Aami Dan adanwo

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

1. Kini aami iyipada smoothing nipa awọn ibi-afẹde ikẹkọ?
   • O rọ awọn ibi-afẹde ọkan-gbona nitorina kilasi otitọ wa ni isalẹ 1.0 ati pe awọn miiran wa diẹ sii ju 0 lọ
   • O yọ iṣẹ isonu kuro
   • O npa awọn kilasi ti ko tọ
   • O ṣe ilọpo meji oṣuwọn ẹkọ
2. Iye ti o wọpọ fun paramita didan epsilon jẹ:
   • 0.9
   • 0.1
   • 5.0
   • 0.0
3. What is one intended effect of replacing hard one-hot targets with smoothed targets?
   • Force every class to have equal probability
   • Guarantee the largest possible logit for the true class
   • Avoid requiring the labeled class to receive a target probability of exactly one
   • Remove the need for labeled training data
4. Which landmark architectures are prominent examples of label-smoothing use?
   • AlexNet and LeNet only
   • ResNet-50 and VGG only
   • AlphaGo ati AlphaZero
   • Ibẹrẹ-v3 ati Amunawa
5. What can label smoothing do to calibration?
   • It can improve calibration in some settings, but the effect is task- and model-dependent
   • It guarantees perfect calibration for every classifier
   • It cannot affect predicted probabilities
   • It always makes confidence equal accuracy
6. Kini isale isalẹ ti aami didan?
   • O nigbagbogbo din išedede
   • O le nu itanran laarin-kilasi ibajọra, ipalara imo distillation
   • O nilo lẹmeji data ikẹkọ
   • O ṣiṣẹ nikan lori ọrọ
7. Why can label smoothing complicate knowledge distillation?
   • It converts every class label into an image
   • It forces the student to use a different tokenizer
   • It can remove information in logits about similarities among incorrect classes
   • It prevents the teacher from producing logits
8. What should a practitioner verify when comparing label-smoothing implementations?
   • Whether the model has a web browser
   • Whether training uses a validation filename
   • Whether logits are stored as text
   • Whether epsilon is distributed over all classes or only the non-target classes