Quiz moyenne de poids stochastique

Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.

1. Lan mooy moyenne poids stochastic ci dëgg?
   • Diisaayu model bi ñu jël ci poñ yu bari ci diggu tàggat yaram
   • Xalaatal yu bari ci xeetu model yuñ tàggat bu wuute
   • degrade yi ñu xayma ci lote yu ndaw yépp
   • Tolluwaayu jàng yi ñuy jëfandikoo ci jamono bu nekk
2. Lan moo waral SWA di gëna suqali généralisasioŋ bi?
   • Dafay gëna xam tàggat yaram bi
   • Dafay gis minima yu gëna yaatu, gëna plat yu gëna dëgër ci done yuñu gisul
   • Dafay yokk limu parametre yi ci model bi
   • Dafay dindi soxla ensemble validation
3. Ban jéego bu gëna am solo la SWA soxla ci reso yiy jëfandikoo normalisasioŋ ci lots?
   • Taggaat reso bi ci noonu
   • Ñaari yoon yokk dayo lote bi
   • Beneen paas ngir xaymawaat lim yi BN di daw
   • Di dindi bépp couche dropout
4. Ban jeffin juy jàng mooy gëna bari luñu koy jëfandikoo ci wàllu moyenne SWA?
   • Tolluwaayu jàng daa wàññeeku ba amul dara
   • Njàngat bu benn ngir dakkal
   • Amul benn njàng ndax dañu tëj tàggat yaram
   • Tolluwaayu jàng bu sax wala buy wëréelu
5. Lan mooy njëgu inference bu SWA buñu ko méngale ak ensemble bu cosaan bu N models?
   • Lu tollu ci benn xeetu njëg la, ndax limuy génne mooy benn xeetu poid
   • Dafa seer N yoon
   • Dafay nuru ak daw model N
   • Mënu ñu ko jëfandikoo ngir def ay tegtal
6. Why does the batch-normalization adjustment belong after SWA weight averaging?
   • The averaging step deletes every convolution
   • Its running statistics must be recomputed for the averaged weights
   • It converts batch normalization into layer normalization
   • It restores the original optimizer learning rate
7. An SWA running mean combines n previous samples with a new weight vector w_i. Which update gives equal weight to all n+1 samples?
   • w_SWA + n · w_i
   • (w_SWA + w_i) divided by n
   • (n · w_SWA + w_i) divided by (n + 1)
   • n · (w_SWA − w_i)
8. Why is SWA commonly started during the later part of training?
   • It requires the model to remain at random initialization
   • It can only average models with different architectures
   • It replaces the training data with validation examples
   • It averages nearby useful checkpoints from a trajectory after the model has made substantial progress