Dropout ati Sitokasitik Regularization adanwo

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

1. Kini yiyọ kuro ṣe lakoko ikẹkọ?
   • Npa awọn iwuwo rẹ kuro patapata
   • Laileto ṣeto ida kan ti awọn abajade neuron si odo lori iwe-iwọle siwaju kọọkan
   • Ilọpo meji oṣuwọn ẹkọ
   • Ṣe afikun awọn ipele diẹ sii si nẹtiwọọki naa
2. Kini idi ti yiyọ kuro ni ilọsiwaju gbogbogbo?
   • O mu ki awọn nọmba ti sile
   • O yọ iwulo fun data ikẹkọ kuro
   • O ṣe onigbọwọ aṣiṣe ikẹkọ odo
   • O ṣe idilọwọ awọn neuronu lati isọdọtun ati fi agbara mu laiṣe, awọn aṣoju to lagbara
3. Kini yoo ṣẹlẹ si sisọ silẹ ni akoko idanwo (itọkasi)?
   • O wa ni pipa ati nẹtiwọọki kikun ti lo
   • O ṣubu paapaa awọn neuronu diẹ sii
   • O npa idaji awọn iwuwo rẹ kuro
   • O randomizes awọn aami
4. Oṣuwọn yiyọkuro ti p = 0.5 ni ipele kan tumọ si ni aijọju kini lakoko ikẹkọ?
   • Gbogbo awọn neuronu kuro
   • Ko si awọn neuronu ti o kan
   • Nipa idaji awọn neuronu ti wa ni odo jade kọọkan kọja
   • Layer ti paarẹ
5. Kini isọ silẹ nigbagbogbo ṣe apejuwe bi isunmọ?
   • A nikan laini ifaseyin
   • Igi ipinnu
   • A Fourier yipada
   • Ijọpọ ti ọpọlọpọ awọn nẹtiwọọki tinrin ti o pin awọn iwuwo
6. Kini Monte Carlo dropout ti a lo fun?
   • Iyara soke data ikojọpọ
   • Iṣiro aidaniloju awoṣe nipa titọju yiyọ kuro lọwọ ni itọkasi
   • Fifun nẹtiwọọki si awọn iwuwo diẹ
   • Iforukọsilẹ data ikẹkọ laifọwọyi
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