Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Kuacha na Kudhibiti Udhibiti wa Stochastic

Jaribu vinyago nasibu, modi za mafunzo na uelekezaji, viwango vilivyogeuzwa, vikomo vya urekebishaji shirikishi, kuacha shule kwa MC na ubashiri tofauti.

Njia za mwongozo zinazohusianaKuacha na Udhibiti wa Stochastic
Swali 1 ya 8

Kuacha shule hufanya nini wakati wa mafunzo?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Kuacha shule hufanya nini wakati wa mafunzo?
    • Hufuta uzani kabisa
    • Nasibu huweka sehemu ya matokeo ya neuroni hadi sufuri kwa kila pasi ya mbele
    • Huongeza kasi ya kujifunza mara mbili
    • Huongeza tabaka zaidi kwenye mtandao
  2. Kwa nini kuacha shule kunaboresha ujanibishaji?
    • Inaongeza idadi ya vigezo
    • Inaondoa hitaji la data ya mafunzo
    • Inahakikisha kosa la mafunzo ya sifuri
    • Inazuia niuroni kubadilika kwa kushirikiana na kulazimisha uwakilishi usio na nguvu na thabiti
  3. Nini kinatokea kwa kuacha shule wakati wa mtihani (maelekezo)?
    • Imezimwa na mtandao kamili unatumiwa
    • Inapunguza neurons zaidi
    • Inafuta nusu ya uzito
    • Inabadilisha lebo
  4. Kiwango cha kuacha p = 0.5 kwenye safu kinamaanisha nini wakati wa mafunzo?
    • Neurons zote zinaondolewa
    • Hakuna niuroni zinazoathiriwa
    • Takriban nusu ya nyuroni zimetolewa kwa kila pasi
    • Safu imefutwa
  5. Ni nini kuacha shule mara nyingi hufafanuliwa kama kukadiria?
    • Urejeshaji wa mstari mmoja
    • Mti wa maamuzi
    • Mabadiliko ya Fourier
    • Mkusanyiko wa mitandao mingi nyembamba inayoshiriki uzani
  6. Monte Carlo dropout inatumika kwa nini?
    • Kuharakisha upakiaji wa data
    • Kukadiria kutokuwa na uhakika kwa mfano kwa kudumisha kuacha shule katika makisio
    • Kubana mtandao kwa uzani mdogo
    • Kuweka lebo data ya mafunzo kiotomatiki
  7. With inverted dropout, why is no extra activation scaling normally needed at ordinary inference?
    • The keep probability becomes exactly zero
    • Inference always averages new random masks
    • Surviving activations were already scaled during training
    • Every weight is divided by the number of layers
  8. The same input yields varying predictions because dropout remains active during evaluation. What explains the variation?
    • New random masks sample different subnetworks on each pass
    • The input changes its label by itself
    • The network has no trained weights
    • Batch size must be one