Ikibazo gihuza ibibazo · Hagati Urwego

Ikigereranyo Cyibiro Bikigereranyo

Kugenzura uburemere bwa SWA impuzandengo, imyitwarire-igipimo cyo kwiga, icyiciro-gisanzwe cyo kwisubiramo nuburyo icyitegererezo kimwe cyagereranijwe gitandukanye na ensemble.

Inzira zijyanye no kuyoboraKugereranya Ibiro Bikomeye
Ikibazo 1 Bya 8

Kugereranya Uburemere bwa Stochastique ni ikihe?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Kugereranya Uburemere bwa Stochastique ni ikihe?
    • Ibipimo by'icyitegererezo byatoranijwe ku ngingo nyinshi bitinze mu myitozo
    • Ubuhanuzi bwa moderi zitandukanye zahuguwe zitandukanye
    • Gradients zabazwe kuri mini-matsinda yose
    • Igipimo cyo kwiga gikoreshwa muri buri gihe
  2. Kuki SWA ikunda kunoza rusange?
    • Ifata mu mutwe amahugurwa yashyizweho neza
    • Irasanga minima yagutse, iryoshye cyane ifite imbaraga kumakuru atagaragara
    • Yongera umubare wibipimo murugero
    • Bikuraho ibikenewe byo kwemeza
  3. Ni izihe ntambwe yinyongera SWA isaba imiyoboro ikoresha ibice bisanzwe?
    • Kongera imyitozo kuva kera
    • Kongera inshuro ebyiri
    • Inyongera yimbere yo kubara BN ikora imibare
    • Kuraho ibice byose byabata ishuri
  4. Ni iyihe myitwarire yo kwiga-isanzwe ikoreshwa mugihe cyo kugereranya SWA?
    • Igipimo cyo kwiga cyangirika kugeza kuri zeru
    • Igipimo cyo kwiga cya kimwe kugirango gihamye
    • Nta gipimo cyo kwiga na gato kuva amahugurwa yahagaritswe
    • Igipimo gihoraho cyangwa cyikigereranyo, ugereranije igipimo cyo kwiga
  5. Nigute ibiciro bya SWA byagereranijwe ugereranije na gakondo ya moderi ya N?
    • Igura hafi yicyitegererezo kimwe, kuva ibisubizo nimwe murwego rwo gupima
    • Nibihe N bihenze cyane
    • Birasa no gukora N moderi
    • Ntishobora gukoreshwa mugusuzuma
  6. Why does the batch-normalization adjustment belong after SWA weight averaging?
    • The averaging step deletes every convolution
    • Its running statistics must be recomputed for the averaged weights
    • It converts batch normalization into layer normalization
    • It restores the original optimizer learning rate
  7. An SWA running mean combines n previous samples with a new weight vector w_i. Which update gives equal weight to all n+1 samples?
    • w_SWA + n · w_i
    • (w_SWA + w_i) divided by n
    • (n · w_SWA + w_i) divided by (n + 1)
    • n · (w_SWA − w_i)
  8. Why is SWA commonly started during the later part of training?
    • It requires the model to remain at random initialization
    • It can only average models with different architectures
    • It replaces the training data with validation examples
    • It averages nearby useful checkpoints from a trajectory after the model has made substantial progress