Kwenzekeni
Ucwaningo olushicilelwe ku-Knowledge and Information Systems lwethula indlela yokufunda ejulile yokuthola ubungozi benkontileka ehlakaniphile ngokuhlaziya ukulandelana kwe-opcode yobude obuguquguqukayo kusukela ku-Ethereum transaction replays. Imodeli, eyakhelwe ekwakhiweni kweChordMixer nezindlela zokugcinwa, igwema ukunqamula imikhondo yokubulawa, ithola ukunemba okungu-93.5% kuzo zonke izinhlobo eziyisikhombisa zobungozi.
Abacwaningi abavela eNyuvesi yase-Guangzhou bashicilele ucwaningo ku-Knowledge and Information Systems oluchaza ngemodeli entsha yokufunda ejulile yokutholwa sengozini yenkontileka ehlakaniphile. Ukuqamba okusha okuyisisekelo kuwukusetshenziswa kokulandelana kwe-opcode yobude obuguquguqukayo okususelwe ku-Ethereum transaction replays, esikhundleni sekhodi yomthombo emile noma ukulandela ngomkhondo okuncishisiwe.
Imodeli isebenzisa i-architecture ye-ChordMixer, indlela yokunaka ye-neural enwebekayo eyenzelwe ukulandelana kobude obuhlukahlukene, ehlanganiswe nendlela yokulondoloza evela emndenini we-Retentive Network. Le ndlela yokulungiswa okukabili ivumela imodeli ukuthi ithwebule kokubili amasiginesha okuziphatha omhlaba wonke okuhlasela kanye nokuqhuma kwemiyalelo okusolisayo kwasendaweni.
Ithimba lilinganise imodeli ngokumelene nezinhlobo eziyisikhombisa ezihlukene zokuba sengozini, okuhlanganisa ukungena kabusha nokuncika kwesitembu sesikhathi. Ucwaningo lubika ukunemba okungama-93.5% kanye ne-F1-score yama-90.6%, izitholi ezisebenza kahle kakhulu ezincike ekufakweni kobude obugxilile. Idatha ithathwe kubahloli be-blockchain basesidlangalaleni njenge-Etherscan, ebonisa ukuziphatha kwenkontileka yomhlaba wangempela.
Imininingwane yomthombo: bioengineer.org ↗
Kungani kubalulekile
Lolu cwaningo lubhekana nobuthakathaka obubalulekile besakhiwo kumathuluzi okuphepha e-blockchain azenzakalelayo: ukulahlekelwa kolwazi lapho ukulandelwa kobude obuguquguqukayo kuphoqelelwa okokufaka kwenethiwekhi ye-neural yobude obugxilile. Ngokulondoloza umongo ogcwele wokwenziwa kwenkontileka ehlakaniphile, imodeli ingakwazi ukuhlonza ubungozi obufihlekile njengokungena kabusha kanye nokuchichima okuphelele okungase kugejwe ukuhlaziya okuvamile okumile noma ukuhlaziya okuncishisiwe okuguquguqukayo. Lokhu kubalulekile emkhakheni we-DeFi, lapho izinkontileka ezingaguquleki nezigidigidi zezimpahla zenza amaphutha ezokuphepha abe yinhlekelele. Umsebenzi ubonisa ukuthi ukucutshungulwa kobude obuguquguqukayo bomdabu kungadlula izindlela ezijwayelekile zokunamathisela/zokuncisha, kunikeze isisekelo esiqinile sokuqapha okuqhubekayo, okuzenzakalelayo kokuvikeleka kwezimpahla eziku-on-chain.
Izitholi zokufunda zomshini ezivamile zokuvikela i-blockchain ngokuvamile zinciphisa ukulandelana okude kwe-opcode ukuze kulingane okokufaka kwemodeli yobude obugxilile, okungenzeka kulahle yona kanye imiyalelo eveza ubungozi. Le ndlela entsha iqeda lokho kulahlekelwa kolwazi, inikeza umbono ophelele wokwenziwa kwenkontileka.
Kuhlelo lwe-decentralized finance (DeFi) ecosystem, lapho izinkontileka zingenakuguqulwa futhi ukuxhashazwa kungadla izigidigidi, amathuluzi azenzakalelayo angahlaba umkhosi ngokunembile ubungozi ngaphambi noma ngemva kokuthunyelwa abalulekile. Le modeli inikeza indlela ebheke okuqhubekayo, ekuqashweni kokuphepha kwesikhathi sangempela kwezimpahla ezikuchungechunge.
Ucwaningo luphinde lube nemithelela ebanzi ekusetshenzisweni kwe-AI, okubonisa ukuthi izakhiwo eziklanyelwe ubude obumangelengele, bokulandelana bomhlaba wangempela zingadlula izindlela ezijwayelekile zokupheda kanye nezindlela zokucubungula ngaphambilini kwezinye izizinda ezifana ne-genomics nedatha yenzwa.
I-Interactive Mechanism: Indlela Esebenza Ngayo Ngempela
Hlola ubuchwepheshe obuyisisekelo ngemuva kwalokhu kuthuthukiswa ngokuhlanganyela.
Which component of an AI application is the machine-learning model itself?
Ongakubuka ngokulandelayo
Gada ukukhishwa komthombo ovulekile noma ukuhlanganiswa kwezentengiso kwaleli payipi lokuthola ubude obuguquguqukayo. Bheka ukutholwa yizimiso ze-DeFi noma amafemu okuhlola ukuvikeleka njengesendlalelo sokubuyekezwa komuntu. Qaphela uma ngabe lesi sakhiwo siba yindinganiso yeminye imisebenzi yokuhlaziya ukulandelana kobude obuguquguqukayo ku-AI, njengedatha ye-genomic noma yenzwa.
Bheka ukukhishwa kwekhodi yemodeli noma izisindo ezinkundleni ezifana ne-GitHub ukuze uhlole ukukhiqizwa kabusha kanye namandla okutholwa ngumphakathi.
Gada amafemu okuhlola ukuvikeleka kanye namaphrothokholi e-DeFi ukuze ahlanganiswe le ndlela yokuthola ubude obuguquguqukayo kuzitaki zazo zokuphepha ezijwayelekile.
Qaphela uma ngabe inhlanganisela ye-ChordMixer neyokugcina iba yisithenjwa esijwayelekile sokuphatha ukulandelana kobude obuguquguqukayo kwezinye izinhlelo zokusebenza ze-AI.