Repository for simulating, performing parameter/model recovery, and fitting for two exemplar behavioral models in Python. This repo is relevant for Saez laboratory members/collaborators adapting behavioral models for sequential DM tasks:
- Attention at choice and learning model = Rescorla-Wagner variant that implements selective attention to relevant dimensions in a multidimensional state space. This model is based on Leong & Radulescu et al (2017).
- Bayesian inference model = The Bayesian model uses probablistic inference to compute the probability distribution over the identity of the rewarding dimension and feature given all past trials in a multidimensional state space. This model is based on Wilson & Niv (2012).
Simulations are conducted for a multidimensional decision-making task based on Leong & Radulescu et al. (2017) & Wilson & Niv (2012). A maximum likelihood approach based on Wilson & Collins (2019) is used for model/parameter fitting.
Model scripts
- Attention at choice and learning model:
ACL.py - Bayesian inference model:
BI.py
Simulation scripts
simulation.py: Script for model simulation.
Fitting scripts
param_recovery.py: Script for parameter recovery.model_recovery.py: Script for model recovery.model_fitting.py: Script for model fitting with experimental data (for example purposes one example simulated dataset will be used. This is available under the following directoryexample_data/