quiz bu lëkkale ak tegtal · Diggante Niveau
Quiz buy dagg degrade
Dafay saytu dagg norm ak valeur, baña yàq yoon wi, saytu gradient buy kalaate ak dagg misaal bu nekk ci tàggat yaram bu wuute.
Laaj 1 de 8
Lan la dagg gradient clip-by-norm di njëkka am ci di wàññi yeesal bi?
Xoolaatal laaj bu nekk8 laaj
Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.
- Lan la dagg gradient clip-by-norm di njëkka am ci di wàññi yeesal bi?
- Jubluwaayu yeesali gradient bi
- Dooley gradient bu nekk
- Limu paramet model yi
- Jadwalu tolluwaayu jàng
- Ci clip-by-norm ak threshold c, lan mooy xew su norm gradient bi g_norm weesee c?
- Degrade yi dañu leen tëral ci nul
- Bépp degrade dañu koy yokk ci c/g_norm
- Taggat yaram dafay taxaw
- Tolluwaayu jàng yokk na ñaari yoon
- Ban jafe-jafe lañu jagleel gradient clipping ngir xeex?
- Degrade yuy réer
- Done yu yeex
- Degrade yuy kalaate
- Overfitting ci ensemble validation bi
- Lan moo wuutale clip-par-valeur ak clip-par-norme?
- Dafay faral di tëye yoon wiñ yeesal
- Du dox ludul ci diggu convolution
- Dina dindi gradient yi yépp
- Dafay tënk bépp cër ci degrade bi ci benn rang, bu mëna soppi yoon wi
- A minibatch produces an unusually large global gradient norm. What does clip-by-norm do before the optimizer update?
- Scales the full gradient vector down to the configured threshold when the norm exceeds it
- Changes the model’s labels
- Drops every gradient component independently
- Increases the learning rate until the loss recovers
- Ci tàggat yaram bu wuute (DP-SGD), ban cër la dagg gradient ci misaal bu nekk?
- Dafay encrypter done yi
- Dafay tënk njeexitalu misaal bu nekk suko defee bruit buñ kalibre mën na gaaraati nëbbëtu
- Dafay yokk dayo lots bi
- Dina dindi soxla fonction perte
- The global gradient norm is twice a clipping threshold c. By what factor is the gradient scaled?
- Ñaar
- One quarter
- One half (c divided by 2c)
- It is left unchanged
- Why does DP-SGD clip each example’s gradient before adding calibrated noise?
- To make all examples have the same label
- To remove the need for a privacy budget
- To increase the norm of rare-class gradients
- To bound an individual example’s influence on the released update