Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Udhibiti

Angalia malengo ya urekebishaji, adhabu za L1 na L2, kuacha shule, kuacha mapema, nguvu ya adhabu na mifano ya nambari ya lengo lililoratibiwa.

Njia za mwongozo zinazohusianaUrekebishaji
Swali 1 ya 8

Why would a practitioner add regularization when a model fits training-specific noise?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Why would a practitioner add regularization when a model fits training-specific noise?
    • To guarantee zero training error
    • To remove the need for validation data
    • To guarantee a faster GPU kernel
    • To discourage overfitting and evaluate whether unseen-data performance improves
  2. Ni mbinu ipi ya urekebishaji ambayo ina uwezekano mkubwa wa kuongeza uzani hadi sifuri, kwa kuchagua vipengele kwa ufanisi?
    • Urekebishaji wa L2 (kuoza kwa uzito)
    • Udhibiti wa L1
    • Urekebishaji wa kundi
    • Kuongeza kiwango cha kujifunza
  3. Jinsi gani kuacha shule kunarekebisha mtandao wa neva wakati wa mafunzo?
    • Inafuta safu kabisa kutoka kwa mtandao
    • Inaongeza uzani wa mraba kwa utendaji wa upotezaji
    • Huzimisha nyuroni bila mpangilio kila hatua ili zisiweze kukabiliana
    • Inaacha mafunzo wakati hasara ya uthibitishaji inapoongezeka
  4. Katika upotezaji wa kawaida 'kosa + lambda × adhabu', kuongeza lambda hufanya nini?
    • Inaimarisha adhabu, ikisukuma mfano kuelekea suluhisho rahisi
    • Hupunguza adhabu na kuruhusu uzani mkubwa
    • Huondoa neno la makosa kabisa
    • Haina athari kwa uzito
  5. Kwa nini kuacha mapema kunaweza kuzingatiwa kama njia ya kuhalalisha?
    • Kwa sababu inaongeza adhabu ya uzito kwa hasara
    • Kwa sababu inasimamisha mafunzo kabla ya mfano kuanza kuweka kelele kwenye data
    • Kwa sababu huongeza uwezo wa mfano kwa wakati
    • Kwa sababu huondoa neurons kwa nasibu
  6. What can happen when regularization is strengthened beyond what a task needs?
    • Both training and test accuracy must improve
    • The penalty becomes unrelated to the objective
    • Validation becomes unnecessary
    • The model can underfit useful structure even if variance decreases
  7. For a regularization example with weights minus two and one, what are the unscaled L1 and sum-of-squares L2 penalties?
    • L1 is three and L2 is five
    • L1 is minus one and L2 is three
    • L1 is five and L2 is three
    • Both penalties are minus two
  8. A regularized objective has prediction error two, lambda one-half, and penalty six. What is its total value?
    • Nane
    • Tatu
    • Tano
    • Moja