Kedis ku xidhan hagaha · Dhexdhexaad Heerka

Kediska Xisaabinta itimaalka

Waxay hubisaa qeexitaannada cabbirka, jaantusyada la isku halleyn karo, ECE, cabbirida heerkulka iyo sababta tayada ansaxinta ay uga duwan tahay saxnaanta kala soocidda.

Wadooyinka haga ee la xidhiidhaQiyaasida ixtimaalka
Su'aal 1 ee 8

Qaabka ayaa si fiican loo jaangooyay haddii, saadaasha ka mid ah ay ku calaamadiso 80% kalsooni, dhacdadani dhacdo:

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Qaabka ayaa si fiican loo jaangooyay haddii, saadaasha ka mid ah ay ku calaamadiso 80% kalsooni, dhacdadani dhacdo:
    • Qiyaastii 80% wakhtiga
    • 100% waqtiga
    • Qiyaastii 20% wakhtiga
    • Dhab ahaan hal mar
  2. Muxuu cabirka heerkulka ku sameeyaa soosaarka shabakada neerfaha?
    • Isbedelka fasalka la saadaaliyay
    • Wuxuu u qaybiyaa lojisyada scalar la bartay ka hor softmax, hagaajinta kalsoonida
    • Wuxuu ku daraa buuqa calaamadaha tababarka
    • Waxay dib u tabobaraysaa shabakada xoqan
  3. Waa maxay jaantuska la isku halleyn karo?
    • Luminta tababarka iyo waagii
    • Cabbirka qaabka iyo saxnaanta
    • Itimaalka la saadaaliyay iyo inta jeer ee la arkay
    • Heerka waxbarashada iyo cabbirka dufcadda
  4. Cabbiraadda Qaladka Calibration ee la filayo (ECE):
    • Saxnaanta-1-sare ee moodeelka
    • Tirada xuduudaha ee qaabka
    • Kala duwanaanshaha gradients
    • Farqiga celceliska ah ee miisaanka u dhexeeya kalsoonida iyo saxnaanta qashinka
  5. Shabakadaha neerfaha qoto dheer ee casriga ah ayaa inta badan lagu tilmaamaa sida:
    • Kalsooni xad dhaaf ah
    • Kalsooni darro
    • Si sax ah loo habeeyey si toos ah
    • Aan awoodin in uu soo saaro itimaalka
  6. Waa maxay sababta qaabka aadka saxda ah uu weli ugu baahan karo qiyaasid?
    • Moodooyinka saxda ah had iyo jeer waxay leeyihiin eber khalad hagaajin ah
    • Saxnaanta iyo is-habaynta ayaa ah sifooyin kala duwan
    • Calibration kaliya waxay khusaysaa dib u noqoshada
    • Saxnimadu waxay dammaanad qaadaysaa ixtimaalka daacadda ah
  7. On which data is temperature scaling typically fit in the guide’s calibration workflow?
    • The final test set using classification accuracy
    • Only the training labels without model outputs
    • A held-out validation set using negative log-likelihood
    • The model’s hidden weights using mean squared error
  8. Why can temperature scaling improve calibration without changing top-1 accuracy?
    • It changes the model’s predicted class labels randomly
    • It retrains the feature extractor
    • It replaces every logit with the same value
    • Dividing logits by a positive scalar changes confidence but preserves their ordering