Can't always get closed form for the marginals. OTOH your question makes me feel like maybe I'm the one missing something obvious 
Why do people estimate the predictive posterior by regular sampling when it's exactly as hard as computing marginal likelihood? Please tell me I'm missing something obvious.
-
-
-
I mean, ML is integrating the product of likelihood and prior over all parameters; PP is integrating the product of likelihood and posterior over the same space. There are countless papers on approximating ML in various ways, but the standard prescription for PP is just sampling.
- Još 3 druga odgovora
Novi razgovor -
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.