VQ-VAE iyo daahsoonaha Discrete
VQ-VAE waxay ku cadaadisaa sawirada, maqalka, ama fiidyowga xayndaab yar oo koodh gaar ah oo laga soo qaatay buug-codeed la bartay, halkii ay ka ahaan lahaayeen tirooyin joogto ah.
Dulmar
This discrete bottleneck lets powerful sequence models like Transformers treat media as 'tokens', much like words.
quusid qoto dheer
VQ-VAE (Vector Quantized Variational Autoencoder), oo uu soo bandhigay van den Oord iyo asxaabtiisa DeepMind sanadka 2017, waa cod-bixiye kaas oo meeshiisu qarsoon tahay. Codeeyuhu wuxuu u rogaa sawirka inuu u beddelo shabaqyo joogto ah; vector kasta ayaa markaa la jeexjeexay gelitaanka ugu dhow ee buug- codeed ee la bartay (tiro-ururinta vector). Decoder-ku wuxuu dib-u-dhisayaa sawirka koodka la tiriyay. Sababtoo ah qarsoodiyadu hadda waa eray bixin kooban oo tilmaame ah, nooc gaar ah ayaa baran kara qaybintooda wuxuuna soo saari karaa waxyaabo cusub. Cunto karinta laba-marxale ah waxay awood u leedahay DALL-E 1, Jukebox ee muusiga, iyo VQGAN, kaas oo ku daraya khasaare maskaxeed iyo iska caabin dib-u-dhiska fiiqan. VQ-VAE-2 waxay soo ururisay qaraaro badan si ay u soo saarto sawiro daacadnimo sare leh.
Aragtida Farsamada
Tallaabada qiyaasidda (argmin-deriska ugu dhow) ma kala duwana, sidaas darteed VQ-VAE waxay isticmaashaa qiyaas toos ah: gradients si toos ah ayaa looga koobiyeeyay gelinta qalabka dib loogu celiyo soo saarista codeeyaha sidii haddii qiyaasiddu tahay aqoonsiga. Tababarku waxa uu isku daraa khasaare dib-u-dhis ah, luminta buug-yaraha oo jiidaya wax-soo-saarka codeer-ka, iyo lumitaanka ballan-qaadka ilaalinta cod-bixiyaha oo ay ka go'an tahay koodka la doortay. Fashilka caadiga ah waa burburka buug-gacmeedka, halkaasoo dhowr kood la isticmaalo.
Saamaynta Istiraatijiyadeed
Xawaaraha iyo miisaanka
Visual AI wuxuu si otomaatig ah u samayn karaa baadhista, ogaanshaha, iyo sumadaynta hawlaha miisaanka.
Xulashada dhismayaasha
Kooxaha hal-abuurka leh waxay hindise karaan fikradaha si dhakhso leh iyagoo leh dib-u-eegis buugeed yar.
Kooxda iyo socodka shaqada
Hawlgalladu waxay isticmaali karaan calaamadaha muuqaalka iyo muuqaalka kuwaas oo markii hore adkeyd in la farsameeyo.
Mustaqbalka VQ-VAE iyo daahsoonaanshaha Discrete
Daahsoonayaasha kala duwan ayaa udub dhexaad u ah riixitaanka qaabab kala duwan oo midaysan kuwaaso calaamadiya sawirada, maqalka, iyo muuqaalka ereyada isku midka ah ee qoraalka. Hagaajinta sida tirooyinka haraaga ah iyo kuwa dhamaadka ah, buugaag codeed oo waaweyn, iyo dheellitirka isticmaalka wanaagsan ayaa yareynaya burburka iyo xoojinta daacadnimada. Sida moodooyinka ujeedadoodu tahay inay fahmaan oo ay dhaliyaan habab kala duwan, calaamadeeyayaal adag oo lagu dhisay fikradaha VQ-VAE waxay ahaan doonaan walxo aasaasi ah, oo sii tartameysa oo isku daraya hababka faafinta qarsoon ee joogtada ah.
