/cs839_rl_project_for_submission

Primary LanguagePythonMIT LicenseMIT

Environment constraint

  • The repo assumes the usage of Ubuntu (especially 20.04) because of the package dependency.

Git clone and move current directry

git clone git@github.com:YorkNishi999/cs839_rl_project_for_submission.git
cd /path/to/repo

Build environment

conda env create -f gym-adserver.yml
conda activate gym-adserver
pip install gym-adserver

Comparison analysis

  1. Check below in all agents py files.
# outdir = './outputs/sensitivity/{agent}.txt' # for sensitivity test
outdir = './outputs/{agent}.txt' # for comparison test
  1. Then do the command.
python comparison_test.py
  1. You can see output images in ./outputs/imgs

sensitivity analysis

ucb1

  1. Check below.
outdir = './outputs/sensitivity/ucb1.txt' # for sensitivity test
# outdir = './outputs/ucb1.txt' # for comparison test
  1. Then do the command.
python sensitivity_ucb.py
  1. You can see output images in ./outputs/sensitivity/

Gradient Bandit

  1. Check below.
outdir = './outputs/sensitivity/softmax.txt' # sensitivity test
# outdir = './outputs/softmax.txt' # comparison.py
  1. Then do the command.
python sensitivity_softmax.py
  1. You can see output images in ./outputs/sensitivity/

egreedy

  1. Check below.
outdir = './outputs/sensitivity/egreedy.txt' # sensitivity 
# outdir = './outputs/egreedy.txt' # conparison
  1. Then do the command.
python sensitivity_egreedy.py
  1. You can see output images in ./outputs/sensitivity/