Hacker News new | ask | show | jobs
by MarkusQ 53 days ago
"the model can alter its priors"

The model doesn't have high-level priors in the Bayesian sense (though you could have priors about it).

The low-level priors it does have (the weights) are not modified by the context.