Unugyada xusuusta ee muddada-gaaban
Unugyada xusuusta muddada-gaaban (LSTM) waa nooc gaar ah oo ah unug shabakad neerfaha ah oo soo noqnoqda oo loo dhisay in lagu xasuusto macluumaadka taxanaha dheer.
Dulmar
They solved the vanishing-gradient problem that crippled earlier RNNs, powering a decade of breakthroughs in language, speech, and translation.
quusid qoto dheer
Waxaa soo bandhigay Sepp Hochreiter iyo Jurgen Schmidhuber sanadkii 1997, unugga LSTM wuxuu ilaaliyaa 'dawlad unug' kaasoo u dhaqma sida suunka wareejinta xusuusta ee ku socda taxanaha. Saddex albaab oo la bartay ayaa gacanta ku haya: Irridka illowda ayaa go'aamiya waxa la tirtirayo, albaabka wax lagu gelinayo ayaa go'aamiya macluumaadka cusub ee la kaydinayo, albaabka wax soo saarka ayaa go'aamiya waxa lagu soo bandhigayo wax soo saarka unugga. Irid kastaa waxay isticmaashaa sigmoid (wax soo saarka 0 ilaa 1) si uu u noqdo bedel jilicsan. Sababtoo ah gobolka unugga waxaa lagu cusboonaysiiyaa inta badan isugeynta halkii lagu celcelin lahaa isku dhufashada, jaranjarooyinka waxay u qulquli karaan dib-u-dhacyo waqti badan iyada oo aan la sii yaraanin eber, u oggolaanaya LSTM-yada inay bartaan ku-tiirsanaanta boqolaal tillaabo. Transformers ka hor, LSTM-yadu waxay taageereen Google Turjumi, aqoonsiga hadalka, iyo qoraalka qoraalka.
Aragtida Farsamada
Hagaajinta luminta-gradient waxay ka timaadaa cusboonaysiinta toosan ee gobolka unugga: c_t = f_t * c_{t-1} + i_t * g_t. Albaabka illowda f_t (sigmoid) wuxuu joogi karaa meel u dhow 1, abuurista 'carousel qalad joogto ah' si calaamadaha khaladku ay uga badbaadaan faafinta-ilaa-waqtiga dheer ee dheer. Albaabada laftoodu waa lakabyo yar yar oo neerfaha ah (sigmoid for gating, tanh ee qiyamka musharraxiinta), dhamaantood waxaa si wadajir ah loogu tababaray faracyo hoose. Albaabkani wuxuu u ogolaanayaa shabakadu inay bartaan waxa la hayo iyo waxa la tuurayo.
Saamaynta Istiraatijiyadeed
Go'aamo cad
Waxay kaa caawinaysaa inaad kala saartid sheegashooyinka farsamada cad iyo luqadda suuq-geynta.
Qiimaha iyo miisaaniyada
Waxaad waydiin kartaa su'aalo fulineed oo wanaagsan ka hor inta aadan lacag ama waqti bixin.
Kooxda iyo socodka shaqada
Kooxaha fahamka la wadaago waxay sameeyaan wax soo saar, siyaasad, iyo go'aano waxbarasho oo wanaagsan.
Mustaqbalka Unugyada Xasuusta ee Muddada Dheer
Transformers ayaa inta badan la wareegay LSTM-yada hawlaha luqadeed ee baaxadda leh sababtoo ah waxay barbar socdaan isku xigxiga waxayna qabtaan macnaha fog ee dareenka, halka habka LSTM uu calaamad u yahay hal tallaabo markiiba. Weli, LSTM-yadu waxay ahaanayaan kuwo qiimo u leh baahinta, daahitaanka hooseeya, iyo goobaha xaddidan ee kheyraadka, iyo xogta taxanaha-waqtiga yar. Shaqadii u dambaysay sida xLSTM (2024) waxay dib u eegtaa oo ku casriyaysaa qaab-dhismeedka qaab-dhismeed cusub oo leh gating cusub iyo xusuus si ay ugu tartamaan cabbirka, iyagoo muujinaya fikradda aan dhammaan.
