Amabwiriza ya AI

VQ-VAE na Bitinze

VQ-VAE ikanda amashusho, amajwi, cyangwa videwo mumurongo muto wa code ya discret yakuwe mubitabo byize, aho kuba imibare ikomeza.

2 min somaIbiherutse kuvugururwa

Incamake

Uku gucikamo ibice kureka ibintu bikomeye bikurikirana nka Transformers ifata itangazamakuru nk 'ibimenyetso', nkamagambo.

Kwibira cyane

VQ-VAE (Vector Quantized Variational Autoencoder), yatangijwe na van den Oord na bagenzi be muri DeepMind muri 2017, ni autoencoder ifite umwanya wihishe utagaragara. Kodegisi ihindura ishusho murusobe rwibice bikomeza; buri vector noneho ifatirwa kumurongo wacyo hafi muri codebook yize yo gushiramo (kugereranya kwa vector). Decoder yubaka ishusho kuva kode zingana. Kuberako ibyihishe ubu ari amagambo atagira ingano yerekana indangagaciro, icyitegererezo cyihariye gishobora kwiga kugabana kwabo no kubyara ibintu bishya. Ibi byiciro bibiri byokoresha imbaraga DALL-E 1, Jukebox kumuziki, na VQGAN, byongera igihombo cyo kwiyumvisha no guhangana kugirango kwiyubaka bikarishye. VQ-VAE-2 yashyizeho imyanzuro myinshi kugirango itange amashusho yubudahemuka.

Ubushishozi

Intambwe yo kubara (argmin hafi-umuturanyi ushakisha) ntaho itandukaniye, bityo VQ-VAE ikoresha igereranya-igereranya: gradients yandukuwe biturutse kuri decoder yinjiza igaruka kuri kodegisi nkaho kubara byari indangamuntu. Amahugurwa akomatanya igihombo cyo kwiyubaka, igihombo cya codebook gikurura ibyerekezo bisohoka kuri kodegisi, hamwe nigihombo cyo kwiyemeza gukomeza kodegisi yiyemeje kode yahisemo. Kunanirwa bisanzwe ni codebook gusenyuka, aho kode nkeya zikoreshwa.

Ingaruka z'Ingamba

Umuvuduko n'igipimo

AI igaragara irashobora gukora igenzura, gutahura, no gutondekanya imirimo kurwego.

Kubaka amahitamo

Amakipe arema arashobora prototype ibitekerezo byihuse hamwe nintoki nkeya.

Itsinda hamwe nakazi

Ibikorwa birashobora gukoresha amashusho nibimenyetso bya videwo byari bigoye gutunganya.

Ejo hazaza ha VQ-VAE na Bitinze Bitinze

Ibanga ryihariye ni ingenzi mu gusunika kuri moderi imwe ihuriweho yerekana amashusho, amajwi, na videwo mu magambo amwe n’inyandiko. Gutezimbere nkibisigisigi bisigaye kandi bitagira ingano, kodegisi nini, hamwe no kuringaniza imikoreshereze myiza bigabanya gusenyuka no kuzamura ubudahemuka. Nkuko icyitegererezo kigamije gusobanukirwa no kubyara muburyo butandukanye, ibimenyetso bikomeye byubakiye kubitekerezo bya VQ-VAE bizakomeza kuba urufatiro, bigenda bihatana kandi bigahuza nuburyo bukwirakwizwa bwihishwa.

Gushyira mu bikorwa Isi

DALL-E 1 yakoresheje icyerekezo cya VQ-VAE cyihariye kugirango Transformer ibashe gukora amashusho nkurutonde rwibipimo bya codebook.

VQGAN yahujije VQ-VAE hamwe nigihombo cyo guhangana no gushishoza kugirango itange ibimenyetso bisobanutse neza, byerekana amashusho yerekana ibihangano.

__AIU_PROTECTED_10 __ 'Jukebox yakoresheje VQ-VAE kumajwi mbisi, ikomatanya umuziki muma code yihariye yo kwerekana imiterere.

VQ-AA

Ingaruka & Kurinda

Uburenganzira bwishusho hamwe no kwemererwa birashobora guhinduka ibyago byemewe n'amategeko niba ibimenyetso bidasobanutse.

Imikorere yicyitegererezo irashobora gutandukana kumurika, demografiya, nibidukikije.

Ibyiza byibinyoma birashobora kutamenyekana keretse niba ibyiringiro byateganijwe bikurikiranwa.

Igishushanyo mbonera

1

Sobanura ibipimo byo kwemererwa kugiciro, kwibutsa, nibiciro byamakosa.

2

Gerageza hamwe namakuru ajyanye nuburyo nyabwo bwo gukora.

3

Ongeraho isubiramo ryabantu kubwizere buke cyangwa guhanura cyane.

4

Kurikirana icyitegererezo cya drift hanyuma uhindurwe nyuma ya kamera cyangwa dataset ihinduka.

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 VQ-VAE and Discrete Latents quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Tangira ikibazo

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Ubuyobozi bukurikira

Kuvanga Bitinze hamwe na Interpolation

Ibibazo bikunze kubazwa

Niki VQ-VAE na Discret Latents?

VQ-VAE ikanda amashusho, amajwi, cyangwa videwo mumurongo muto wa code ya discret yakuwe mubitabo byize, aho kuba imibare ikomeza. Uku gucikamo ibice kureka ibintu bikurikirana bikurikirana nka Transformers ifata itangazamakuru nk '' ibimenyetso ', nkamagambo.

Niki gitandukanya umwanya wihishe wa VQ-VAE itandukanye na VAE isanzwe?

VQ-VAE isimbuza ibyihuta bikomeza hamwe na code yihariye yakuwe muri codebook yize ikoresheje kubara kwa vector.

Nigute VQ-VAE itambutsa gradients binyuze mu ntambwe yo gutandukanya itandukanye?

Ikigereranyo cyanyuze mu kugereranya kopi ya decoder-yinjiza gradient mu buryo butaziguye kuri kodegisi isohoka, ifata umubare nkirangamuntu yinyuma yinyuma.

'Gusenyuka kwa codebook' ni iki?

Gusenyuka kwa Codebook bibaho mugihe icyitegererezo gishingiye kuri code nkeya, guta igice kinini cya codebook kandi bikangiza ubudasa.

Ni ukubera iki ibanga ryihariye rifite akamaro ko kwerekana imiterere hamwe na Transformers?

Indice zidasanzwe zigize amagambo atagira ingano, bityo moderi zikurikirana nka Transformers zirashobora kwiga no kwigana ikwirakwizwa ryazo nkamagambo.

Niki VQGAN yongeyeho hejuru yuburyo bwibanze bwa VQ-VAE?

VQGAN yongerera VQ-VAE hamwe nivangura nigihombo cyo gushishoza, itanga ibisobanuro, ibisobanuro birambuye byubatswe.