Ingirabuzimafatizo ndende yo kwibuka
Ingirabuzimafatizo ndende (LSTM) selile ni ubwoko bwihariye bwimikorere ya neural neteur yubatswe kugirango yibuke amakuru murwego rurerure.
Incamake
They solved the vanishing-gradient problem that crippled earlier RNNs, powering a decade of breakthroughs in language, speech, and translation.
Kwibira cyane
Yatangijwe na Sepp Hochreiter na Jurgen Schmidhuber mu 1997, selile LSTM ikomeza 'selile selile' ikora nkumukandara wa convoyeur yibuka unyuze murukurikirane. Amarembo atatu yize aragenzura: irembo ryibagirwa rihitamo icyo gusiba, irembo ryinjira rihitamo amakuru mashya yo kubika, naho irembo risohoka rihitamo icyo ryerekana nkibisohoka muri selire. Buri rembo rikoresha sigmoid (isohoka 0 kugeza 1) kugirango ikore nka switch yoroheje. Kuberako selile yimikorere ivugururwa cyane cyane mubyongeweho aho kugwiza inshuro nyinshi, gradients zirashobora gusubira inyuma mugihe kinini cyintambwe zitagabanutse kuri zeru, kureka LSTM ikiga kwishingikiriza kuntambwe amagana. Mbere y'abahindura, LSTMs yashimangiye Google Guhindura, kumenyekanisha imvugo, no kubyara inyandiko.
Ubushishozi
Gukuraho-buhoro buhoro gukosora biva muri selire ya reta yegereye umurongo: c_t = f_t * c_ {t-1} + i_t * g_t. Kwibagirwa irembo f_t (sigmoid) irashobora kuguma hafi ya 1, igakora 'guhora ikosa karuseli' kuburyo ibimenyetso byamakosa bikomeza gusubira inyuma-mugihe-kinini mugihe kirekire. Amarembo ubwayo ni mato mato (sigmoid yo gutambuka, tanh kumico yabakandida), bose bahuguwe hamwe na gradient descente. Irembo rireka umuyoboro wiga icyo ugomba kubika nicyo guta.
Ingaruka z'Ingamba
Ibyemezo bisobanutse
Iragufasha gutandukanya ibyifuzo bya tekiniki bisobanutse nururimi rwo kwamamaza.
Igiciro na bije
Urashobora kubaza ibibazo byiza byo gushyira mubikorwa mbere yo gukoresha amafaranga cyangwa igihe.
Itsinda hamwe nakazi
Amakipe asangiye ibitekerezo akora ibicuruzwa byiza, politiki, nibyemezo byo kwiga.
Ejo hazaza h'igihe kirekire cyo kwibuka
Abahinduzi barenze cyane LSTMs kubikorwa binini byururimi kuko bigereranywa bikurikiranye kandi bigafata intera ndende binyuze mubitekerezo, mugihe LSTM itunganya ibimenyetso intambwe imwe imwe. Biracyaza, LSTMs ikomeza kuba iyagaciro mugutemba, gutinda-gutinda, hamwe nibikoresho-bigenwa igenamiterere, no kumwanya muto-wamakuru. Ibikorwa bya vuba nka xLSTM (2024) bisubiramo kandi bigezweho byubaka hamwe na gati nshya hamwe nibuka kugirango duhatane kurwego, byerekana igitekerezo kitarangiye.
Gushyira mu bikorwa Isi
Guhindura imashini isobanura hakiri kare Google Sobanura sisitemu yimitsi mbere yuko Transformers ifata.
Kumenyekanisha imvugo-mwandiko mu bafasha amajwi hamwe na software yerekana.
Guteganya indangagaciro zizaza mugihe gikurikiranye nkibisabwa ingufu, gusoma sensor, cyangwa ibiciro byimigabane.
Gukora inyandiko cyangwa umuziki ikimenyetso kimwe icyarimwe hamwe na autocompleting ikurikirana.
Ingaruka & Kurinda
Amakipe atandukanye arashobora gukoresha ijambo rimwe muburyo butandukanye, sobanura intera hakiri kare.
Ibipimo birashobora kugaragara bikomeye mugihe imikorere-yisi-itaringaniye.
Kwirengagiza ubuziranenge bwamakuru na gahunda yo gusuzuma akenshi bitanga ibisubizo byoroshye.
Igishushanyo mbonera
Tangira nururimi rusobanutse rwibisubizo ukeneye.
Toranya intsinzi imwe hamwe nuburyo bumwe bwo gutsindwa mbere yo kwipimisha.
Koresha umuderevu muto hamwe namakuru ahagarariye, ntabwo ari demo yashizweho.
Inyandiko aho Utugingo ngufi twa Memory selile zifasha kandi nuburyo bworoshye bworoshye.
Komeza Ubushakashatsi
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
Ubuyobozi bukurikira
Ubuyobozi bwa GPU bwo kwibuka no gucamo ibice
Ibibazo bikunze kubazwa
What is Long Short-Term Memory Cells?
Ingirabuzimafatizo ndende (LSTM) selile ni ubwoko bwihariye bwimikorere ya neural neteur yubatswe kugirango yibuke amakuru murwego rurerure. Bakemuye ikibazo cyo kuzimira-buhoro buhoro cyamugaye mbere ya RNN, bitanga imyaka icumi yiterambere mu mvugo, imvugo, no guhindura.
Ni ayahe marembo atatu agenzura amakuru atemba muri selile isanzwe ya LSTM?
LSTM ikoresha irembo ryibagirwa (ibyo gusiba), irembo ryinjira (ibyo kubika), hamwe n irembo risohoka (ibyo gushyira ahagaragara), buriwese sigmoid yize.
Ni ikihe kibazo gikomeye na RNNs mbere LSTMs yakemuye?
RNNs isanzwe ibura gradients ibuza kwiga intera ndende; LSTM yongeyeho selile leta ireka gradients ikomeza.
Ninde watangije LSTM, kandi nuwuhe mwaka?
Sepp Hochreiter na Jurgen Schmidhuber basohoye LSTM mu 1997.
Kuki leta ya LSTM selile ifasha gradients kubaho mugihe cyintambwe nyinshi?
Ivugurura ryongeweho ('ikosa rihoraho karuseli') ryirinda kugwiza inshuro zigabanya gradients, bityo ibimenyetso byamakosa bigaruka hejuru yigihe kirekire.
Ni ubuhe buryo bwo gukora amarembo ya LSTM akoresha muburyo bworoshye kuri / kuzimya?
Amarembo akoresha sigmoid, ifite 0-kuri-1 ibisohoka umunzani uko amakuru anyura, akora nka switch yoroshye.