WASerstein GAN
Wasserstein GAN (WGAN) ni igishushanyo mbonera cyamahugurwa ya GAN ikoresha intera ya Wasserstein aho gutakaza min-max yambere.
Incamake
It makes notoriously unstable GAN training far more reliable and gives a loss value that actually correlates with image quality.
Kwibira cyane
GANs yumwimerere itoza imiyoboro ibiri mugukurura intambara: generator ikora amashusho yimpimbano kandi ivangura igerageza kubibona. Ibi bikunze gusenyuka cyangwa guhagarara kuko igihombo cyivangura ntacyo kivuga kubijyanye niterambere. WGAN, yatangijwe na Arjovsky, Chintala, na Bottou mu 2017, isimbuza ivangura n '' umunegura 'utanga amanota yerekana uko ishusho igaragara ku gipimo gikomeza aho gushyira mu byiciro-vs-mpimbano. Intego y'amahugurwa ihinduka intera ya Wasserstein (isi-yimuka) hagati yukuri no gutanga amakuru. Iyi ntera itanga gradients yoroshye, isobanutse nubwo mugihe ibyo kugabura byombi guhuzagurika, bikagabanya cyane uburyo bwo gusenyuka no gutuma igihombo kigabanuka ikimenyetso cyiza.
Ubushishozi
Intera ya Wasserstein yapimye mu buryo bwimbitse 'akazi' kugirango ihindure ikirundo cy'umwanda (gukwirakwiza impimbano) mu kindi (nyacyo). Kubara bishingiye kubintu bibiri bya Kantorovich-Rubinstein, bisaba kunegura kuba 1-Lipschitz (gradients imipaka). WGAN yumwimerere yabishyize mubikorwa muburyo bwo gukuramo uburemere buke; Nyuma WGAN-GP yasimbuye gukata ibihano bihanishwa buhoro buhoro asunika buhoro buhoro ibipimo byo kunegura kuri 1, imyitozo ihamye.
Ingaruka z'Ingamba
Umuvuduko n'igipimo
AI igaragara irashobora gukora igenzura, gutahura, no gutondekanya imirimo kurwego.
Build choices
Amakipe arema arashobora prototype ibitekerezo byihuse hamwe nintoki nkeya.
Itsinda hamwe nakazi
Ibikorwa birashobora gukoresha amashusho nibimenyetso bya videwo byari bigoye gutunganya.
Kazoza ka Wasserstein GAN
Ubushishozi bwibanze bwa WGAN, ko guhitamo intera ikwirakwiza bigira ireme ryiza, biracyumvikana muburyo bwo kubyara ibintu. Mugihe moderi yo gukwirakwiza ubu yiganje mugushushanya kwishusho, ibitekerezo byiza-byo gutwara biva muri WGAN byongeye kugaragara muburyo bwo guhuza, uburyo bwa Schrodinger-ikiraro, hamwe no gutandukanya moderi yo gukwirakwiza mumashanyarazi yihuta. Tegereza intego-yuburyo bwa Wasserstein kugirango ukomeze kumenyesha uburyo bwimvange aho imyitozo ihamye hamwe nigihombo gifatika, cyane cyane mubumenyi nubumenyi buke.
Gushyira mu bikorwa Isi
Kubyara amasura nuburyo bwo gufotora aho vanilla GANs yaguye kugeza kubisubizo bike
Gukora amashusho yubuvuzi bwa syntetique, nka MRI cyangwa amateka ya histologiya, kugirango yongere gake yanditseho imibare
Kwerekana ibice-kugongana mubyabaye-imbaraga za fiziki zigereranya aho imyitozo ihamye ari ngombwa
Gukora nkibipimo fatizo mubushakashatsi bwa ML kuko igihombo cyacyo cyerekana icyitegererezo cyiza kuruta amahugurwa
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
Sobanura ibipimo byo kwemererwa kugiciro, kwibutsa, nibiciro byamakosa.
Gerageza hamwe namakuru ajyanye nuburyo nyabwo bwo gukora.
Ongeraho isubiramo ryabantu kubwizere buke cyangwa guhanura cyane.
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 Wasserstein GAN 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
ESRGAN na GAN Ikirenga-Icyemezo
Ibibazo bikunze kubazwa
What is Wasserstein GAN?
Wasserstein GAN (WGAN) ni igishushanyo mbonera cyamahugurwa ya GAN ikoresha intera ya Wasserstein aho gutakaza min-max yambere. Bituma imyitozo izwi cyane idahinduka GAN yizewe cyane kandi itanga agaciro k'igihombo gifitanye isano nubwiza bwibishusho.
Nibihe bipimo WGAN ikoresha kugirango igereranye kugabanwa kwukuri kandi kubyara?
WGAN isimbuza intego yumwimerere ya Jensen-Shannon nintera ya Wasserstein, itanga gradients yoroshye nubwo kugabana bitarenze.
Muri WGAN, umuyoboro wari ivangura wahinduwe iki, kandi kuki?
Abanegura batanga amashusho kurwego rukomeza aho gutondekanya vs vs mpimbano, aribyo bituma Wasserstein akora neza.
Kuki abanenga WGAN bagomba guhagarikwa kuba 1-Lipschitz?
Ibice bibiri byemerera WGAN kugereranya intera ya Wasserstein ifata gusa imirimo ya 1-Lipschitz, bityo gradients zo kunegura zigomba kuba ziteganijwe.
Nigute WGAN yumwimerere yashyize mubikorwa imbogamizi ya Lipschitz?
Urupapuro rwa mbere rwa WGAN rwakoresheje gukuramo uburemere buke; WGAN-GP nyuma yaje kuyisimbuza igihano cyoroheje.
Ni ikihe kibazo WGAN igabanya cyane ugereranije na vanilla GAN?
WGAN yoroshye ya gradients curb uburyo bwo gusenyuka kandi bitanga igihombo gifitanye isano nubwiza bwikitegererezo.