ໷ҭࡤȂ஡ٸᐄڐ໷ҭޠ឵ܓ੬ኊᙾඳԚ᜹րᒶᐆӪ໕ȂέһܿစᡜᇅӪ໕ϟ໣ޠ឵ ܓᙾඳዀྦȂࠍ࢑ٸᐄߓ 4 ܛۢဏޠ᜹րӪ໕ઑଳ឵ܓߓ࣐лȂᙥҦᙾඳࡤܛூژޠ ᜹րӪ໕ઑଳȂໍՅѠпॏᆘяུޠ᜹ր᠍२ȄԄშ 8 ࣁ୞᜹ր Ontology შپᇴ݃Ȉ ٻң޲ᒶᐆίၸᓨࡋӉଡ଼ႬኈޠኈбၦଊᇅႬኈ׸ቌڣȂڐᔭ਱ᄣ्Ϝᇴ݃ԫၦଊ໷ ҭӶ৐዆ȃຝ᠚ȃႬኈޠ᜹րᆱ࡚Ϝ᐀ԥऌ׭ȃਣۧȃᓘᄙޠ੬ኊ4пІΚૢ੬ኊᜱᗥ ԆϜޠӅ໤ᜱᗥԆȂӱԫഇႇ Itv ޠϸݚᙾඳ஡ԫၦଊϸٵһܿᙾඳԚ᜹րᒶᐆӪ໕ ூژ(0,1,0,1,0,0,0,1,0,0,0,0,0,0) 5ȂՅᙾඳޠРԓࠍ࢑ٸᐄߓ 4 ᜹րᒶᐆӪ໕឵ܓߓȂ Itv ࠍٸᐄϸٵᔭ਱Ґٙޠ឵ܓᄣ्஡ԫԪһܿᄈᔗژܛ឵ޠ᜹րӪ໕឵ܓȂூژ᜹ր ᒶᐆӪ໕ࡤȂ๗ӬؑΚ์ᒶᐆһܿޠ᜹րᒶᐆӪ໕ CnTmIkȂ֊Ѡூژ᜹րӪ໕ઑଳȄ ഷࡤ Preference Learning AgentȞPLAȟٸᐄུޠ᜹րӪ໕ઑଳ౱Ңུޠ᜹ր᠍२Ȃӕ ஡ԫུޠ᜹ր᠍२ᇅٻң޲঩ԥޠ᜹ր᠍२ձ҂ྥְϾޠ೏౪ȂԄԫȂ֊Ѡᕖூತᑗ ٯᏱಭٻң޲һܿစᡜޠ᜹ր᠍२ȂҼѠᗘռӱ྄ᆓঅܛആԚޠኈ៫ȂԄ Preference Learning Agent Algorithm ܛұȄ

Preference Learning AgentȞPLAȟॏᆘ᜹ր᠍२Ȟcategory weightȟޠᅌᆘݳȄڐ

ϜȂ K M N xi I T C N n M m K k k m n ∗ ∗

¦¦¦

=1 =1 =1 ) (

! (Formula 5) ֊࢑஡᜹րӪ໕ઑଳȞcategory vector matrix, CVMȟϜޠ᜹րӪ໕ᙾඳԚ᜹ր᠍२Ȟcategory weightȟȂڐᅌᆘᡓᒯ࢑џॏᆘؑ՞᜹ ր឵ܓೞᒶңޠԪ኶ഷࡤڦ҂ְঅȂ਍ٿхߓԫ᜹ր឵ܓޠ२्ܓȂ֊ٻң޲ޠൊԂ โ࡚ȄپԄٻң޲ᒶᐆޠΫঐၦଊϸٵ໷ҭϜऌ׭᜹ޠ឵ܓխΠϳঐȂҦ Itv ஡ԫΫ ঐၦଊ໷ҭᙾඳԚၦଊᒶᐆӪ໕ӕഇႇ Compute CVMȂ஡ؑঐҦ itv ᙾඳяٿޠ၍ݏ ӬڂକٿܛூژޠӪ໕ઑଳȂ٦ቅӶ᜹րӪ໕ઑଳϜࠍѠॏᆘяऌ׭឵ܓޠ᜹ր᠍२ ࣐ 0.6Ȟ6/10ȟȂέؑᆎၦଊϸٵ໷ҭཽಓӬӼঐ᜹ր឵ܓȂഷࡤತॏяٿޠӪ໕֊Ѡ хߓٻң޲ޠٻңಭᄜпІٻң୒ԂȂစҦαख़ႇโЇ᙮ॏᆘȂРѠᆘяٻң޲ঐ΢ ತॏޠ᜹ր᠍२ȂٯѠӶίԪණٽၦଊఽ൑ޠਣ঑Ȃഇႇ᜹ր᠍२ᑣᒶၦਠఽ൑Ȃႁ ژ҂ಌٻң޲ၦଊᒶᐆစᡜޠҭޠȄ 4 ڐϜ੬ኊঅڏԥΚঐᓘᄙޠᇮཏᐚȞontologyȟȂԫ ontology खᓄؑ՞੬ኊޠᜱᗥԆңٿᄈᔗᔭ਱ᄣ ्ϜޠၦଊȂҼ֊ᙾඳᔭ਱ᄣ्Ԛ࣐ᒶᐆӪ໕ޠዀྦȄ 5 ٸߓ 4 ੬ኊ᐀ԥਣۧȃऌ׭ȃඉѡȃΚૢȃၽ୞ȃߴୋȃ۩ঢ়Νᆎ੬ኊέؑᆎ੬ኊಡϸ࣐୞ᄙᇅᓘ ᄙڎᆎȂ࢑ࢉ Itv Ҧ੬ኊ ontology Ϝޠ੬ኊӫᆏܗᜱᗥԆȞԄშ 7ȟᙾඳޠ᜹րᒶᐆӪ໕ȂڐϜ᜹րᒶ ᐆӪ໕ٸߓϟΝᆎ੬ኊٸז௷ӗȂέؑᆎ੬ኊٸזϸ࣐୞ᄙᇅᓘᄙޠ᜹րȂூ 14 ᒶ໷ޠ᜹րᒶᐆӪ ໕Ȅ

Preference Learning Agent Algorithm

နȃःفРݳᄃᡜ๗ݏ

CICM ၦଊϸٵ҂ѯΞ࢑ഇႇ JXTA [L. Gong, 2001] ᇅ P-Grid ϸයԓޠၦਠᓾ๗ ᄻٿ໡ึ׈ԚȂٯ஡ϟԋ၇ӶؑΚঐ Peer ϟ ՘୞၇ဋαȄJXTA ࢑Κ৉໡ܺ঩ۗጇޏ ҐޠᘉᄈᘉڟۢȂѠᡲӤΚᆪၰαޠӉե၇ဋȞԄȈКᐡȃPDAȃPCȃժ݉Ꮳ๊ȟձ ၦਠޠһඳᇅᖓᛯȄȞߤᓄΚ஡ᙐ౲ᇴ݃ CICM ϟөዂಣϤ୞ᜱഀȄȟ ࣐ᡜᜍ CICM ҂ѯϟቌঅȂשউໍ՘ΠΚഀ՜ޠᄃᡜȄ೼ٳᄃᡜޠҭޠԥڎঐР ӪȈ (1) ᡜᜍ CICM ၦଊଔᝧ࡚ዂ࠯ᄈܼၦଊϸٵᆪၰϜၦଊ߰ٚ޲ޠ੒ଷ૗ΩȄ

