Daidaita rukuni
Daidaita rukuni wata dabara ce da ke daidaita horon cibiyar sadarwar jijiyoyi ta hanyar daidaita fasalulluka a cikin ƙananan ƙungiyoyin tashoshi, mai zaman kansa ga kowane misali.
Dubawa
It matters because, unlike Batch Normalization, it works well even when batches are tiny.
Zurfafa nutsewa
Matsakaicin daidaitawa yana kiyaye lambobin suna gudana ta hanyar hanyar sadarwa mai ma'ana mai kyau, wanda ke sauri da daidaita horo. Batch Normalization yana yin haka ta hanyar ƙididdige ma'ana da bambance-bambancen kowane fasali a cikin duka ƙaramin tsari, amma hakan yana sa ya zama mara ƙarfi lokacin da batches ƙanƙanta ne, tunda ƙididdiga ta zama hayaniya da rashin dogaro. Daidaita rukuni, wanda Wu and He suka gabatar a cikin 2018, yana cire batch daga lissafin gaba ɗaya. Ga kowane misali guda ɗaya, yana raba tashoshi zuwa ƙayyadaddun adadin ƙungiyoyi, sannan ya daidaita kowace ƙungiya ta amfani da ƙimar wannan misalin kawai. Saboda lissafin bai taɓa dogara da wasu misalan a cikin tsari ba, aikin yana tsayawa tsayin daka ko rukunin yana riƙe da hotuna 32 ko ɗaya kawai, yana sa ya shahara a ganowa, rarrabuwa, da ayyukan hangen nesa mai nauyi.
Fahimtar Fasaha
Ƙungiya ta al'ada tana ƙididdige ma'ana da bambance-bambance akan ma'auni na sararin samaniya da kuma kan tashoshi a cikin kowace ƙungiya, kowane samfurin. Sannan ya daidaita zuwa sifili ma'ana da bambance-bambancen raka'a kuma yana amfani da sikelin kowane tashoshi da aka koya (gamma) da motsi (beta). Yana ƙaddamar da wasu tsare-tsare: tare da rukuni ɗaya ya zama Layer Normalisation, kuma tare da tashoshi ɗaya a kowace ƙungiya ya zama Daidaita Misali. Ƙididdigar rukuni shine hyperparameter, sau da yawa saita zuwa 32.
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 Daidaita Ƙungiya
Daidaita rukuni yana tsayawa zaɓin zaɓi a duk inda batches dole ne su zama ƙanana, kamar gano babban ƙuduri da rarrabuwa, 3D da ƙirar bidiyo, da ƙarancin horon ƙwaƙwalwar ajiya. Hakanan an haɗa shi cikin gine-ginen gine-ginen da ake amfani da su sosai kamar U-Nets a cikin samfuran watsawa. Kamar yadda samfura ke girma da kuma matsa lamba na ƙwaƙwalwar ajiya yana tura masu girma dabam, masu zaman kansu na yau da kullun, Tsarin Rukuni tsakanin su tare da Layer Norm, wataƙila za su kasance tsoffin tubalan ginin, tare da ci gaba da bincike kan hybrids da zaɓin daidaitawa.
Aiwatar da Gaskiyar Duniya
Gano abu da rarrabuwar misali (misali, Mask R-CNN salon salo) waɗanda aka horar da ƙananan batches kowane-GPU.
Kasusuwan baya na U-Net a cikin masu samar da hoto na watsawa, inda Rukuni na al'ada ya daidaita ma'auni.
3D da cibiyoyin sadarwar bidiyo inda babban ƙwaƙwalwar ajiya ke amfani da ƙarfin juzu'i zuwa ɗaya ko biyu.
Kyakkyawan daidaita manyan nau'ikan hangen nesa akan ƙayyadaddun kayan aiki inda ƙananan batches ke sa ƙididdigar Batch Norm ta zama abin dogaro.
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 Daidaita rukuni ke taimakawa kuma inda mafi sauƙi hanyoyin 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 Group Normalization 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
Daidaita Ladan Rukuni a cikin RLHF
Tambayoyin da ake yawan yi
What is Group Normalization?
Daidaita rukuni wata dabara ce da ke daidaita horon cibiyar sadarwar jijiyoyi ta hanyar daidaita fasalulluka a cikin ƙananan ƙungiyoyin tashoshi, mai zaman kansa ga kowane misali. Yana da mahimmanci saboda, ba kamar Batch Normalization ba, yana aiki da kyau koda lokacin da batches ƙanƙanta ne.
Menene babban fa'idar Daidaita Rukuni akan Daidaita Batch?
Saboda Rukuni na al'ada yana ƙididdige ƙididdiga a kowane misali maimakon a ko'ina, ba a lalata shi da ƙananan batches, sabanin Batch Norm.
A kan menene Ma'anar Rukuni ya lissafta ma'anarsa da bambancinsa?
Yana rarraba tashoshi zuwa rukuni kuma yana daidaita kowace ƙungiya ta amfani da ƙimar misali guda ɗaya kawai da wuraren sarari, yin watsi da sauran rukunin.
Daidaita rukuni ya zama Daidaita Layer a cikin wane yanayi na musamman?
Tare da rukuni ɗaya, duk tashoshi an daidaita su tare kowane misali, wanda shine daidaitaccen Layer Layer.
Daidaita rukuni ya zama Daidaita Misali a cikin wane yanayi na musamman?
Idan kowane rukuni yana riƙe tashoshi ɗaya, kowane tasho yana daidaita shi da kansa akan kowane misali, wanda shine Daidaita Misali.
Bayan daidaitawa, menene Al'adar Ƙungiya ke aiki don maido da sassaucin wakilci?
Kamar sauran yadudduka na daidaitawa, Ƙungiya ta al'ada ta haɗa da ma'auni da za'a iya koyo da sigogin motsi don hanyar sadarwar zata iya gyarawa ko daidaita daidaitattun kamar yadda ake buƙata.