Idanwo Didinku Sharpness-Aware

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

1. Iru kere wo ni SAM gbiyanju lati wa?
   • O kere alapin nibiti pipadanu duro kekere kọja agbegbe ti awọn iwuwo
   • O kere julọ ti o ṣeeṣe
   • Awọn kere pẹlu awọn julọ sile
   • Eyikeyi o kere de ọdọ ni igbesẹ kan
2. Bawo ni ọpọlọpọ siwaju-sẹhin kọja ni igbesẹ SAM boṣewa nilo?
   • Idaji kọja
   • O fẹrẹ to meji
   • Gangan ọkan
   • Mẹwa tabi diẹ ẹ sii
3. Kini iṣakoso hyperparameter rediosi rho ni SAM?
   • Iyara ibajẹ ẹkọ
   • Awọn nọmba ti ikẹkọ epochs
   • Iwọn agbegbe ni ayika awọn iwuwo ti o ni aabo lodi si awọn spikes pipadanu
   • Awọn iṣeeṣe dropout
4. Kini akọkọ ti awọn igbesẹ meji SAM ni aṣetunṣe kọọkan?
   • Din awoṣe
   • Apapọ awọn iwuwo ti gbogbo awọn ti o ti kọja awọn igbesẹ
   • Tun ipo oluṣeto
   • Ṣe igbesẹ gigun ti iwọn rho si aaye ti o buruju ti o wa nitosi
5. Kini idi ti SAM nigbagbogbo ṣe ilọsiwaju agbara lati ṣe aami ariwo?
   • Nitoripe minima alapin ko ni itara lati ṣe akori awọn akole ibaje
   • Nitoripe o npa awọn aami alariwo rẹ laifọwọyi
   • Nitoripe o nlo awọn ọrọ ti o tobi ju
   • Nitori ti o meteta awọn ipele iwọn
6. Ewo ninu iwọnyi jẹ iyatọ ti a ṣe lati dinku idiyele iṣiro SAM?
   • ReLU
   • ESAM (SAM ti o munadoko)
   • BatchNorm
   • AdaGrad
7. In SAM’s second pass, which gradient is used to update the original model weights?
   • The gradient of a random label
   • The first pass gradient without evaluating the perturbed point
   • The gradient evaluated at the perturbed nearby weights
   • A gradient from a different model checkpoint
8. If SAM’s neighborhood radius rho is set to zero, what happens to its perturbation step?
   • The model parameters are all set to zero
   • The loss becomes independent of the weights
   • The optimizer takes an infinite step
   • There is no weight perturbation, so the neighborhood reduces to the current point