(2) ᡜᜍ Preference Learning Agent (PLA) Ᏹಭٻң޲ၦଊᒶᐆ୒Ԃٿႇᘯၦଊ໷ҭཫ ൷๗ݏޠ૗ΩȄ

ѫѵȂӶᕽਞຠզࡿዀР८Ȉ

(1) ܼ CICM ၦଊ߰ٚ޲ޠ੒ଷ૗ΩᄃᡜϜȂ஡п acceptance rate ձ࣐ຠզ CICM ޠࡿ ዀȞmetricȟȄڐϜȂacceptance rate ңٿΠ၍ᐍঐၦଊϸٵᆪၰޠְᒌโ࡚Ȃٯй ഇႇၦଊ߰ٚ޲ޠ acceptance rate ٿᔯᡜ࢑֐ CICM ૗஋ᒲրяၦଊϸٵᆪၰϜء ԥଔᝧ࡚ޠၦଊঐᡞȄ

(2) Ӷ PLA ٻң޲ၦଊᒶᐆ୒ԂᄃᡜϜȂп ranking difference ձ࣐ຠզ Preference Learning Agent ޠА࡚ȞmetricȟȄй ranking difference ࢑ңٿᒌ໕ Preference Learning Agent ܛᏱಭژޠၦଊٻңᒶᐆಭᄜ࢑֐ಓӬٻң޲ޠಭᄜȄ

ഷࡤޠᄃᡜ๗ݏࠍ࢑ഇႇ೼ٳА࡚ܛ೪ॏޠຠզړ኶ȞEvaluative Functionȟٿᡜ ᜍ CICM ዂ࠯ӶᘉᄈᘉၦଊϸٵޠႇโϟϜ࢑֐૗ණٽΚঐၦଊਞ౦ࢻ೾ܓᇅ҂ᒌܓ ޠၦଊϸٵᕘძȄՅ Evaluative Function ޠϵԓ೪ॏᇅ྆܉ᇴ݃ԄίӗܛұȈ

Function CVLΰSrt,Cfα returns the category weights of peer X learned evolved by

Preference Learning Agent.

InputΚSrt, Cf

1. Attain CWold from the profile of the peerX. 2. Compute CVM = ItvȐSrtȑ∪ItvȐCfȑ.

3. Calculate CWnew = K M N xi I T C N n M m K k k m n ∗ ∗

¦¦¦

=1 =1 =1 ) ( . 4. Return α∗CWold+(1−α)∗CW_new.

Evaluative Function 1 ᙥҦϸٵ౦ѠпΠ၍ᐍঐၦଊϸٵᆪၰޠϸٵโ࡚࢑֐ࣁ ๝ȄԫѵȂпᐍᡞޠܣ๙౦ȃmalicious peer ӶᆪၰϜϸոޠЩ౦ȂᙥҦЩ౦঩ࠍޠಜ ॏ௱፤Ȃ௱ข࢑֐ԥႁژࡋᢏޠၦଊଔᝧ࡚ຠզȄѫѵȂҐःفҼଭᄈঐրޠၦଊ߰ ٚ޲ٿᔯຝ࢑֐Ӷ CICM ዂ࠯ޠϸٵᆪၰϜཽೞܣ๙ϸٵȂٯ෉గ૗Ҧ׊ᢏᇅཌྷᢏڎ ঐِ࡚ٿᔯຝ CICM ዂ࠯Ȅᐍᡞޠ malicious peer rate Щ౦ାਣȂ࿌ Evaluative Function 1 བ௦ߗ 1ȞҼ֊ܣ๙౦ȃmalicious peer rate ௦ߗΚЩΚȟȂпЩ౦௱፤Ȃڎ޲Щپࣻ ߗߓұԥࡋᢏޠୣϸяڏၦଊଔᝧޠঐᡞȄЇϟȂ୆೪ᐍᡞޠ malicious peer rate Щ౦ մਣȂEvaluative Function 1 བ௦ߗ 0Ȅ࢑ࢉȂ࿌ Evaluative Function 1 ԥ૗Ωୣϸяڏ ၦଊଔᝧޠঐᡞਣȂҐःف஡ѠпߗΚؐ௱፤ȂӶ CICM ዂ࠯ޠၦଊϸٵϟίȂ૗ႁ ژᄈ๊ޠၦଊϸٵȄ

Evaluative Function 2 ᙥҦٻң޲ЗϜঅூ๊ಒȞrankingȟ໸זȂᇅ preference learning agent ޠ๗ݏٿЩၷȂࣽࣽ࢑֐ಓӬٻң޲ၦଊᒶᐆޠٻңಭᄜȄՅ ranking difference А࡚ޠԂᚾޠຠզԄ Evaluative Function 3Ȃٸᐄᄃᡜਣޠᄃር๊ಒৰޒݸᇅ ഷৰޒݸޠЩ౦Ȃ؛ۢഇႇ Preference Learning Agent ܛூژޠ௷ז๗ݏޠᓀᇳ౦Ȃ࢑ ࢉȂ࿌ Evaluative Function 36ޠ๗ݏ຺ାਣߓұ preference learning agent Ᏹಭޠ๗ݏ຺ ৰȄ

ᜱܼ Evaluative Function 2ȃ3 ᖟپᇴ݃ԄίȂ࿌ preference learning agent ௷яޠ ၦ ଊ ዀ ޠ ๊ ಒ ࣐ ̋ ɵ ̌ ɵ ̍ Ȃ Յ ٻ ң ޲ ޠ Ӳ 㔴 ޠ ၦ ଊ ๊ ಒ ࣐ ̌ ɵ ̋ ɵ ̍ Ȃ ᙥ Ҧ Evaluative Function 2 שউѠпޤၿ๊ಒৰ࣐ 2Ȃڐॏᆘႇโ࣐ B:(3-2)! -! A:(2-3)! +! C:(1-1)ȄڐϜȂೞմզޠዀޠޑ๊ಒпҔ኶ߓұȂೞାզޠዀޠޑ๊ಒп॓኶ٿߓұȂ ӱԫ࿌๊ಒৰ຺τਣȂࠍߓұ Preference Learning Agent ޠᏱಭ๗ݏᇅٻң޲ᄃርޠᒶ ᐆৰ౵຺τȄՅϑԫ਱پٿᇴȂഷৰޒݸ࣐ 4Ȃ਴ᐄ Evaluative Function 3 Ѡᆘяᓀᇳ ౦࣐ 50%Ȟ2/4ȟȄ

6 ϵԓϜޠ real case of difference ߓұစขၑᄃᡜࡤூژޠ઼๊ৰȂworst case of difference ߓұٸᐄᄃ ᡜ೪ॏޠၦଊ໷ҭঐ኶ॏᆘяഷৰޒݸίޠ઼๊ৰȂڐϜഷৰޠޒݸ࣐ٻң޲ޠᒶᐆᇅ PLA Ᏹಭޠ๗ ݏᓟ׏ᄈ።Ȅ

Evaluative Function 1 = (1 - acceptance rate) ʏ malicious peer rate

Evaluative Function 2 =

_¦

= − n i ki PLAItemRan nki peerItemRa 1 ) (

peerItemRank i, the information item rank provided by the peer. PLAItemRank i, the information item rank provided by PLA.

