I have a model that contains 5 parameters. let me describe my motivation and later the problem I am facing.
-
One model parameter has a non-informative prior, I used
pm.HalfFlat()
because the parameter is practically un-identifiable… I want to obtain a proper posterior. In the past I’ve been usingpm.Uniform()
but the posterior was always very similar to the prior. -
The execution gets stagnated exactly at the half… See the next terminal outputs.
Sampling 4 chains, 0 divergences: 50%|████████████████████████████████████████████████████████████ | 202001/404000 [3:51:12<6:39:55, 8.42draws/s]
Sampling 4 chains, 0 divergences: 50%|███████████████████████████████████████████████████████████████ | 22000/44000 [22:30<37:57, 9.66draws/s]
Even changing (reducing) the number of samples I get this problem. Any advice will be helpful…