Guide-yakabatana mibvunzo · Zvakaoma Nhanho

Contrastive Learning Quiz

Dzidzira kusiyanisa kwakanaka vaviri vaviri, kunyudza kudonha, uye tembiricha-yakayerwa kufanana, kusanganisira kuti mitsetse yeMoCo inosiyana sei kubva kune inference cache.

Nzira dzekutungamira dzinoenderanaContrastive Learning
Mubvunzo 1 ye 8

Ndechipi chinangwa chikuru chekudzidza kwakasiyana?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. Ndechipi chinangwa chikuru chekudzidza kwakasiyana?
    • Dhonza mapairi akafanana (akanaka) pamwe chete uye sunda maviri akasiyana (negative) akaparadzana munzvimbo yekumisikidzira.
    • Fanofanotaura iwo chaiwo pixel kukosha kwemufananidzo
    • Rongedza zvese zvinoiswa muimwe yezvirongwa zvakagadziriswa
    • Kuderedza nhamba yemuenzaniso parameters
  2. Munzira yekuzvitarisira yemufananidzo seSimCLR, mbiri yakanaka inogadzirwa sei?
    • Mifananidzo miviri yakasiyana yekirasi yakafanana yakanyorwa
    • Mufananidzo uye mufananidzo wakasarudzika wakasarudzwa kubva kune dataset
    • Kuwedzera kuviri kwakangoerekana kwaitika (semuenzaniso, zvirimwa, jitter yemavara) emufananidzo mumwe chete
    • Mufananidzo wakapetwa nemasiki ayo echikamu
  3. Nderipi basa rekurasikirwa rinonyanya kuenderana nekudzidza kwakasiyana?
    • Kureva kukanganisa kwakapetwa kaviri
    • InfoNCE
    • Yambuka-entropy pamusoro pezvinyorwa zvekirasi
    • Hinge kurasikirwa kweSVMs
  4. Sei nzira dzakaita seMoCo dzichishandisa memory bank kana mutsara?
    • Kuchengeta mavara ekugadzirisa zvakanaka
    • Kuchengeta fungidziro dzekupedzisira kuti dziite nekukurumidza
    • Kumanikidza modhi kune nharembozha
    • Kupa mienzaniso yakawanda yakaipa pasina kuda yakakura batch size
  5. Idambudziko ripi iro BYOL neSimSiam vanonyatso dzivirira pasina kushandisa zvakajeka zvisiri izvo?
    • Overfitting kune zvinyorwa zvekudzidzira
    • Kuputika gradients mune optimizer
    • Chiratidziro chinopunzika, uko zvese zvakamisikidzwa zvinofanana
    • Vanishing kudzidza mwero
  6. CLIP inotambanudza sei pfungwa yekusiyanisa?
    • Iyo inoenzanisa mifananidzo neayo zvinyorwa zvinyorwa munzvimbo yakagovaniswa yekumisikidza
    • Inosiyanisa maviri ekuwedzera eiyo imwechete odhiyo clip
    • Inobvisa kudiwa kwechero neural network
    • Inoshandisa kusimbisa kudzidza pachinzvimbo chekufanana
  7. A contrastive image encoder maps every photo to the same vector. Which failure does this illustrate?
    • A useful separation of image categories
    • An increase in the number of negative samples
    • A perfect nearest-neighbor representation
    • Representation collapse
  8. Why is MoCo’s training queue different from caching completed inference responses?
    • It serves final chatbot responses without model execution
    • It supplies encoded comparison examples for a contrastive training objective
    • It replaces every positive image with a label
    • It removes the need to update any encoder