multi-armed bandit simulation models in python
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Stories in Ready

###Bandit Model Research Simulations This work begins by replicating the multi-armed bandit simulation of Posen & Levinthal (2012), "Chasing a Moving Target: Exploitation and Exploration in Dynamic Environments", Management Science 58(3), pp.587-601. It extends this into a general-purpose bandit simulation with additional variables that can be manipulated for further research.

###Contributors Joseph W Clark,

###File Structure

  • All files in this directory (except define experiments.
  • The output subfolder will hold the data and log output from simulation experiments.
  • The analysis subfolder contains some sample scripts for visualizing the data with Python.
  • Files in the bandito subfolder run the actual experiments:
    • /bandito/ runs one replication of a bandit simulation.
    • /bandito/ runs an experiment with any specified set of treatments and number of replications.
    • /bandito/ contains algorithms for turbulence, strategy, beliefs, and payoffs. To implement a new algorithm, place it there.

###How to Use Copy the file and modify to set up your experiment. Run it with something like the following:


Warning: You may want to change the number of replications for a faster test. Posen & Levinthal used 25000 replications for each experimental treatment, and replicating their experiments may take hours or days.

Tested on Mac OS X with Python 3.5.2