Ikibazo gihuza ibibazo · Hagati Urwego

Ikibazo cya Bottleneck Kubaka

Kugenzura kugabanya umuyoboro, kubara kuzigama, ibisigisigi byahinduwe, kuvugurura urwego rwo hasi hamwe nimbogamizi zisangiwe inyuma yibibazo.

Inzira zijyanye no kuyoboraBottleneck Ubwubatsi
Ikibazo 1 Bya 8

Mubice bya ResNet, ni uruhe ruhare rwambere rwa 1x1?

Reba buri kibazo8 ibibazo

Ibisubizo nibisobanuro bigaragara mugihe ufata ikibazo.

  1. Mubice bya ResNet, ni uruhe ruhare rwambere rwa 1x1?
    • Kugabanya umubare wimiyoboro mbere yo kwemeza 3x3
    • Kongera imiterere yikibanza
    • Gushyira mu bikorwa icyiciro cya nyuma
    • Kongera urusaku kubisanzwe
  2. Ni ukubera iki icyuho gituma imiyoboro yimbitse ihendutse kubara?
    • Ikuraho ibikorwa byose
    • Ikora ibikorwa bihenze kumiyoboro mike
    • Ikoresha intoki nini ahantu hose
    • Isimbuka gusubira inyuma
  3. Muri autoencoder, icupa ryerekana mbere na mbere:
    • Guhitamo neza
    • Imiterere ya dosiye
    • Ni bangahe ibyinjijwe bigabanijwe
    • Ingano yicyiciro
  4. MobileNetV2 ikoresha niyihe variant yigitekerezo cyacitse?
    • Gutobora neza
    • Gusubira inyuma
    • Nta gucogora na gato
    • Guhindura ibisigisigi bisigaye (kwagura noneho amasezerano)
  5. Nigute LoRA ikoresha amahame agenga imiterere yindimi nini?
    • Mugushyiramo umutoza wo hasi-urwego (rugufi) icyuho muguhindura ibiro
    • Mugusiba ibitekerezo
    • Mugukuba kabiri uburemere bwose
    • Kubara kuri 1 bit
  6. Inzira isanzwe ya ResNet ihagarikwa ikurikira niyihe nzira?
    • Mugari, mugari, mugari
    • Mugabanye, gutunganya, kugarura (urugero, 256 kugeza 64 kugeza 256)
    • Gito, ubugari, bugufi
    • Bisanzwe kuri buri cyiciro
  7. A ResNet block reduces 256 channels to 64 before its 3×3 convolution. Ignoring the 1×1 layers, what is the 3×3 layer’s approximate multiply-add cost relative to 256 channels?
    • About one-quarter
    • About one-half
    • About one-sixteenth
    • About four times as much
  8. An autoencoder uses a narrow latent code. What constraint does this impose on reconstruction?
    • The encoder must preserve every input pixel exactly
    • The network can no longer use learned weights
    • The decoder must expand the latent code before encoding
    • The decoder must reconstruct the input from a compressed representation