ΚȃCICM ᄈၦଊ߰ٚ޲ޠ੒ଷ૗Ωϟᄃᡜ

Ӷᄃᡜᕘძޠ೪ۢϜȂშ 9 ᡘұԫಣᄃᡜޠᄃᡜࢻโ྆܉ȄҐᄃᡜዂᔤၦଊϸٵ ᆪၰϜڎᜟၦଊঐᡞϟ໣ໍ՘ၦଊϸٵޠࣁ୞ȂΚᜟ॓ೱၦଊ໷ҭࢦၛȂΚᜟ॓ೱၦ ଊ໷ҭϸٵȂٯйӶᄃᡜޠႇโϟϜࢦၛଊਁޠ೪ۢпङ 100/200/300/400/500/600 Ԫ ޠᄃᡜԪ኶ತໍໍ՘๗ݏޠϸݚȂഷࡤпङ 600 Ԫၦଊࢦၛ໕ޠᄃᡜԪ኶ٿ֖౫ CICM ᄃᡜޠ๗ݏȄڐϜӶؑԪޠᄃᡜҐःف୆೪஡ၦଊ߰ٚ޲ޠ౱Ң౦೪ۢӶ 0.2 ޠЩ౦ 7 Ȃձ࣐ᄃᡜޠஆᙄᕘძໍ՘ၦଊϸٵޠᄃᡜȄዂᔤᄃᡜޠᖃԪ኶Ӕ࣐ 12 ԪȂؑԪޠ ᄃᡜਣ໣ङ౲࣐ 2.5 ϊਣȄѫѵ࣐Πᡜᜍ CICM ૗ၽңӶөᆎၦଊ߰ٚ޲ঐᡞϸոޠ ᕘძϟαȂҐःفӕ஡ၦଊ߰ٚ޲ޠ౱Ң౦೪࣐ۢ 0.6ȂٯйӤαख़ᄃᡜᕘძໍ՘ᄃ ᡜȂڐ๗ݏԄࡤख़ᄃᡜ๗ݏܛұȄ ๻ࡳᇷಛঁ߫ ृขس෷Δข سᇷಛଡ᧯ ެࡳਢܡᇷಛ ։ࠆΔᚏژٌ ࣐ᆖ᧭ ᇷಛ։ࠆृေ ۷ᇷಛಥ᣸৫ ᇷಛᓮޣृ࿇ נᇷಛᓮޣ შ 9.CICM ᄃᡜࢻโ೪ॏ Ӷၦଊঐᡞޠ೪ॏαȂҦܼӶҐःفؑ՞ၦଊঐᡞܛٻңޠᆪၰᕘძ୥኶Ȟᆪၰ ഁ౦ȃࢻ໕४ښȟᇅ௒ძ୥኶Ȟϸٵ๋౲ዂԓȟڐ኶অ੬ܓഎ឵ܼᚕය࠯ޠԥ४໲Ӭ অȂԄߓ 5Ȅ࢑ࢉȂҐःف؛ۢӶዂᔤၦଊঐᡞਣȂпᓎᐡ༅኶ޠРԓ౱ҢؑԪᄃᡜ Ϝၦଊঐᡞ profile Ϝޠᆪၰᕘძᇅ௒ძ୥኶ȂԄԫ߰ѠዂᔤяϛӤঐᡞޠᕘძᇅ՘࣐ ዂԓȂпٽᄃᡜٻңȂᔕ၅ѬпڎᜟȞၦଊ໷ҭࢦၛȃၦଊ໷ҭϸٵȟٿዂᔤᘉᄈᘉ ၦଊϸٵᆪၰޠϛ٘Ȅ 7 Ԝኧॶࣁჴᡍޑଷ೛Ǵ଺Ԝࢎ೛ࢂࣁΑा଺ჴᡍෳ၂Ǵᆶ੿ჴၗૻϩ٦ᆛၡޑᕉნ ೛ۓคᜢǴ٠Ъ٩ᏵԜࢎ೛നࡕӧ೸ၸჴᡍٰ่݀ᔠຎ೭٤ၗૻߡًޣޑঁᡏࢂցૈ ೏ CICM ܌׹๊Ƕ

ߓ 5.ၦଊঐᡞ୥኶឵ܓߓ ᜹ր ୥኶឵ܓ ϸٵ๋౲ዂԓ8 {"All"ȃ"Part"ȃ"Deny"ȃ"Friend"ȃ"Equal"ȃ"LocationYes"ȃ "LocationNo"ȃ"SpeedFast"ȃ"SpeedNormal"ȃ"SpeedSlow"} α༉Ȉ{64ȃ128ȃ256ȃ512ȃ1500} ᆪၰഁ౦ȃࢻ໕४ښ ίၸȈ{512ȃ1500ȃ2000} ߓ 6 ࣐قಜᄃᡜ୥኶ޠ೪ॏᇅඣख़Ȃңٿߓұᄃᡜᕘძ୥኶ޠٻңᇅ೪ۢȄ ߓ 6.قಜᄃᡜ୥኶ ᡑ኶ӫᆏ অ஀ ඣख़ Ⴒ೪অ

Contributive Score natural number ଔᝧ࡚ϸ኶,CICM ޠຠզ๗ݏ 0 Alpha 0..1 ҂ྥࡿ኶,؛ۢ CICM Ᏹಭ።ᐍޠഁ౦ 0.5 Threshold natural number ዀྦঅ,؛ۢ࢑֐ႁژϸٵዀྦ 15 Acceptance Rate 0..1 ϸٵၦଊԚѓޠᐡ౦ 0 Rate of Malicious Peers 0..1 ϛϸٵၦଊޠঐᡞӶϸٵᆪၰϜޠЩ౦ 0.2

ॷӒȂಒΚঐᄃᡜ(Free Rider Rate = 0.2Ȃၦଊࢦၛ໕ 630 Ԫ)ޠ๗ݏ࢑п acceptance rate ٿ֖౫ᐍᡞၦଊᆪၰޠၦଊϸٵ౦ȂҼ֊ᢏࢦᐍঐၦଊϸٵᆪၰ࢑֐ࣁ๝Ȅڐᄃ ᡜ኶ᐄϸݚԄშ 10Ȅ

Average Acceptance/Rejection Rates

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1 2 3 4 5 6 7 8 9 10 11 12 Experiment Ratio Acceptance Rate Rejection Rate

Free Rider (Preset)

შ 10.ᐍᡞϸٵ౦҂ְ኶

8 ϸٵ๋౲ዂԓᇴ݃ȂAllȈϛ४ۢӉեዂԓпഷାޠ๋౲᠍२ϸٵȂPartȈϸࢻࢻ໕ϸٵȂDenyȈϛϸٵȂFriendȈ ԂЅϸٵȂEqualȈᄈ๊ၦଊ໕ϸٵȂLocationYes/NoȈٸᐄ SpeedFast/Normal/Slo ೪ۢࢻ໕ഁ࡚ϸٵȄ

