Maswali ya Kujifunza kwa Kina ya Bayesian

Majibu na maelezo huonekana unapojibu maswali.

1. Mtandao wa neural wa Bayesian unashughulikia vipi uzani wake tofauti na ule wa kawaida?
   • Inachukulia kila uzani kama usambazaji wa uwezekano badala ya dhamana moja isiyobadilika
   • Inatumia uzani kamili badala ya desimali
   • Huondoa uzito kabisa
   • Inashiriki uzito mmoja kwenye tabaka zote
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. Kwa nini makadirio kama uelekezaji wa tofauti au MCMC inahitajika?
   • Kwa sababu mitandao ya neural haina kazi ya kupoteza
   • Kwa sababu GPU haziwezi kuongeza
   • Kwa sababu uzani lazima uwe nambari kamili
   • Kwa sababu sehemu halisi ya nyuma zaidi ya mamilioni ya uzani haiwezi kubatilika kwa hesabu
5. Monte Carlo walioacha shule hufanya nini kukadiria kutokuwa na uhakika?
   • Huweka walioacha shule kufanya kazi wakati wa jaribio na huendesha mtandao mara nyingi, kwa kutumia uenezaji wa matokeo
   • Huzima kuacha shule na kufanya mazoezi kwa muda mrefu
   • Huondoa niuroni zote kabisa
   • Hubadilisha hasara kwa kelele nasibu
6. Maoni tofauti hurekebisha makadirio ya nyuma kama shida ya aina gani?
   • Tatizo la kupanga
   • Tatizo la uboreshaji ambalo hupunguza tofauti za KL (kupitia ELBO)
   • Tatizo safi la sampuli nasibu bila lengo
   • Tatizo la muundo wa maunzi
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