Dhaqangelinta Adduunka-dhabta ah
Turjumidda mishiinka hore Google Turjun habka neerfaha ka hor inta aanay Transformers la wareegin.
Aqoonsiga hadalka-ka-qoraalka ee caawiyayaasha codka iyo software-ka hadalka.
Saadaalinta qiyamka mustaqbalka ee taxanaha wakhtiga sida baahida tamarta, akhrinta dareemayaasha, ama qiimaha saamiyada.
Abuuritaanka qoraal ama muusig hal calaamad markiiba iyo dhammaystirka taxanaha.
Khatarta & Dariiqyada Ilaalada
Kooxo kala duwan ayaa laga yaabaa inay isla erey u isticmaalaan si kala duwan, marka hore u qeex baaxadda.
Tilmaamaha ayaa u ekaan kara kuwo xooggan halka waxqabadka dhabta ah ee dunidu aanu sinnayn.
In la iska indho tiro tayada xogta iyo qorshayaasha qiimayntu waxay inta badan abuurtaa natiijooyin jilicsan.
Qorshe Hawleedka Dhaqangelinta
Ka bilow qeexidda luqadda cad ee natiijada aad u baahan tahay.
Dooro hal cabbir guusha iyo hal xaalad guuldarro ka hor tijaabada.
Ku orod duuliye yar oo wata xogta matale, ee ma aha bandhig muuqaal ah.
Dukumeenti halka ay ku caawinayaan Unugyada xusuusta muddada-gaaban iyo meelaha hababka fudud ay ka fiican yihiin.
Sii wad Sahaminta
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
Hagaha xiga
Maareynta Xusuusta GPU-da iyo Kala-jabinta
Su'aalaha soo noqnoqda
What is Long Short-Term Memory Cells?
Unugyada xusuusta muddada-gaaban (LSTM) waa nooc gaar ah oo ah unug shabakad neerfaha ah oo soo noqnoqda oo loo dhisay in lagu xasuusto macluumaadka taxanaha dheer. Waxay xalliyeen mushkiladdii sii liidata ee curyaamisay RNN-yadii hore, iyaga oo awood u yeeshay toban sano oo horumarro xagga luqadda, hadalka, iyo tarjumaada ah.
Saddexda albaab keebaa kantaroola xogta ku socota unugga LSTM ee caadiga ah?
LSTM waxa ay isticmaashaa albaabka illowda (waxa la tirtiro), albaabka wax lagu shubo (waxa la kaydinayo), iyo albaabka wax soo saarka (waxa la soo bandhigayo), mid walba sigmoid la bartay.
Dhibaato noocee ah ayaa RNN-yadii hore ee LSTM-yadu xaliyeen?
Heerka caadiga ah ee RNN-yada waxay la ildaran yihiin jaangooyooyin luminaya oo ka hortagaya barashada ku-tiirsanaanta muddada-dheer; Gobolka unugga wax-ku-darka ee LSTM wuxuu u ogolaanayaa gradients inay sii jiraan.
Yaa soo bandhigay LSTM, iyo sanadkee?
Sepp Hochreiter iyo Jurgen Schmidhuber ayaa daabacay LSTM sanadkii 1997.
Waa maxay sababta gobolka unugga LSTM uu u caawiyo jaranjarooyinka inay ku badbaadaan tallaabooyin waqti badan ah?
Cusboonaysiinta dheeriga ah ('karoosalka qaladka joogtada ah') ayaa ka fogaanaya isku dhufashada soo noqnoqda ee hoos u dhigaya jaangooyooyinka, markaa calaamadaha khaladku waxay dib ugu soo noqdaan muddo dheer.
Waa maxay shaqada firfircoonida ee irdaha LSTM sida caadiga ah u isticmaalaan inay u dhaqmaan sida daar/daminta jilicsan?
Gates waxay isticmaalaan sigmoid, kaas oo wax soo saarkiisa 0-ilaa-1 uu cabbirayo inta macluumaadku dhaafo, una dhaqmaya sida furaha jilicsan.