Maswali yanayounganishwa na mwongozo · Ngumu Kiwango

Maswali ya Malengo ya InfoNCE na SimCLR

Hukagua lengo la InfoNCE chanya dhidi ya hasi, maoni na kichwa cha makadirio ya SimCLR, na majukumu ya halijoto na saizi ya bechi.

Njia za mwongozo zinazohusianaMaelezo na Malengo ya Simclr
Swali 1 ya 8

Je, lengo la InfoNCE linahimiza mwanamitindo kufanya nini?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Je, lengo la InfoNCE linahimiza mwanamitindo kufanya nini?
    • Fanya swali lifanane zaidi na chanya yake ya kweli kuliko hasi nyingi
    • Kariri thamani halisi za pikseli za kila picha
    • Punguza idadi ya vigezo vya mfano
    • Bashiri neno linalofuata katika sentensi
  2. Katika SimCLR, jozi chanya inaundwaje?
    • Kwa kuoanisha picha na lebo yake ya kibinadamu
    • Kwa kutumia nyongeza mbili tofauti za nasibu kwa picha moja
    • Kwa kuchagua picha mbili tofauti kabisa
    • Kwa kuzungusha mkusanyiko mzima wa data
  3. Joto τ lina jukumu gani katika NT-Xent / InfoNCE?
    • Inaweka kiwango cha kujifunza
    • Inadhibiti ukubwa wa kundi moja kwa moja
    • Inapima alama za kufanana; chini τ kunoa usambazaji na kusisitiza hasi ngumu
    • Huamua idadi ya tabaka kwenye kisimbaji
  4. Ni nini kinatokea kwa kichwa cha makadirio ya SimCLR baada ya kufanya mazoezi ya awali?
    • Inakuwa kiainishaji cha mwisho
    • Imegandishwa na kutumika kama kisimbaji
    • Imenakiliwa kwa hitimisho
    • Inatupwa; uwakilishi kabla ya kichwa kutumika kwa kazi za chini ya mkondo
  5. Kwa nini SimCLR ilitegemea saizi kubwa za batch?
    • Kutoa mifano mingi hasi ndani ya kila hatua ya mafunzo
    • Ili kupunguza matumizi ya kumbukumbu
    • Ili kuepuka kutumia augmentations
    • Ili kuondoa hitaji la parameter ya joto
  6. In SimCLR, which examples usually act as negatives for one anchor view?
    • The matching view of the same image
    • The other augmented views in the batch that come from different images
    • Only examples with the same class label
    • Views produced by a second encoder trained on labels
  7. Why does SimCLR use a nonlinear projection head before computing its contrastive loss?
    • It replaces the image encoder
    • It supplies class labels during pretraining
    • The learned projection helps the contrastive objective while the representation before the head can transfer downstream
    • It removes the need for data augmentation
  8. For an InfoNCE anchor, what should happen to the positive similarity relative to sampled negatives?
    • Every negative should have a higher score than the positive
    • All scores must be identical
    • The positive is excluded from the objective
    • The positive should receive a higher score than the negatives