gym-discrete

Gym-style API

The domanin features a continuos state and a dicrete action space.

The environment initializes:

• cross-sectional dataset with variables X_a, X_s, Y and N observations;
• logit model fitted on the dataset, retrieving parameters \theta_0, \theta_1, \theta_2;

The agent:

• sees all patients;
• predict risk of admission \rho, using initialized parameters
• sample an action (50 possible values between -2 and 2)
• if risk > 0.2:
  • replace Xa by g, where g(\rho, Xa) is obtained using the patient's risk and Xa value
• else:
  • do not intervene, X_a stays the same
• give reward equal to average risk of admission, using predicted Y, initial parameters and sampled values

To install

• git clone https://github.com/claudia-viaro/gym-discrete.git

• cd gym-discrete

• !pip install gym-discrete

• import gym_discrete

• env =gym.make('discrete-v0')

To change version

• change version to, e.g., 1.0.7 from setup.py file
• git clone https://github.com/claudia-viaro/gym-discrete.git
• cd gym-discrete
• python setup.py sdist bdist_wheel
• twine check dist/*
• twine upload --repository-url https://upload.pypi.org/legacy/ dist/*