Ҧშ 10 Ϝޠ X ໆٿࣽѠпࣽژؑΚԪᄃᡜ҂ְϸٵ౦ޠഎႁژΝԚпαȂၦଊ ϸٵޠࣁ୞ӶᆪၰϜ૗໸ւໍ՘ȂٯґӱӼΠ CICM ޠຠզقಜՅᏳयႇ࡚ଔᝧ࡚ຠ ໕Ȃٻூၦଊঐᡞϟ໣ณݳϸٵၦଊၦྜޠ౫ຬȂέ௄ؑԪᄃᡜߒ෉ژᄃᡜҒ෉ޠႇ โϟϜ CICM എ૗ԥਞޠຠզၦଊঐᡞȄՅйᐍᡞޠܣ๙౦എାܼᄃᡜ೪ॏޠၦଊ߰ ٚ޲ޠЩ౦ 0.2ȂпᐍᡞޠᢏᘉٿࣽȂ௦ڨ౦ӶήԪӲᘫᗎ༗ጤޠᗎ༗ίȂ҂ְϸٵ ౦૗஋ᛨۢޠᆱࡼӶ 75%ѿѢޠЩ౦ȂйؑԪᄃᡜϟϜณ྄ᆓঅޠ౱ҢȄ࢑ࢉȂҐः فߒؐ௱፤ഇႇ CICM ޠຠզȂၦଊϸٵࣁ୞ӶᆪၰϜϬѠ໸ւȃࣁ๝ޠໍ՘Ȃйԥ ܣ๙ၦଊ߰ٚ޲ӶᆪၰϜໍ՘ၦଊϸٵޠ૗ΩȄԫѵȂҐःفӶп Evaluative Function 1 ޠॏᆘٿϸݚၦଊϸٵ౦ᇅၦଊ߰ٚ޲Щ౦ϟ໣ޠᜱ߾ȂҦ Evaluative Function 1 ழ Τ 12 Ԫᄃᡜ҂ְޠϸٵ౦Ȟ0.734003724ȟᇅᄃᡜ೪ॏޠၦଊ߰ٚ޲Щ౦Ȟ0.2ȟூژ ޠ኶অ࣐ 1.329981Ȟ(1-0.734003724)/0.2ȟȂᐍᡞՅّȂCICM ܛܣ๙ޠၦଊঐᡞ࢑ା ܼၦଊ߰ٚ޲ޠЩ౦Ȅп׊ᢏޠِ࡚ٿࣽ CICM ޠၦଊϸٵࣁ୞Ȃၦଊ߰ٚ޲࢑Ѡп ೞ CICM ႇᘯ௭ޠȄ ௦ίٿȂҐःفӤኻٻң acceptance rate ԫА࡚ٿᒌ໕Ȃӕഇႇཌྷᢏޠِ࡚ȂҦ ঐրၦଊ߰ٚ޲ޠၦଊϸٵ௒םٿໍΚؐᡜᜍ CICM ࢑֐઎ޠ؈๙ΠၦଊϸٵᆪၰϜ ޠၦଊ߰ٚ޲Ȃڐᄃᡜ኶ᐄᇅϸݚშԄშ 11 ܛұȈ

Free Rider Rejection Ratio

0 20 40 60 80 100 120 140 1 2 3 4 5 6 7 8 9 10 11 12 Experiment Fr ee Ri der S iz e 0 0.2 0.4 0.6 0.8 1 Rej ec tion Rati o

Free Rider Size Times of Rejection Rejection Rate შ 11.ၦଊ߰ٚ޲҂ְೞܣ๙౦ Ҧშ 11 ޠ X ໆٿࣽѠпࣽژؑΚԪᄃᡜ҂ְೞܣ๙౦ޠഎႁژΝԚѿѢȂშϜ Y ໆѿᜟ࣐ؑԪᄃᡜၦଊ߰ٚ޲ޠ኶໕ᄈᔗӶ൑Ԫᄃᡜޠߞ఩შȂՅ Y ໆѢᜟ࣐ؑԪ ᄃᡜၦଊ߰ٚ޲ೞܣ๙ޠЩ౦ᄈᔗӶ൑Ԫᄃᡜޠ׸ጤშȄ࢑ࢉȂၦଊϸٵޠࣁ୞ଷΠ ӶᆪၰϜ૗໸ւໍ՘ȂՅйᄈܼᄃᡜϜܛ೪ॏޠ 20%ޠၦଊ߰ٚ޲എ૗ԥਞޠೞџ ଷȂ12 Ԫޠᄃᡜ҂ְܣ๙౦࣐ 77.6%ȄՅйၦଊ߰ٚ޲ೞܣ๙ޠЩ౦ӶήԪӲᘫᗎ༗

ጤޠᗎ༗ίȂၦଊ߰ٚ޲ޠೞܣ๙౦ԥ٘ᅛᛈЁژ 90%ѿѢޠᗎ༗ȂйؑԪᄃᡜϟϜ ณ྄ᆓঅޠ౱ҢȄ࢑ࢉȂϛ፤пᐍᡞܗ࢑ঐրޠᢏᘉٿࣽȂҐःفഎ૗௱፤ഇႇ CICM ޠຠզȂၦଊϸٵࣁ୞ϛկӶᆪၰϜϬѠ໸ւȃࣁ๝ޠໍ՘ȂՅй૗ԥਞޠܣ๙௭ၦ ଊ߰ٚ޲ӶᆪၰϜໍ՘ၦଊϸٵࣁ୞Ȅ Յ࣐Πᡜᜍ CICM ૗ၽңӶөᆎၦଊ߰ٚ޲ঐᡞϸոޠၦଊϸٵᆪၰᕘძϟαȂ Ґःفӕ஡ၦଊ߰ٚ޲ޠ౱Ң౦೪࣐ۢ 0.6Ȃໍ՘ᄃᡜขၑȂڐᄃᡜ኶ᐄԄϸݚშ 12 ܛұȄҦშ 12 ޠ X ໆٿࣽѠпࣽژၦଊ߰ٚ޲ޠತᑗ኶໕ȂშϜ Y ໆ࣐ၦଊ߰ٚ޲ ޠತᑗၦଊϸٵ౦ᇅತᑗܣ๙౦ȂԫԪᄃᡜစᐤΠ 630 ԪޠၦଊϸٵȂΚӔԥ 352 ՞ ၦଊ߰ٚ޲ȂڐϜԥ 89 ՞ၦଊ߰ٚ޲ೞڐуၦଊঐᡞ௦ڨϸٵȂ࢑ࢉᐍөၦଊ߰ٚ޲ ޠထᡞԥ 74.7%Ȟ(352-89)/352ȟޠঐᡞೞܣ๙௭ȂέӶತᑗܣ๙ϸٵԪ኶ޠႇโϜȂ ҂ְޠܣ๙౦ङӶ 83.5%Ȅ࢑ࢉȂпαख़ڎಣᄃᡜ೪ॏȞfree rider rate = 0.2, 0.6ȟᇅᄃ ᡜ኶ᐄ๗ݏٿࣽȂҐःف૗ᡜᜍഇႇ CICM ໍ՘ၦଊϸٵ૗ԥਞޠܣ๙௭ϸٵᆪၰϜ ө᜹࠯ϸҁޠၦଊ߰ٚ޲ӶᆪၰϜໍ՘ၦଊϸٵࣁ୞Ȅ ˙̅˸˸ʳ˥˼˷˸̅ʳ˔˶˶˸̃̇˴́˶˸˂˥˸˽˸˶̇˼̂́ʳ˥˴̇˼̂ ˃ˁ˃ ˃ˁ˅ ˃ˁˇ ˃ˁˉ ˃ˁˋ ˄ˁ˃ ˄ˁ˅ ˄ ˅ˉ ˈ˄ ˊˉ ˄˃˄ ˄˅ˉ ˄ˈ˄ ˄ˊˉ ˅˃˄ ˅˅ˉ ˅ˈ˄ ˅ˊˉ ˆ˃˄ ˆ˅ˉ ˆˈ˄ ˙̅˸˸ʳ˥˼˷˸̅ʳ˦˼̍˸ ˥˴ ̇˼̂ ˔˶˶̈̀̈˿˴̇˸˷ʳ˔˶˶˸̃̇˴́˶˸ʳ˥˴̇˸ ˔˶˶̈̀̈˿˴̇˸˷ʳ˥˸˽˸˶̇˼̂́ʳ˥˴̇˸ შ 12.ၦଊ߰ٚ޲ತᑗೞ௦ڨ౦/ೞܣ๙౦

