Hi. I’m trying to model experimental data for multiple participants using a hidden Markov model based on pymc-extras’ DiscreteMarkovChain. I have a working model which uses a for loop to create a DiscreteMarkovChain distribution, emission matrix, and observed data for each participant, but it seems …

For this you can define your own distribution with CustomDist. def dist(init_state, transition_p, end_t, size): def step(t, state_tm1, transition_p, end_t): state_t = pm.Categorical.dist(p=transition_p[state_tm1]) # If t >= end_t, we encode as -1 # We need to use a Dira…

Using DiscreteMarkovChain with batch dimensions and data of different lengths

Questions version agnostic

ddg.palmer March 4, 2025, 10:41pm 3

Thank you so much!

Topic		Replies	Views
Implementing the random() function for HMM distribution Questions	3	1285	February 8, 2019
Continuous Time Markov Chain Questions	1	1502	February 18, 2019
Hidden Markov Model - Estimating Transition and Emission CPDs from multiple sequences - not working Questions	11	3900	December 3, 2020
Plate notation: shape parameter/for loop Questions	2	2516	July 30, 2017
RV-Dependent Observations from Irregularly-Shaped Dataset Questions	8	987	July 26, 2017

Using DiscreteMarkovChain with batch dimensions and data of different lengths

Related topics