s******a 发帖数: 184 | 1 对于连续变量做Bayesian analysis, likelihood function往往是0或者非常小的数.程
序往往就blowup 了或根本产生不了什么有意义的结果。
请问有没有数值上的解决方案,或者对于连续变量的Bayesian analysis有特殊的
方法。 | z****g 发帖数: 1978 | 2 don't know what you are talking about. Anytime there is Bayesian, there is
at least a numerical postier distribution. Even if it is small, you still
need to normalize it. Don't you use the log-likelihood function? |
|