ΡȃPLA ᏱಭಓӬٻң޲ၦଊᒶᐆಭᄜϟᄃᡜ

๻ ࡳ ௽ ᐛ ֺ ෷ ։ ಻ Δ ข س ᇷ ಛ ႈ ؾ ࠉ ᖕ ଺ ๻ ࡳ ข س ᇷ ಛ ႈ ؾ ᛀ ᧭ ʳˣ̅˸˹˸̅˸́˶˸ʳ ˟ ˸˴̅́˼́˺ʳ˔ ˺˸́̇ ಝ ᒭ ʳˣ̅˸˹˸̅˸́˶˸ʳ ˟˸˴̅́˼́˺ʳ˔ ˺˸́̇

შ 13.Preference Learning Agent ᄃᡜࢻโ೪ॏ Average Rejection Rate =

შ 13 ࣐ Preference Learning Agent ᄃᡜޠᄃᡜࢻโ྆܉ȂҭޠӶܼᄃᡜዂᔤၦଊ ϸٵᆪၰϜޠၦଊঐᡞӶᒶᐆ P-Grid ཫ൷ၦଊ໷ҭޠ๗ݏਣȂPreference Learning Agent Ᏹಭٻң޲ᒶᐆ୒Ԃᇅᑣᒶၦଊ໷ҭ๗ݏޠࣁ୞ȄӶؑԪޠ Preference Learning Agent ዂᔤᄃᡜϸրໍ՘Πङ 1000/5000/10000 Ԫޠၦଊᒶᐆࣁ୞Ȃңٿଌጜ Preference Learning Agent ٿႲӒଌጜ Preference Learning Agent ።ᐍ Category Ontology ϟϜޠ੬ ኊ឵ܓ᠍२অȞfeature weightȟȂйᏱಭޠ҂ྥࡿ኶೪࣐ۢ 0.59ȂഷࡤпႲӒଌጜ 10000 Ԫޠ Preference Learning Agent ٿ֖౫ໍ՘ᄃᡜ֖౫๗ݏȂٯйٸᐄԫଌጜ๗ݏໍ՘ 1000 ԪޠᄃᡜขၑȂᡜᜍᏱಭޠ๗ݏ࢑֐ಓӬ࿌ߒዂᔤٻң޲ၦଊᒶᐆޠၦଊ໷ҭϸ ପȂഷࡤޠᄃᡜ๗ݏп 10000 Ԫޠࣁ୞ขၑձ࣐ၦਠϸݚޠ኶ᐄȄ

ڐϜȂዂ࠯ଌጜࣁ୞೪ۢΠڎᆎၦଊ໷ҭޠϸପȄಒΚ᜹࠯ޠၦਠڐၦଊ໷ҭϸ ପпڎঐၦଊ੬ኊȞfashionȃtechnologyȟ࣐л्੬ኊȄಒΡ᜹࠯ޠၦਠڐϸପпϥঐ ၦଊ੬ኊȞclassicȃgeneralȃsportȃhealthȃhomeȟ࣐л्੬ኊȄዂᔤᄃᡜޠԪ኶Ӕ࣐ 12 ԪȂPreference Learning Agent ޠଌጜᇅዂ࠯ޠขၑؑԪ౱Ң 10 ঐၦଊ໷ҭٿขၑȂ ؑԪޠᄃᡜਣ໣ङ౲࣐ 1.5 ϊਣȄ

ॷӒȂᄃᡜޠಒΚഌӌ҇໹Ӓໍ՘ Preference Learning Agent ൊԂዂ࠯ޠଌጜȄଌ ጜޠၦਠԥڎᆎ᜹࠯Ȃϸր࣐пڎঐȞ᜹࠯ΚȟȃϥঐȞ᜹࠯Ρȟၦଊ໷ҭ੬ኊ࣐лޠ ၦଊ໷ҭϸପ౱Ңၦଊ໷ҭٿଌጜ Preference Learning Agent ٿ።ᐍ category ontology ϜޠΝঐ੬ኊ᠍२Ȟfeature weightȟȂٯйໍ՘ 12 ԪޠᄃᡜȂٸᐄᄃᡜޠ኶ᐄҐःف Ⴒ෉૗ᡜᜍ࿌ Preference Learning Agent ८ᄈϛӤޠ੬ኊൊԂਣ૗ԥਞޠᏱಭяٻң ޲ޠ੬ኊᒶᐆൊԂȂڐᄃᡜϸݚშߓԄშ 14ȃ15 ܛұȄ Preference Learning (1) 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Dyn ami c Sta tic Dyn ami c Sta tic Dyn ami c Sta tic Dyn ami c Sta tic Dyn ami c Sta tic Dyn ami c Sta tic Dyn ami c Sta tic

Fashion Technology Classic General Sport Health Home

Ontology Category Feature Weight 1TH 2TH 3TH 4TH 5TH 6TH 7TH 8TH 9TH 10TH 11TH 12TH შ 14.҂ְ category ontology ੬ኊ᠍२׸ጤშȞ᜹࠯Κȟ 9 ԫ೏ޠ҂ྥࡿ኶ߓ PLA ತᑗᙠޠစᡜᇅᏱಭུޠ୒Ԃစᡜϟ໣ޠЩ౦ȂҼѠߓұ PLA Ᏹಭޠഁ౦Ȃ Ґःفᇰུۢᙠޠစᡜϟ໣२्ܓٸኻȂࢉӶԫ೪ۢ 0.5 ޠ኶অ࣐Ґःفޠᄃᡜ୆೪ձ࣐ᄃᡜܛңȂ կӶᄃርᕘძϜԫᏱಭഁ౦࢑Ѡпٸᐄঐᡞޠ΢ੀ੬፵ՅԥܛϛӤȄ

