README.md

Steps to generate figures

Note that all the following commands are run under the root directory i.e., risk_aware_mnl.

Preliminary

Install package https://github.com/Alanthink/banditpylib.

Figure 1

• Run worst_regret.sh to generate 10 random input instances and start simulations
• Move generated files in formats data_*.out and params_*.json to arxiv folder (currently arxiv contains all files generated from last time of running)
• Run python3 mnl_bandit.py --final to generate Figure 1 (worst_regret.pdf)

Figure 2

Real parameters are already manipulated and stored in file real_params.json. See file car_data_processing.ipynb on the code to manipulate original data. Then just run

python3 mnl_bandit.py --cvar_data --cvar_fig \
--horizon=1000000 --freq=1000 --card_limit=100 --trials=40 --processes=40 \
--random_neighbors=10 --percentile=5

to generate simulation data (cvar_data.out) and Figure 2 (cvar_fig.pdf).