Kediska Barashada Deep Bayesian

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Sidee bay shabkada neerfaha Bayesian ula dhaqantaa miisaankeeda si ka duwan kan caadiga ah?
   • Waxay ula dhaqantaa miisaan kasta sida qaybinta itimaalka halkii ay ka ahaan lahayd hal qiime go'an
   • Waxay isticmaashaa miisaan isku dhafan halkii ay ka isticmaali lahayd jajab tobanle
   • Waxay ka saartaa miisaanka gebi ahaanba
   • Waxay la wadaagtaa hal miisaan dhammaan lakabyada
2. What can a Bayesian approach provide that a single fixed-weight network does not directly represent?
   • A guarantee that every confidence score is calibrated
   • A distribution of predictions that reflects uncertainty about model weights
   • A proof that the model will detect every unfamiliar input
   • A way to remove all noise from the data
3. How can epistemic uncertainty change as relevant, representative training data are added?
   • It is guaranteed to reach zero with any added data
   • It is identical to label noise
   • It may decrease under the chosen model and data assumptions
   • It always increases linearly
4. Waa maxay sababta qiyaasaha loo baahan yahay sida kala-duwanaanta kala-duwanaanta ama MCMC?
   • Sababtoo ah shabakadaha neerfaha ma laha shaqo lumis
   • Sababtoo ah GPU-yadu ma samayn karaan isku-darka
   • Sababtoo ah miisaanku waa inuu noqdaa tirooyin
   • Sababtoo ah dhabarka saxda ah ee ka badan malaayiinta miisaanku waa mid aan xisaabin karin
5. Maxay ka tagtaa Monte Carlo si loo qiyaaso hubaal la'aanta?
   • Wuxuu ka dhigaa ka-tegidda mid firfircoon waqtiga imtixaanka wuxuuna maamulaa shabakadda marar badan, isagoo isticmaalaya faafinta wax soo saarka
   • Deminta ka bixista oo tareen dheer
   • Ka saara dhammaan neerfayaasha si joogto ah
   • Wuxuu khasaaraha ku beddelaa buuq aan toos ahayn
6. Kala duwanaanshuhu wuxuu dib u dhigayaa qiyaasta dambe inay tahay mushkilad noocee ah?
   • Dhib kala soocida
   • Dhibaatada hagaajinta ee yaraynaysa kala duwanaanshaha KL (iyada oo loo marayo ELBO)
   • Dhibaato muunado random saafi ah oo aan ujeeddo lahayn
   • Dhibaatada naqshadaynta qalabka
7. How does Bayes’ rule relate a prior, likelihood, and posterior?
   • The posterior is the prior plus the likelihood, without normalization
   • It is the likelihood divided by the prior without normalization
   • The posterior is proportional to the prior times the likelihood
   • It is the fixed point estimate of the most likely weight vector
8. How does MCMC approximate a Bayesian neural-network posterior?
   • By training a single point estimate with dropout disabled
   • By ranking model outputs with a fixed rule
   • By minimizing an ELBO with a simpler variational family
   • By drawing samples through a Markov chain whose stationary distribution targets the posterior