Samfuran Ƙarfafa Maki
Samfurin ƙirƙira tushen maki suna ƙirƙirar bayanai ta hanyar koyon gradient na rarraba bayanai - alkiblar da ke sa kowane samfurin hayaniya ya yi kama da ainihin bayanai.
Dubawa
This score-function view unifies diffusion models with stochastic differential equations and underpins many modern image generators.
Zurfafa nutsewa
Madadin yuwuwar yin ƙira kai tsaye, ƙirar tushen ƙima suna koyon makin: girman girman yuwuwar log dangane da shigarwar. Sanin hanyar da za a ƙulla samfurin don ƙara yiwuwarsa ya isa ya samar da sababbin bayanai. Aikin Yang Song da Stefano Ermon na 2019 sun horar da hanyar sadarwa don kimanta wannan maki a cikin matakan amo da yawa ta amfani da rashin daidaituwar maki, sannan suka samar da samfurori tare da kuzarin Langevin - akai-akai tare da ci gaba da ƙara ɗan ƙara. Takardar maki-SDE ta 2021 ta nuna cewa yaduwa da samfuran tushen maki fuskoki biyu ne na tsari iri ɗaya da aka kwatanta ta hanyar ma'auni mai ban sha'awa. Mahimmanci, kowane SDE yana da daidaitaccen ƙayyadaddun ƙayyadaddun 'yiwuwar kwarara' ODE wanda ke raba rataye iri ɗaya, yana ba da damar ainihin dama da samfur mai sauri.
Fahimtar Fasaha
Ƙididdiga ƙididdiga na tsaftataccen bayanai kai tsaye yana da wuyar gaske inda bayanai ba su da yawa, don haka ana horar da ƙirar akan bayanan da hayaniyar Gaussian ke damun su a ma'auni masu yawa. Rashin daidaituwar maki yana ba da maƙasudi: makin rarraba amo ya yi daidai da alkiblar amo da aka raba ta da bambance-bambancen amo, don haka tsinkayar amo da tsinkayar maki ainihin abu ɗaya ne. Samfurin yana warware SDE na baya-lokaci (ko daidaitaccen yuwuwar- kwarara ODE) yana farawa daga tsantsar Gaussian amo.
Dabarun Tasiri
Gudu da sikelin
Kayayyakin AI na iya sarrafa aiki da bincike, ganowa, da ayyuka masu alama a sikelin.
Gina zaɓuɓɓuka
Ƙungiyoyin ƙirƙira za su iya samar da ra'ayoyi cikin sauri tare da ƙarancin bita da hannu.
Ƙungiya da aikin aiki
Ayyuka na iya amfani da siginar hoto da bidiyo waɗanda a baya suke da wahalar aiwatarwa.
Makomar Samfuran Ƙarfafa Maki
Tsarin maki-SDE shine ingin ka'idar da ke bayan yawancin ci gaban AI. Masu warware lambobi masu sauri, mafi kyawun jadawalin amo, da yuwuwar-gudanar ODE suna ba da damar samar da kusan lokaci-lokaci da kuma ainihin yuwuwar kimantawa. Irin wannan ra'ayin daidaita ma'auni yana yaduwa fiye da hotuna zuwa tsarin sauti, kwayoyin halitta da tsarin gina jiki, gizagizai, da kwaikwaiyo na kimiyya, yayin da daidaito da tsarin daidaitawa suna gina kai tsaye akan waɗannan tushe na lokaci-lokaci don rage tsararraki zuwa ɗimbin matakai.
Aiwatar da Gaskiyar Duniya
Hayaniya-Yanayin Cibiyoyin Cibiyoyin Sadarwa (NCSN) masu samar da fuskokin hoto ta hanyar bin matakan da aka koya ta hanyar haɓakar Langevin.
Sake gina hoton likitanci, kamar haɓakar MRI, inda ƙimar da aka koya ke aiki a matsayin kafin a cika bayanan sikanin da ba a ƙima ba.
Ƙirƙirar tsarin kwayoyin halitta da furotin a cikin gano magunguna, ƙirar ƙirar atomic 3D tare da rarraba tushen maki.
Haɗin sautin kalaman sauti inda ƙirar ƙira ke nuna tsaftataccen magana ko kiɗa, kamar a cikin masu amfani da ƙira.
Hatsari & Tsare-tsare
Haƙƙoƙin hoto da yarda na iya zama haxarin doka idan ba a fayyace ba.
Ayyukan samfuri na iya bambanta a ko'ina cikin haske, ƙididdiga, da mahalli.
Ƙarya tabbataccen ƙila ba za a iya lura da shi ba sai dai idan an kula da ƙofofin amincewa.
Taswirar Hanya
Ƙayyade ma'auni na karɓa don daidaito, tunowa, da farashi na kuskure.
Gwada tare da bayanan da suka dace da ainihin yanayin samarwa.
Ƙara bita na ɗan adam don ƙarancin amincewa ko tsinkaya mai tasiri.
Bi diddigin ƙirar ƙira kuma sake ingantawa bayan canje-canjen kamara ko saitin bayanai.
Ci gaba da Bincike
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 Score-Based Generative Models 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
Jagora na gaba
Samfuran Da Aka Gina Makamashi
Tambayoyin da ake yawan yi
What is Score-Based Generative Models?
Samfurin ƙirƙira tushen maki suna ƙirƙirar bayanai ta hanyar koyon gradient na rarraba bayanai - alkiblar da ke sa kowane samfurin hayaniya ya yi kama da ainihin bayanai. Wannan kallon-aiki-aiki yana haɗa nau'ikan watsawa tare da ma'auni daban-daban na stochastic kuma yana tallafawa yawancin masu samar da hoto na zamani.
Menene ainihin 'maki' da waɗannan samfuran suka koya?
Makin shine girman girman bayanan log dangane da shigarwar - yana nunawa a alkiblar da ke ƙara yuwuwar samfurin.
Me yasa ake horar da ƙirar tushen maki akan bayanai tare da amo a ma'auni da yawa?
A cikin ƙananan ƙananan yankuna, ƙimar gaskiya ba ta da kyau a bayyana; rikitattun bayanai tare da matakan amo da yawa yana sa madaidaicin maki a ko'ina.
Wanne tsarin samfuri suka yi amfani da farkon tushen makin don samar da bayanai?
Langevin kuzarin kawo cikas akai-akai matsar da samfur zuwa ga ci da kuma allura ƙaramar amo, a hankali canza bazuwar amo zuwa na haƙiƙa bayanai.
Wace babbar alaƙa ce takardar maki-SDE ta 2021 ta kafa?
Song et al. haɗe-haɗe da samfuran tushen maki a ƙarƙashin tsarin ma'auni daban-daban, tare da madaidaicin ƙayyadaddun yuwuwar kwarara ODE.
Ta yaya tsinkayar hayaniyar ke da alaƙa da tsinkayar ƙima a cikin ƙin madaidaicin maki?
Rashin daidaituwar maki yana nuna maki daidai da mummunan amo da aka raba ta hanyar bambance-bambancen amo, yin hasashen amo da tsinkayar maki daidai da manufa iri ɗaya.