Preference Learning (2) 0 0.05 0.1 0.15 0.2 0.25 0.3 Dy na m ic St at ic Dy na m ic St at ic Dy na m ic St at ic Dy na m ic St at ic Dy na m ic St at ic Dy na m ic St at ic Dy na m ic St at ic

Fashion Technology Classic General Sport Health Home

Ontology Category Feature Weight 1TH 2TH 3TH 4TH 5TH 6TH 7TH 8TH 9TH 10TH 11TH 12TH შ 15.҂ְ category ontology ੬ኊ᠍२׸ጤშȞ᜹࠯Ρȟ Ҧშ 14 ᇅშ 15 ޠ X ໆٿࣽѠпࣽژؑΚԪᄃᡜ҂ְၦଊ໷ҭ੬ኊ᠍२ȂშϜ Y ໆ࣐ؑԪᄃᡜၦଊ໷ҭ੬ኊ᠍२অᄈᔗӶ൑Ԫᄃᡜޠ׸ᘉϟαȄშ 14 ࣐᜹࠯ΚၦਠȂ ڐ๗ݏϸҁᗎӪ fashionȃtechnology ڎঐ੬ኊঅȂშ 15 ࣐᜹࠯ΡၦਠȂڐ๗ݏϸҁᗎ Ӫ Classicȃgeneralȃsportȃhealthȃhome ϥঐ੬ኊঅȂӶშαޠ׸ጤѠпఽྀޠࣽя ณ፤࢑᜹࠯Κܗ࢑᜹࠯ΡޠጤშȂτयαޠϸҁᗎ༗എᇅ঩ۗޠၦਠ᜹࠯ࣻಓȂйؑ ԪᄃᡜϟϜณ྄ᆓঅޠ౱ҢȄӱԫȂഇႇЩ౦኶ᐄޠᇴ݃ȂҐःفᡜᜍΠഇႇ Preference Learning Agent ޠၦଊᒶᐆ୒ԂᏱಭਞݏၗԄᄃᡜ೪ॏܛႲ෉Ȅ

Average Rank Difference (1)

-15 -10 -5 0 5 10 15 20 25 1 2 3 4 5 6 7 8 9 10 11 12 Experiement Ranki ng Average underRestimateRank Average overRestimateRank Average rankingDifference

3rd degree polynomial regression(rankingDifference)

Average Rank Difference (2) -15 -10 -5 0 5 10 15 20 25 30 1 2 3 4 5 6 7 8 9 10 11 12 Experiement Ranking Average underRestimateRank Average overRestimateRank Average rankingDifference

3rd degree polynomial regression(rankingDifference)

შ 17.Preference learning ҂ְ๊ಒৰȞ᜹࠯Ρȟ

ഷࡤȂ਴ᐄ Evaluative Function 2ȃ3 ޠ ranking difference А࡚Ȃᡜᜍ Preference Learning Agent ࢑֐૗Ҕጃޠ௷זȞsortingȟڎᆎၦਠ࠯ᄙޠၦଊᒶᐆኻԓȄҦშ 16 ᇅშ 17 ޠ X ໆٿࣽѠпࣽژؑΚԪᄃᡜ҂ְ๊ಒৰȂშϜ Y ໆ࣐ؑԪᄃᡜ๊ಒৰϸ ኶ᄈᔗӶ൑Ԫᄃᡜޠ׸ᘉϟαȄშ 16 ࣐᜹࠯ΚޠၦਠȂშ 17 ࣐᜹࠯ΡၦਠȂڐᆎڎ ᆎၦਠ᜹࠯ޠ๗ݏпήԪӲᘫᗎ༗ጤϸݚȂڐ๊ಒৰߴࡼᛨۢȂڐঅङ౲Ӷ 18~21 ѿ ѢၾٗȂйؑԪᄃᡜϟϜณ྄ᆓঅޠ౱ҢȄ ഷࡤȂ਴ᐄ Evaluative Function 3 ԫΚຠզړ኶ٿຠ໕๊ಒৰޠ๗ݏ࢑֐૗ᡜᜍ PLA ޠᏱಭڏԥ࡚߭ȄڐϜȂҦܼؑԪᄃᡜܛ౱Ңޠၦଊ໷ҭ኶໕࣐ 10 ঐ໷ҭȂܛ пڐഷৰޒݸޠ๊ಒৰঅ࣐ 50ȂέҦ 12 ԪޠᄃᡜϜҐःفѠпூژ๊ಒৰ҂ְঅङ ౲࣐ 19.77366667Ȃ࢑ࢉ Evaluative Function 3 ࣐ 0.3954733334Ȟ19.77366667/50ȟȂҼ ֊ PLA ԥ 61.5%ޠҔጃ౦ȄࢉҐःفᡜᜍ Preference Learning Agent ܛᏱಭٻң޲ޠ ၦଊᒶᐆ୒Ԃڏԥ 6 Ԛпαޠ߭З૗ಓӬᄃርޠ୒ԂᒶᐆȄ

Ӄȃ๗፤

ӧόዴۓޑคጕᗺჹᗺၗૻϩ٦ᕉნϐΠǴᙖҗҁࣴزගрȨ௃ნϯၗૻଅ᝘ࡋ ኳࠠȐContextualized Information Contribution Model, CICMȑȩǴճҔ P-Grid ϩණԄᓯ ӸࢎᄬᓯӸҬܰ࿶ᡍǴаीᆉၗૻঁᡏӧၗૻϩ٦ᆛၡύޑၗૻଅ᝘ࡋǴ჋၂ှ،ၗ ૻ ϩ ٦ ύ ޑ ၗ ૻ ߡ ً ޣ Ȑ free riderȑ ޑ ᝼ ᚒ Ƕ ӕ ਔ Ǵ ӧ Ȩ ௃ ნ ϯ ၗ ૻ ଅ ᝘ ࡋ ኳ ࠠ ȐContextualized Information Contribution Model, CICMȑȩޑᗺჹᗺၗૻϩ٦ޑၸำϐ ύǴԵቾډၗૻࠔ፦ǵၗૻሽॶǵਔਏ܄ǵϩ٦฼ౣǵ௃ნӢનȐcontext factorȑǵᆛ ၡᓎቨϷၩڀȐdeviceȑޑਏૈȐperformanceȑ฻ӢηǴ٠ЪճҔ΢ॊӢηǴаϩණ ԄीᆉȐdistributed computingȑࣁ୷ᘵǴຑ՗ঁᡏჹܭԜၗૻϩ٦ᆛၡޑၗૻଅ᝘ࡋ Ȑinformation contributionȑǶ٠Ъ೸ၸ Preference Learning AgentȐPLAȑᏢಞঁᡏၗૻ ᒧ᏷ޑಞᄍᆶୃӳǴӆᙖҗѳ౽Ҭܰᒧ᏷࿶ᡍޑᐒڋǴගٮঁᡏঁΓϯޑၗૻᒧ᏷Ƕ ਥᏵ CICM ޑჴᡍ่݀ᡉҢǴҁፕЎ܌ගϐࣴزБݤૈᆢ࡭᏾ᡏϩ٦ᆛၡޑၗૻ ϩ٦ࢲ๎Ǵ٠Ъૈ࣬྽Ӧ׹๊ၗૻߡًޣޑفՅӸӧӧᆛၡύǶќѦǴځҭԖᏢಞ٬ Ҕޣၗૻᒧ᏷ୃӳٰၸᘠၗૻ໨Ҟཛྷ൨่݀ޑૈΚǶඤقϐǴҁፕЎගٮΑ΋ঁૈԖ ਏ࿶ᔼคጕᗺၗૻϩ٦ᆛၡǵዴߥঁᡏϐ໔ၗૻϩ٦ჹᆀޑᐒڋǴ຾Զബ೷рόӕܭ ໺಍ޑ୘཰ኳԄǶ! ҂ٰࣴزஒᝩុׯ๓ PLA ᏢಞޑБݤǴаගଯྗዴࡋǶёૈБݤхࡴፓ᏾Ꮲಞೲ ౗ǵׯ๓ཥᙑٿᅿᏢಞ࿶ᡍҔܭᏢಞ่݀ޑКٯǵ೛ۓόӕޑᜢᗖӷख़ा܄ٰளډ׳ ᆒጏޑ่݀ΕගܹᜢᗖӷϩᜪޑБݤǵࡌҥᓉᄊ܈ࢂ୏ᄊޑӷ৤ǵ٩Ᏽ੝ۓޑЎҹဂ ܈ၗૻ໨Ҟဂ຾Չϣ৒ޑှ݋аளډᜢᗖӷȐtermsȑࡌҥᜢᗖӷ৤฻Ƕ

