Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

Pipadanu Idojukọ fun Idanwo Aiṣedeede

Checks the focal-loss modulator, class imbalance, the role of alpha and gamma, and how well-classified and hard examples differ.

Jẹmọ itọsọna awọn ọnaPipadanu Idojukọ Fun Wiwa Aiṣedeede
Ibeere 1 ti 8

Isoro koko wo ni wiwa nkan ipele kan jẹ pipadanu idojukọ ti a ṣe lati koju?

Awotẹlẹ gbogbo ibeere8 ibeere

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

  1. Isoro koko wo ni wiwa nkan ipele kan jẹ pipadanu idojukọ ti a ṣe lati koju?
    • Iwontunwonsi to gaju laarin awọn nkan iwaju ti o ṣọwọn ati ipilẹ irọrun lọpọlọpọ
    • Iyara ifọkansi ti o lọra ti awọn aṣawari ipele-meji
    • Aini ti ike data ikẹkọ
    • Overfitting to tobi ohun
  2. Kini ifosiwewe modulating ti ipadanu idojukọ pọ si lori agbelebu-entropy?
    • log(p_t)
    • (1 - p_t)^gamma
    • p_t^2 nikan
    • 1 / p_t
  3. Kini yoo ṣẹlẹ si pipadanu idojukọ nigbati a ṣeto gamma paramita idojukọ si 0?
    • O di aṣiṣe onigun mẹrin tumọ si
    • O foju gbogbo awọn apẹẹrẹ rere
    • O di arinrin agbelebu-entropy
    • O diverges si ailopin
  4. Kini gamma ti o pọ si ṣe si irọrun, awọn apẹẹrẹ iyasọtọ daradara?
    • Ṣe alekun ilowosi ipadanu wọn
    • Fi ilowosi wọn silẹ ko yipada
    • Yọ wọn kuro ninu dataset
    • Dinku ilowosi ipadanu wọn diẹ sii ni agbara
  5. Eyi ti aṣawari faaji ṣe afihan pipadanu idojukọ?
    • RetinaNet
    • YOLOv1
    • Yiyara R-CNN
    • AlexNet
  6. Ipa wo ni paramita alpha nigbagbogbo n ṣiṣẹ lẹgbẹẹ gamma?
    • O ṣeto oṣuwọn ẹkọ
    • O ṣe iwọntunwọnsi iwuwo ibatan ti rere dipo awọn kilasi odi
    • O nṣakoso ipinnu aworan
    • O asọye awọn nọmba ti oran apoti
  7. For a hard example with p_t close to zero, what is the focal modulating factor (1 − p_t)^gamma likely to do?
    • Drive the factor to zero
    • Turn the example into a positive label
    • Keep the factor relatively close to one so the example retains substantial loss weight
    • Replace the cross-entropy term with alpha
  8. A dense detector has many easy background locations. What effect does focal loss aim to have on their combined training influence?
    • Increase their count by duplicating them
    • Remove all positive object examples
    • Make every example receive an identical loss
    • Reduce their dominance so training can focus more on sparse hard examples