Hi, I’m trying to implement a series of models for binary adjacency matrices. Some of these models have multi-modal posteriors, so I decided to use PyMC’s SMC approach. For several models this works like a charm, but the following Stochastic Block Model I have not been able to get to work. It can …

That’s a more structural problem with the logp evaluation: import pymc as pm import aesara.tensor as at with pm.Model() as m: x = pm.Categorical("x", p=[0.1, 0.3, 0.6]) y = pm.Normal("y", mu=at.constant([0, 1, 2])[x]) m.compile_logp()({"x": 4, "y": 0}) # IndexError I think by default the C…

Model works with NUTS/prior predictive, but not with SMC?

Questions v5

mhinne June 10, 2022, 2:43pm 4

No worries

Might the issue be related to SMC not using discrete proposals within the sampling chains?

Topic		Replies	Views
Bug (?): NaN SMC weights and IndexOutOfBoundsError in theano Questions	4	986	October 28, 2018
TypeError in NUTS Questions	5	656	May 9, 2018
Model yeilding no samples Questions	1	395	July 31, 2018
Pm.step_methods() doesn't include pm.SMC() in pymc3 v 3.8 Questions	6	820	April 17, 2020
Sampling time is very long (minutes per sample) Questions	13	4527	November 2, 2017

Model works with NUTS/prior predictive, but not with SMC?

Related topics