Graph Neural Networks
Hanyoyin sadarwa na jijiyoyi (GNNs) samfura ne waɗanda ke koyo kai tsaye akan bayanan da aka tsara jadawali - nodes ɗin da aka haɗa ta gefuna - ta hanyar wucewa da tara bayanai tsakanin maƙwabta.
Dubawa
They matter because much of the real world is relational: social networks, molecules, road maps, and recommendation systems are all graphs that grids and sequences cannot naturally represent.
Zurfafa nutsewa
GNN yana aiki ta hanyar wucewar saƙo. Kowane kumburi yana farawa da sifa mai siffa, kuma a cikin kowane Layer kowane kumburi yana tattara saƙonni daga maƙwabtansa, yana haɗa su da aiki mara canzawa kamar jimla, ma'ana, ko max, kuma yana sabunta nasa wakilci. Stacking L yadudduka yana ba da damar bayanai su yaɗa L hops a cikin jadawali, don haka haɗaɗɗen kulli na ƙarshe yana nuna mafi girman unguwarsa, ba kawai haɗin kai tsaye ba. Bambance-bambancen sun bambanta ta yadda suke tarawa: Graph Convolutional Networks suna amfani da matsakaicin matsakaicin maƙwabci, samfuran GraphSAGE da tara ƙayyadaddun adadin maƙwabta don haɓakawa, kuma Hanyoyin Hankali na Hotuna suna koyon ma'auni don haka kumburin yana halartar maƙwabta masu mahimmanci. Ƙunƙwalwar da aka koya, gefuna, ko haɗaɗɗen jadawali gabaɗaya sannan ciyar da rarrabuwa, koma baya, ko shugabannin tsinkaya.
Fahimtar Fasaha
Ma'anar kadarar ita ce rashin daidaituwa: jadawali ba shi da tsari na kumburi, don haka matakin tarawa dole ne ya samar da sakamako iri ɗaya ba tare da la'akari da yadda aka jera maƙwabta ba - don haka jimla, ma'ana, ko max maimakon ƙayyadaddun aiki. Ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun saƙon yana da yawa: tara saƙon da ke wucewa da yawa da kowane kumburin kumburi yana haɗuwa zuwa ƙima ɗaya, yana wanke bambance-bambance masu amfani. Wannan yana ɗaukar zurfin aiki kuma yana motsa haɗin haɗin gwiwa da daidaitawa.
Dabarun Tasiri
Shawarwari masu haske
Yana taimaka muku keɓance bayyanannen da'awar fasaha daga harshen talla.
Kudin da kasafin kuɗi
Kuna iya yin mafi kyawun tambayoyin aiwatarwa kafin kashe kuɗi ko lokaci.
Ƙungiya da aikin aiki
Ƙungiyoyin da ke da fahimtar juna suna yin mafi kyawun samfura, manufofi, da yanke shawara na koyo.
Makomar Graph Neural Networks
GNNs sune tsakiyar AI na kimiyya. DeepMind's GNoME ya yi amfani da su don tsinkayar miliyoyin tsayayyen sabbin sifofi na crystal, kuma samfuran yanayi kamar GraphCast suna wakiltar duniya azaman jadawali don yin hasashen sauri fiye da na'urar kwaikwayo ta kimiyyar lissafi. Bincike yana magance scalability zuwa jadawalai-bakin biliyoyin, hanyoyin sadarwa masu zurfi waɗanda ke ƙin yin laushi, da kuma alaƙar da ke tsakanin GNNs da Transformers (waɗanda ke da mahimmanci a kan cikakkun hotuna masu alaƙa). Yi tsammanin haɗin kai tare da ƙirar tushe da haɓaka amfani a cikin gano magunguna da kimiyyar kayan aiki.
Aiwatar da Gaskiyar Duniya
Hasashen kaddarorin kwayoyin halitta da guba a cikin gano magunguna ta hanyar ɗaukar kwayoyin halitta azaman nodes da haɗin sinadarai azaman gefuna.
Ƙaddamar da shawarwari a kamfanoni kamar Pinterest, inda PinSage ke koyon sakawa akan jadawali na abubuwa da hulɗar mai amfani.
Gano zamba da halasta kuɗaɗe ta hanyar gano alamu masu shakku a cikin jadawali na mu'amala tsakanin asusu.
Hasashen yanayi da zirga-zirga, kamar a cikin GraphCast da ƙirar hanyar sadarwar hanya waɗanda ke wakiltar wurare azaman nodes masu alaƙa.
Hatsari & Tsare-tsare
Ƙungiyoyi daban-daban na iya amfani da kalmar iri ɗaya daban, don haka ayyana iyaka da wuri.
Alamomi na iya yin kama da ƙarfi yayin da aikin zahirin duniya bai yi daidai ba.
Yin watsi da ingancin bayanai da tsare-tsaren kimantawa galibi yana haifar da sakamako mara ƙarfi.
Taswirar Hanya
Fara da ma'anar harshe a sarari na sakamakon da kuke buƙata.
Zaɓi ma'aunin nasara ɗaya da yanayin gazawa ɗaya kafin gwaji.
Gudun ƙaramin matukin jirgi tare da bayanan wakilci, ba saitin demo da aka goge ba.
Takaddun inda Zane-zanen Neural Networks ke taimakawa kuma inda hanyoyin mafi sauƙi suka fi kyau.
Ci gaba da Bincike
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Graph Neural Networks quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Jagora na gaba
Hanyoyin Sadarwar Jijiya
Tambayoyin da ake yawan yi
What is Graph Neural Networks?
Hanyoyin sadarwa na jijiyoyi (GNNs) samfura ne waɗanda ke koyo kai tsaye akan bayanan da aka tsara jadawali - nodes ɗin da aka haɗa ta gefuna - ta hanyar wucewa da tara bayanai tsakanin maƙwabta. Suna da mahimmanci saboda yawancin duniyar gaske tana da alaƙa: cibiyoyin sadarwar jama'a, kwayoyin halitta, taswirorin hanya, da tsarin shawarwari duk jadawali ne waɗanda grid da jeri ba za su iya wakilta ta zahiri ba.
Menene ainihin tsarin lissafin mafi yawan hanyoyin sadarwar jijiyoyi?
GNNs na aiki ta hanyar sa kowane kumburi ya tattara saƙon daga maƙwabtansa, sa'an nan kuma sabunta nasa wakilci, maimaitu a cikin yadudduka.
Me yasa aikin makwabci na GNN ya zama maras bambanci?
Tunda babu wani tsari na canonical ga maƙwabta na kumburi, ayyuka kamar jimla, ma'ana, ko max suna tabbatar da fitarwa bai dogara da yadda aka jera maƙwabta ba.
Menene 'over-smoothing' ke nufi a cikin zurfin GNNs?
Tsara yawan saƙon da ke isar da saƙo yana sa kowane wakilcin kumburi ya haɗu zuwa ƙima ɗaya, yana goge bambance-bambance masu amfani.
Me ya bambanta Cibiyar Kula da Hankali ta Graph (GAT) da ainihin Graph Convolutional Network (GCN)?
GATs suna ba da maƙwabta maƙwabta waɗanda aka koyo, suna barin kumburi ya jaddada waɗanda suka fi dacewa maimakon auna su daidai.
A cikin GNN da aka yi amfani da kwayoyin halitta, menene nodes da gefuna ke wakilta?
Molecules jadawali ne ta dabi'a: atom ɗin nodes ne kuma abubuwan haɗin da ke haɗa su gefuna ne, wanda shine dalilin da ya sa GNNs ya yi fice wajen tsinkayar kaddarorin kwayoyin.