୥ՄНᝧ!

1. [ ྇ ් സ , 2001] ྇ ් സ , P2P Ȃ ᆪ ၰ ޠ τ Ӥ ϟ ၰ ȉ , ኶ ՞ ᢏ ᄇ ޲ , http://www.digitalobserver.com/71-80/78/kevin.htm, ಒ Ν Ϋ Υ ෉ 2001 Ԓ 7 У 4 СȄ

2. A. Oram, editor. Peer-to-Peer: Harnessing the Power of Disruptive Technologies. O'Reilly & Associates, March 2001.

3. [T. Mitchell, 1997] T. Mitchell, Machine Learning, McGraw Hill, 1997.

4. [C. Courcoubetis et al., 2002] C. Courcoubetis and P. Antoniadis, Market Models for P2P Content Distribution, International Workshop on Agents and Peer-to-Peer Computing, Bologna, Italy, 2002.

5. [E. Adar et al., 2000] E. Adar and B. A. Huberman, Free Riding on Gnutella, Internet Ecologies Area Xerox Palo Alto Research Center, 2000.

Fickas, Z. Segall, When Peer-to-Peer comes Face-to-Face: Collaborative Peer-to-Peer Computing in Mobile Ad-hoc Networks, 2001 International Conference on Peer-to-Peer Computing, Linköpings, Sweden, 2001.

7. [K. Aberer et al., 2001] K. Aberer, Z. Despotovic, Managing Trust in a Peer-2-Peer Information System, Proceedings of the Tenth International Conference on Information and Knowledge Management, McLean,VA, 2001.

8. [K. Aberer et al., 2003] K. Aberer, P. Cudre-Mauroux, A. Datta, Z. Despotovic, M. Hauswirth, M. Punceva, R. Schmidt, P-Grid: A Self-organizing Structured P2P System, ACM SIGMOD Record, 32(2), September 2003.

9. L. Gong, JXTA: A Network Programming Environment. IEEE Internet Computing, 5(3), pp. 88-95, May/June 2001.

10. [N. Daswani et al., 2002] N. Daswani, H. Garcia-Molina, and B. Yang, Open problem in file sharing of P2P, Stanford University, Stanford CA 94305, 2002.

11. [N. Daswani et al., 2003] N. Daswani, H. Garcia-Molina, and B. Yang, Open Problems in Data-Sharing Peer-to-Peer Systems, Stanford University, Stanford CA 94305, http://www-db.stanford.edu, 2003.

12. [N. Guarino, 1997] N. Guarino, Understanding, Building and Using Ontologies, International Journal of Human-Computer Studies, 46(2-3), pp. 293-310, February 1997.

13. [R. Jasper et al., 1999] R. Jasper and M. Uschold, A Framework for Understanding and Classifying Ontology Applications, Proceedings of the 12th Workshop on Knowledge Acquisition, Modeling and Management, Voyager Inn, Banff, Alberta, Canada, 1999.

14. [S. Schapp et al., 2002] S. Schapp and R. D. Cornelius,U-Commerce, Accenture, 2002.

15. [S. D. Kamvar et al., 2003] S. D. Kamvar, M. T. Schlosser, H. Garcia-Molina, EigenRep: Reputation Management in P2P Networks, WWW2003, Budapest, Hungary, ACM, 2003.

ߤᓄΚȈCICM ϟөዂಣϟϤ୞ᜱഀ

Ґ፤Нпτ࠯ᗋޑϜЗȞshopping mallȟ࣐௒ძ೪ॏȂӶᘉᄈᘉၦଊϸٵณጤᆪ ၰϟίȂ੒ຳ޲ܗ୧ঢ়Ѡпϸٵ֊ਣޠᗋޑၦଊȂپԄȈ୧ࠣၦଊȃኅ֚ၦଊȃߵ᎜ ࣁ୞ܗ׸ቌڕ๊Ȃй୧ۺϟ໣ҼѠᖟ՘๋౲ᖓ࿘ޠӬձࣁ୞ȞپԄȈၯ୧ঢ়ޠᇰӤћȂ ᖓӫ੒ຳޠ׸ԛȟȄҐ፤Н஡ؑΚঐ੒ຳ޲ְຝ࣐ঐᡞȞpeerȟȂঐᡞܛ׹ᅌޠِՔ࢑ ၦଊණٽ޲η࢑ၦଊίၸ޲Ȃٯഇႇ P-Grid ժ݉Ꮳ൷׳ژҭࠊᗋޑϜЗϲԇӶޠၦଊ Ȟ݉ଡ଼ȟණٽ޲Ȅܼпίᇴ݃ϟშ A1ȞCICM قಜࢻโშȟȂ୆೪ঐᡞ X ࣐ၦଊϸٵ ޲ȃঐᡞ Y ࣐ၦଊ፝ؒ޲ٿձᗀ၍Ȅ ˖˴̇˸˺̂̅̌ʳ ˢ́̇̂˿̂˺̌ ˖̂́̆̇˴́̇̆ʳ˜́˼̇˼˴˿˼̍˼́˺ ˤ̈˸̅̌ʳ˥˸˶˸˼̉˼́˺ ˦̇̅˴̇˸˺̌ʳ˦˸˿˸˶̇˼́˺ ˣˀ˚̅˼˷ʳˡ˸̇̊̂̅˾ ʻ˔˷ˀ˛̂˶ʳˡ˸̇̊̂̅˾ʼ ˣ̅̂˹˼˿˸ ˄ ˖̂́̇˸̋̇̈˴˿˼̍˼́˺ ˖̂́̇̅˼˵̈̇˼̂́ʳ ˖˴˿˶̈˿˴̇˼́˺ ˖̂́̆̇˴́̇̆ʳ˜́˼̇˼˴˿˼̍˼́˺ ˖̂́̇˸̋̇̈˴˿˼̍˼́˺ ˣ̅˸˹˸̅˸́˶˸ʳ˟˸˴̅́˼́˺ ˣ̅̂˹˼˿˸ ˤ̈˸̅̌˼́˺ ˤ̈˸̅̌ʳ˥˸˿̌˼́˺ ˄ ˅ ˉˁ˶ ˆ ˇ ˋˁ˵ ˈˁ˴ ˋˁ˴ ˌ ˄˃ ˄˄ ˣ˸˸̅ʳˬ ˊ ˈˁ˵ ˉˁ˵ ˉˁ˴ ˖˴̇˸˺̂̅̌ʳ ˢ́̇̂˿̂˺̌ ˣ˸˸̅ʳ˫ ˄˅ შ A1. CICM قಜࢻโშ CICM ҂ѯϟ࡛ҴΞ࢑ٸᐄԫ CICM قಜࢻโშᇅ௒ძȄڐϜᇅөقಜϰӈϟ໣ ޠقಜࢻโᇴ݃ԄίȈ ࿌ঐᡞ Y ཫ൷ၦଊ໷ҭٯӪঐᡞ X ึяၦଊϸٵޠ፝ؒȂࢻโᇴ݃ԄίȈ 1. ࿌ঐᡞ XȞϸٵ޲ȟȃYȞ፝ؒ޲ȟః୞ CICM ਣȂ࣐قಜᡑ኶ߒۗϾȄԄȈprofile

