This repository contains code that can simulate the Metropolis-Drift-Diffusion Model.
This code is written for Python 3.
It requires the ddm module, which should be installed manually prior to downloading this code.
After that, simply download this code in a local directory and import its modules from within that
directory (see below).
The code is throughly commented and the documentation is accessible via docstrings (see below).
To perform an interactive simulation, import the interface_console module and run the
run_comparison() function. You will be asked a series of questions on the parameters to use,
then a simulation will be run and the results plotted.
The core algorithms are in the metropolis_ddm module. The timed_metropolis function performs
one simulation run. The metropolis_ddm_hist function runs the model repeatedly and returns a
count of the number of times each alternative was chosen.
For example to run 10³ simulations with 4 alternatives, utilities (1, 3, 5, 7), lower (incumbent) threshold at -1, upper (candidate) threshold at 2, time limit 5, with uniform exploration matrix (the default), you can use:
In [1]: from metropolis_ddm import metropolis_ddm_hist
In [2]: metropolis_ddm_hist([1.0, 3.0, 5.0, 7.0], 1.0, 2.0, 5.0, num_samples=10**3)
Out[2]: array([ 0, 4, 48, 948])In this batch of simulations, alternative 1 was never chosen, alternative 2 was chosen 0.4% of the times, etc.
For more help, try:
In [1]: import metropolis_ddm as mddm
In [2]: mddm.ddm_sample?
In [3]: mddm.DDMSampler?
In [5]: mddm.timed_metropolis?
In [7]: mddm.metropolis_ddm_hist?