I’m starting to look at mixture models. I have a toy liner regression where I’m trying to infer the same latent parameter by choosing the linear operator which has the least noise. Using pm.Mixture, it works as expected: import pymc3 as pm np.random.seed(2) m = 60 n = 3 eps = 0.01 siga = 1 H = np.…

When you write down a PyMC model you are specifying a generative process, not the probability expression (which is the case in Stan). So this line Y = a1*pm.Normal(f'y1', mu = pm.math.dot(H1, x), sd=eps, observed=y) + \ (1-a1)*pm.Normal(f'y2', mu = pm.math.dot(H2, x), sd=eps, observed=y) …

Mixture model using pm.Mixture or writing in parts

Questions version agnostic

lwest August 11, 2023, 6:20pm 5

Great! Thank you very much.

Topic		Replies	Views
4 lines basic mixture model outputs wrong results Questions	0	394	March 12, 2020
How to create a Mixture of First-order Markov Models Questions	1	1861	July 17, 2018
Help with mixture model of MvNormals in pymc3? Questions	4	739	October 24, 2019
Mixture with multiple observations Questions	6	2347	December 14, 2017
Multinomial mixture with observed-values themselves as mixtures Questions	8	1084	March 29, 2019

Mixture model using pm.Mixture or writing in parts

Related topics