Kwayoyin Ƙwaƙwalwar Ƙwaƙwalwar Tsawon Lokaci
Ƙwaƙwalwar Ƙwaƙwalwar Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwaƙwalwa na Ƙwaƙwalwa na Ƙwaƙwal ) ne na musamman wanda aka gina don tunawa da bayanai a cikin jerin dogon lokaci.
Dubawa
They solved the vanishing-gradient problem that crippled earlier RNNs, powering a decade of breakthroughs in language, speech, and translation.
Zurfafa nutsewa
Sepp Hochreiter da Jurgen Schmidhuber suka gabatar a cikin 1997, tantanin halitta na LSTM yana kula da 'yanayin tantanin halitta' wanda ke aiki kamar bel mai ɗaukar ƙwaƙwalwar ajiya yana gudana ta cikin jerin. Ƙofofi uku masu ilmantarwa suna sarrafa ta: Ƙofar mantuwa tana yanke shawarar abin da za a goge, ƙofar shigar ta yanke shawarar abin da sabon bayanin da za a adana, kuma ƙofar fitarwa ta yanke shawarar abin da za a fallasa a matsayin fitarwar tantanin halitta. Kowace kofa tana amfani da sigmoid (fitarwa 0 zuwa 1) don yin aiki azaman sauya mai laushi. Saboda ana sabunta yanayin tantanin halitta galibi ta ƙari maimakon maimaita maimaitawa, gradients na iya komawa baya cikin matakai na lokaci da yawa ba tare da raguwa zuwa sifili ba, barin LSTMs su koyi dogaro da ɗaruruwan matakai baya. Kafin Transformers, LSTMs suna ƙarƙashin Google Fassara, fahimtar magana, da tsara rubutu.
Fahimtar Fasaha
Gyaran ɓarna-gradient yana fitowa daga sabuntawar kusa-daidaita na jihar tantanin halitta: c_t = f_t * c_{t-1} + i_t * g_t. Ƙofar manta f_t (sigmoid) na iya zama kusa da 1, ƙirƙirar 'kuskuren carousel na yau da kullun' don haka siginonin kuskure sun tsira daga baya-ta-lokaci a cikin dogon lokaci. Ƙofofi su kansu ƙananan yadudduka ne na jijiyoyi (sigmoid don gating, tanh don ƙimar ɗan takara), duk an horar da su tare ta hanyar zuriyar gradient. Wannan gating yana ba da damar cibiyar sadarwa ta koyi abin da za ta kiyaye da abin da za a jefar.
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 Kwayoyin Ƙwaƙwalwar Ƙwaƙwalwar Tsawon Lokaci
Masu canji sun mamaye LSTMs don manyan ayyuka na harshe saboda suna daidaitawa a cikin jeri kuma suna ɗaukar mahallin dogon zango ta hanyar hankali, yayin da tsarin LSTM ke nuna alamar mataki ɗaya a lokaci guda. Har yanzu, LSTMs sun kasance masu kima don yawo, ƙarancin jinkiri, da ƙayyadaddun saitunan albarkatu, kuma akan matsakaicin bayanan jerin lokaci. Aiki na baya-bayan nan kamar xLSTM (2024) ya sake duba tare da sabunta gine-gine tare da sabon gating da ƙwaƙwalwar ajiya don gasa a sikelin, yana nuna ra'ayin bai ƙare ba.
Aiwatar da Gaskiyar Duniya
Ƙaddamar da fassarar inji a farkon Google Fassara tsarin jijiya kafin Transformers su karɓi ragamar aiki.
Ƙirar magana-zuwa-rubutu a cikin mataimakan murya da software na ƙamus.
Hasashen ƙima na gaba a cikin jerin lokaci kamar buƙatar makamashi, karatun firikwensin, ko farashin hannun jari.
Ƙirƙirar rubutu ko kiɗa alama ɗaya a lokaci ɗaya da kammala jerin kai tsaye.
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 bayanai inda Kwayoyin ƙwaƙwalwar ajiya na ɗan gajeren lokaci 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 Long Short-Term Memory Cells 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
Gudanar da Ƙwaƙwalwar Ƙwaƙwalwar GPU da Rarraba
Tambayoyin da ake yawan yi
What is Long Short-Term Memory Cells?
Ƙwaƙwalwar Ƙwaƙwalwar Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwararren Ƙwaƙwalwa na Ƙwaƙwalwa na Ƙwaƙwal ) ne na musamman wanda aka gina don tunawa da bayanai a cikin jerin dogon lokaci. Sun warware matsalar ɓata-girma wacce ta gurgunta RNNs a baya, tare da ƙarfafa shekaru goma na ci gaba a cikin harshe, magana, da fassara.
Wadanne ƙofofi uku ne ke sarrafa bayanan da ke gudana a daidaitaccen tantanin halitta na LSTM?
LSTM yana amfani da ƙofar manta (abin da za a goge), ƙofar shigarwa (abin da za a adana), da ƙofar fitarwa (abin da za a fallasa), kowane sigmoid da aka koya.
Wace babbar matsala ta RNNs na baya LSTMs suka magance?
Madaidaitan RNNs suna fama da ɓarna gradients waɗanda ke hana koyan dogaro na dogon zango; Yanayin ƙari na LSTM yana ba da damar gradients su dage.
Wanene ya gabatar da LSTM, kuma a cikin wace shekara?
Sepp Hochreiter da Jurgen Schmidhuber sun buga LSTM a cikin 1997.
Me yasa jihar tantanin halitta ta LSTM ke taimakawa gradients su tsira a kan matakan lokaci da yawa?
Sabuntawar ƙari ('kuskuren carousel na yau da kullun') yana nisantar maimaita maimaitawa wanda ke raguwa gradients, don haka siginonin kuskure suna komawa cikin dogon lokaci.
Wane aikin kunnawa ƙofofin LSTM yawanci ke amfani da su don aiki azaman masu kunnawa/kashe masu taushi?
Gates suna amfani da sigmoid, wanda 0-to-1 fitarwa na sikelin nawa bayanin ya wuce, yana aiki kamar sauya mai laushi.