Ọdịda gradient
Mmụda gradient bụ usoro njikarịcha nke na-ebuga ihe atụ n'ezie na-agbada na mperi dị ala, otu obere nzọụkwụ n'otu oge.
Nchịkọta
It is how learning happens once backpropagation has computed the gradients.
Ime miri emi
Weregodị ya na ị na-eguzo n'akụkụ ugwu na-adịghị mma na-agbalị iru ala ndagwurugwu ka ị na-enwe naanị mkpọda n'okpuru ụkwụ gị. Ọdịda gradient na-eme nke a kpọmkwem maka njehie ihe nlereanya. The gradient na-atụ aka na ntụziaka nke steepe kasị elu na ọnwụ, ya mere algọridim na-aga n'akụkụ nke ọzọ iji belata njehie. A na-achịkwa nha nke usoro ọ bụla site na ọnụego mmụta, hyperparameter dị oke mkpa: oke buru ibu na ihe nlereanya ahụ na-agbapụ na diverges, obere na ọzụzụ ọzụzụ. Na omume, ụdị anaghị adịkarị na-eji usoro data zuru ezu maka nzọụkwụ ọ bụla. Stochastic gradient descent (SGD) na mini-batch variants na-eme atụmatụ gradient site na obere ihe nlele, na-eme ọzụzụ ngwa ngwa ma na-enyere ihe nlereanya ahụ aka ịgbanarị ọnyà na-emighị emi na elu ọnwụ.
Nghọta nka nka
Mmelite ọ bụla na-agbaso iwu dị mfe: ịdị arọ ọhụrụ hà nhata ibu ochie na-ebelata ọnụego mmụta ugboro gradient. Obere-batch gradient mgbada na-agbakọ gradient na obere obere data karịa nhazi niile, na-ere ahịa ziri ezi maka ọsọ na mkpọtụ bara uru. Ndị na-emeziwanye ihe n'oge a dị ka Adam na-ewuli elu na nke a site n'ịgbanwe usoro mmụta dị irè n'otu n'otu ma na-agbakwunye ume, nke na-akwakọba gradients gara aga iji mee ka ọ dị mma ma mee ka ọganihu dịkwuo elu site na mpaghara dị larịị ma ọ bụ nke yiri ndagwurugwu nke ọdịda ọdịda.
Mmetụta atụmatụ
Mkpebi doro anya
Ọ na-enyere gị aka ikewapụta nkwupụta ọrụ aka doro anya na asụsụ ahịa.
Ọnụ ego na mmefu ego
Ị nwere ike ịjụ ajụjụ mmejuputa iwu ka mma tupu itinye ego ma ọ bụ oge.
Team na usoro ọrụ
Ndị otu nwere nghọta na-eme ka ngwaahịa, amụma na mkpebi mmụta ka mma.
Ọdịnihu nke mgbada gradient
A naghị ejikarị mgbada gradient eme ihe naanị ya; ndị na-eme mgbanwe dị ka Adam na AdamW na-achịkwa ọzụzụ buru ibu. Nchọcha na-aga n'ihu na usoro mmụta-ọnụego, atumatu ikpo ọkụ, na usoro nke abụọ na-eji ozi curvature maka njikọta ọsọ ọsọ. Ka ụdị na-eto eto, nkesa na mbelata gradient gafere puku kwuru puku GPU na-aghọ ihe dị mkpa, yana usoro iji kwado mmelite ndị a buru ibu bụ oke na-arụ ọrụ. Echiche bụ isi, soro gradient na-adịghị mma, ga-aga n'ihu, mana igwe ndị dị n'akụkụ nha ọkwa ọkwa na-aga n'ihu.
Mmejuputa n'ezie n'ụwa
Iweda njehie amụma ụdị asụsụ gafere ọtụtụ ijeri akara ọzụzụ site na iji nwelite obere ogbe
Idozi ọnụego mmụta ka ihe onyonyo wee gbakọta ngwa ngwa na-enweghị mfu na-agbawa
N'iji oge na-agba ọsọ ọsọ ọzụzụ nke netwọk njirimara okwu rapaara na ndagwurugwu ọnwụ dị ogologo ma dị warara
Itinye Adam n'ọrụ iji dozie ihe nlere anya na obere dataset ebe ọnụego mmụta kwa-parameter na-enyere aka kwụsie ike
Ihe ize ndụ & okporo ụzọ nche
Otu dị iche iche nwere ike iji otu okwu ahụ mee ihe n'ụzọ dị iche, yabụ kọwapụta oge n'oge.
Ihe nrịbama nwere ike ịdị ike ebe arụmọrụ ụwa na-adaghị adaba.
Ileghara ogo data na atụmatụ nyocha anya na-emepụtakarị nsonaazụ na-adịghị mma.
Map mmejuputa
Malite na nkọwa asụsụ dị larịị nke nsonaazụ ịchọrọ.
Họrọ otu metrik ịga nke ọma na otu ọnọdụ ọdịda tupu nnwale.
Gbaa obere onye na-anya ụgbọ elu nwere data nnọchite anya, ọ bụghị ihe ngosi ngosi na-egbu maramara.
Detuo ebe Gradient Descent na-enyere aka yana ebe ụzọ ndị dị mfe dị mma.
Nọgide na-eme nchọpụta
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 Gradient Descent 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
Ntuziaka na-esote
Stochastic gradient mgbada nwere oge
Ajụjụ a na-ajụkarị
What is Gradient Descent?
Mmụda gradient bụ usoro njikarịcha nke na-ebuga ihe atụ n'ezie na-agbada na mperi dị ala, otu obere nzọụkwụ n'otu oge. Ọ bụ ka mmụta na-eme ozugbo mgbasa ozi gbakọrọ gradients.
Kedu ụzọ mgbada gradient na-ebufe ibu?
Ihe gradient ahụ na-atụ aka na mmụba nke mfu kacha njọ, yabụ iji belata mfu, usoro algọridim na-aga n'ihu (na-adịghị mma).
Kedu ihe na-achịkwa ọnụego mmụta?
Ọnụego mmụta na-atụ ka ogologo nha na-aga n'ihu na mmelite ọ bụla; nnukwu oke ibu, obere obere na-eme ka ọzụzụ na-egbu mgbu dị ngwa ngwa.
Kedu ka stochastic ma ọ bụ obere ogbe gradient siri dị iche na iji setịpụ data zuru ezu?
Obere-batch na stochastic gradient mgbada dị ka gradient na-eji obere sample, nke na-eme ka ọzụzụ dị ngwa ma na-agbakwụnye mkpọtụ enyemaka.
Kedu nsogbu ọnụ ọgụgụ mmụta nke buru oke ibu nwere ike ịkpata?
Nnukwu nzọụkwụ nwere ike ịgafe opekempe wee gbadaa gburugburu ma ọ bụ fesaa, na-eme ka ọnwụ ahụ na-abawanye karịa karịa idozi.
Kedu ihe ọkụ ọkụ na-agbakwunye na mgbada gradient?
Oge oge na-ebu nkezi nke gradients gara aga, na-enyere onye na-ebuli elu aka ịgafe ngwa ngwa site na ndagwurugwu wee mebie oscillations.