Dhaqangelinta Adduunka-dhabta ah
DALL-E 1 waxay isticmaashay calaamade VQ-VAE discrete si uu Transformer-ku u soo saaro sawiro sida taxanaha tusmooyinka koodka.
VQGAN waxay ku dartay VQ-VAE oo leh khasaare iska soo horjeeda iyo garashada si ay u soo saarto calaamado muuqaal ah oo qallafsan, heer sare ah oo loogu talagalay jiilka farshaxanka.
OpenAI's Jukebox waxay codsatay VQ-VAE maqal cayriin ah, oo muusik ku cadaadinaysa koodh kala duwan oo qaabaynta abuurista ah.
VQ-VAE-2 is dulfuulan siyaalo kala duwan oo kala duwan si ay u abuuraan muuqaalo kala duwan, aaminaad sare oo la tartameysa GANs waageedii.
Khatarta & Dariiqyada Ilaalada
Xuquuqda sawirka iyo ogolaanshaha waxay noqon kartaa khataro sharci ah haddii caddayntu aanay caddayn.
Waxqabadka moodeelku wuu ku kala duwanaan karaa iftiinka, tirakoobka, iyo deegaanka.
Wanaagga beenta ah waxa laga yaabaa inaan la dareemin ilaa xadka kalsoonida aan la kormeerin.
Qorshe Hawleedka Dhaqangelinta
Qeex shuruudaha aqbalida ee saxnaanta, dib u celinta, iyo kharashyada khaladka.
Ku tijaabi xogta ku habboon xaaladaha wax soo saarka dhabta ah.
Ku dar dib u eegis bini'aadamka si aad u hesho kalsoonida hoose ama saameeynta sare.
Lasoco moodeel dhaqaaqa oo dib u cusboonaysii kamarada ama xogta kaydinta ka dib.
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 VQ-VAE and Discrete Latents 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
Isku-dhafka Daahsoon iyo Dhex-galka Sawirka
Su'aalaha soo noqnoqda
What is VQ-VAE and Discrete Latents?
VQ-VAE waxay ku cadaadisaa sawirada, maqalka, ama fiidyowga xayndaab yar oo koodh gaar ah oo laga soo qaatay buug-codeed la bartay, halkii ay ka ahaan lahaayeen tirooyin joogto ah. Qoortaan gaarka ah waxay u oggolaanaysaa moodooyinka isku xigxiga ee awoodda leh sida Transformers inay ula dhaqmaan warbaahinta sida 'calaamo', oo la mid ah erayada.
Maxaa ka dhigaya booska qarsoon ee VQ-VAE ka duwan kan caadiga ah ee VAE?
VQ-VAE waxay ku beddeshaa daahsoonayaasha joogtada ah oo ay ku beddesho koodh aan caadi ahayn oo laga soo qaatay buug-koodeed la bartay iyada oo loo marayo cabbirka vector.
Sidee ayay VQ-VAE uga gudubtaa gradients-ka tillaabada xisaabinta ee aan kala duwanayn?
Qiyaasta tooska ah ayaa si toos ah u koobiyaysa jaangooyooyinka wax-soo-gelinta cod-bixiyaha si toos ah wax-soo-saarka codeer-ka, isaga oo ula dhaqmaya qiyaasista aqoonsiga kaska dambe.
Waa maxay 'codebook burbur'?
Burburka Buug-gacmeedku waxa uu dhacaa marka moodeelku uu ku tiirsanaado dhawr kood oo keliya, isaga oo luminaya inta badan buugga koodhka oo dhaawacaya kala duwanaanshaha.
Waa maxay sababta daahsoonayaasha kala duwan ay faa'iido u leeyihiin qaabaynta abuurista ee Transformers?
Indices Discrete waxay sameeyaan eray bixin kooban, markaa moodooyinka isku xigxiga sida Transformers waxay baran karaan oo muunad u qaybin karaan sida ereyada oo kale.
Maxay VQGAN ku soo kordhisay cuntada aasaasiga ah ee VQ-VAE?
VQGAN waxay ku kordhisaa VQ-VAE takoore iyo luminta garashada, taasoo keenta calaamado aad u faahfaahsan oo dib loo dhisay.