Hi all! I have an aversive learning task, in which participant is watching two stimuli (one at a time). One is associated with shocks (30% of the times) and the other is neve paired with shock. We measure SCR as a way to assess if aversive learning had occurred. Now, I try to build a Bayesian mode…

OK. I have played with the number of trials. Things seem much better now. In 60+ trials the recovery is relatively good. Using the following pymc model with pm.Model() as mB: # betaHyper= pm.Normal('betaH', 0, 1) alpha = pm.Beta('alpha', 1,1, shape=n_subj) beta = pm.Normal('beta',0…

Reinforcement learning - help building a model

Questions

ricardoV94 December 13, 2021, 10:22am 8

That could suggest you have some parameter redundancy (or code bug), although 60 trials is not really that much. If you try something like 500 trials for one participant do things get any better?

Topic		Replies	Views
Modeling reinforcement learning of human participant using PyMC3 Questions	8	3224	June 7, 2023
Fitting a simple Reinforcement Learning model to behavioral data with PyMC (Jupyter NB) Sharing	27	3477	April 28, 2023
Fitting Reinforcement Learning model modeling	0	355	November 4, 2023
Behrens' Bayesian Learner Model : How to share parameters between steps? Questions	44	2313	January 7, 2022
Theano.scan and pm.Model() are strangely incompatible Questions	9	1843	May 21, 2019

Reinforcement learning - help building a model

Related topics