Ϝޠঐ΢೪ۢȃၦଊϸٵޠᐤѭစᡜȃcategory ontology Ϝޠ᜹ր᠍२๊೪ۢঅȄ 2. ঐᡞ Y ୏ขԥᜱ௒ძӱφ10ᇅᆪၰᕘძӱφ11ޠᕘძᡑ኶Ȃٯٸᐄঐᡞ Y ঐ΢ሰ 10 قಜٻңޠ௒ძӱφ឵ܓԄίख़໷ҭȄԫ೏ޠ௒ძӱφࡿޠ࢑Π၍፝ؒ޲࿌ίޠၦଊཫ൷๋౲ȂԄ࢑֐ԥঐᡞထ ಣཫ൷ȃၦଊ໷ҭ੬ኊঅȞfeatureȟ๊௒ძཫ൷೪ۢȂпؒӶϸٵޠႇโϜ؂щϸޠߓႁ፝ؒ޲࿌ίޠᄙ࡚ᇅདྷݳȄ 11 قಜٻңޠᆪၰӱφ឵ܓԄίख़໷ҭȂԫ೏ޠᆪၰӱφࡿޠ࢑Π၍፝ؒ޲ᄈܼၦଊϸٵ޲ڐᆪၰᓝቷȞαʝί༉ȟ ޠτϊᒶᐆܗ࢑४ښȃ

ؒȂ๗Ӭᕘძᡑ኶ۢဏࢦၛޠၦਠ໷ҭȄ 3. ঐᡞ Y ٸᐄঐ΢ሰؒȂഇႇ P-Grid ၦଊϸٵᆪၰึя੬ۢၦଊ໷ҭޠၦଊ፝ؒȄ 4. (a) P-Grid ҂ѯٸᐄ፝ؒ޲ޠࢦၛȂཫ൷ၦଊϸٵᆪၰϜಓӬ፝ؒ޲्ؒޠၦଊ໷ ҭȄ (b) P-Grid ҂ѯٸᐄ፝ؒ޲ޠࢦၛȂӲ᙮ҭࠊၦଊϸٵᆪၰϜಓӬ፝ؒ޲ޠܛԥၦ ଊϸٵ޲ޠ໷ҭఽ൑Ȅ 5. (a) ঐᡞ Y ԇڦ categoryOntology.xml Ϝޠ᜹ր᠍२Ȃٯٸᐄԫ᜹ր᠍२२ུ௷ӗ ၦଊ໷ҭޠ໸זȄ (b) ঐᡞ Y ூژҦ᜹ր᠍२௷яޠၦଊ໷ҭఽ൑ࡤȂᒶᐆൊԂޠၦଊ໷ҭȂٯӪ ঐᡞ X ึяၦଊίၸޠ፝ؒȄ (c) ٸᐄঐᡞ Y ޠ፝ؒȂP-Grid ҂ѯ஡ίၸޠଊਁ༉ሏژঐᡞ X ٙαȂҦঐᡞ X ޠଊਁ௦ԞᏣ௦ԞଊਁȄ 6. ٸᐄঐᡞ Y ᒶᐆޠൊԂ໸זȂစҦ Preference Agent Ᏹಭٻң޲ޠၦଊᒶᐆൊԂȂ ٯቹΤ categoryOntology.xml ঔ׾ڐ᜹ր᠍२Ȅ 7. (a) Ӷঐᡞ X ޠଊਁ௦ԞᏣঐᡞ Y ԞژଊਁࡤȂڦூقಜᇅঐ΢೪๊ۢقಜᡑ኶Ȅ (b) ః୞௒ძх౪΢Ȃᇕ໲௒ძӱφ12ᇅᆪၰᕘძӱφ13๊୥኶অȄ 8. ٸᐄؐ᡾ 7ȃ8 ܛᇕ໲ޠ୥኶অᇅᡑ኶Ȃᒶᐆၦଊϸٵޠ๋౲Ȅ 9. ঐᡞ X ٸᐄܛॏᆘяٿޠၦଊϸٵ๋౲Ȃॏᆘ፝ؒ޲ޠၦଊଔᝧ࡚Ȅ 10. ٸᐄၦଊଔᝧ࡚ޠॏᆘ๗ݏȂӲ᙮๞፝ؒ޲ YȂ࢑֐ໍ՘ၦଊϸٵȄ࢑ޠၘ׈Ԛ ၦଊϸٵޠ፝ؒȂЇϟࠍ֐Ȅ 22/!ഷࡤȂ஡ၦଊϸٵޠһܿစᡜखᓄژ profile ᔭϟϜȂٯ๗؃һܿȄ! 12 قಜٻңޠ௒ძӱφ឵ܓԄίख़໷ҭȄԫ೏ޠ௒ძӱφࡿޠ࢑Π၍ϸٵ޲࿌ίޠၦଊϸٵ๋౲Ȃ Ԅ࢑֐ԥӵᘉϸٵȃԂЅϸٵȃܣ๙ϸٵȃᄈ๊ϸٵ๊௒ძ४ښȂпؒӶϸٵޠႇโϜ؂щϸޠߓႁ ϸٵ޲࿌ίޠᄙ࡚ᇅདྷݳȄ 13 قಜٻңޠᆪၰӱφ឵ܓԄίख़໷ҭȂԫ೏ޠᆪၰӱφࡿޠ࢑Π၍ϸٵ޲ᄈܼࢻ໕௢ښޠᒶᐆܗ ࢑४ښȄ