Ikibazo gihuza ibibazo · Hagati Urwego

Ikibazo cya Adam na Adaptive Optimizers Ikibazo

Kugenzura ibihe bya Adam buhoro buhoro, gukosora kubogama, gupima imihindagurikire y'ikirere, epsilon na AdamW igabanije ibiro-bigabanuka. muburyo busanzwe bwo kuvugurura.

Inzira zijyanye no kuyoboraAdam na Adaptive Optimizers
Ikibazo 1 Bya 8

Nibihe bintu bibiri Adamu akurikirana kuri buri kintu?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Nibihe bintu bibiri Adamu akurikirana kuri buri kintu?
    • Ikigereranyo cyo kwiruka cya gradients hamwe na kare ya gradients
    • Ikigereranyo ntarengwa kandi ntarengwa cyigeze kiboneka
    • Igipimo cyo kwiga nubunini bwicyiciro
    • Agaciro k'uburemere n'ikimenyetso cyacyo
  2. Kuki Adamu akoresha kubogama kubigereranyo byigihe?
    • Kuberako gradients zihora ari mbi hakiri kare mumahugurwa
    • Kuberako impuzandengo yimuka itangirira kuri zeru kandi ibogamye kuri zeru ubanza
    • Kugirango igipimo cyo kwiga kibe kinini mugihe
    • Gukuraho amakuru yimbere yo hanze
  3. Ni irihe tandukaniro nyamukuru riri hagati ya Adam na AdamW?
    • AdamW ntabwo ikoresha imbaraga na gato
    • AdamW ikuraho ikigereranyo cyakabiri
    • AdamW yikuramo ibiro kugabanuka kuva kuri gradient-ishingiye ku ivugurura
    • AdamW ikora gusa kumurongo wemeza
  4. Ni uruhe ruhare ijambo rito rya epsilon rigira mu mategeko yo kuvugurura Adamu?
    • Ishiraho imbaraga
    • Igenzura umubare wigihe cyamahugurwa
    • Yongera ubunini bwa gradient
    • Irinda kugabana na zeru kandi igahindura ivugurura
  5. Kuki Adamu akunze kuvugwa ko 'ahuza n'imiterere'?
    • Itanga neza buri kintu ibipimo byacyo byapimye
    • Irahindura imiyoboro yububiko mugihe cyamahugurwa
    • Ihuza dataset nicyitegererezo
    • Ihinduranya hagati yo gutondeka no gusubira inyuma mu buryo bwikora
  6. Which Adam parameter controls the exponential decay rate of its squared-gradient estimate?
    • Beta1
    • Beta2
    • Epsilon
    • The weight-decay coefficient
  7. A parameter repeatedly receives large gradients. Which Adam quantity helps scale down its effective step?
    • Umubare wicyitegererezo
    • The batch-normalization mean
    • The second-moment estimate of squared gradients
    • The AdamW weight-decay term alone
  8. What happens when beta2 is set closer to one, all else equal?
    • The optimizer stops using gradients
    • The first-moment estimate is deleted
    • Weight decay becomes decoupled automatically
    • The squared-gradient estimate decays more slowly and retains a longer history