QuBBD RL

Codebase containing reinforcement learning experiments for SMART-ACT project using the QuBBD data

This repository contains two experiments:

1. Simulator: Completely simulated data and environment. Self defined policy based on random vectors.
2. QuBBD: Policy defined based on Qubbd v3 data.

Directory Structure

The structure of the project as it is now is as follows:

• /root
  • /QuBBD
    • /data (find data on Box)
    • constants.py
    • preprocess.py
    • utils.py
  • /simulator
    • dqn.py
    • model.py
    • plotting.py
    • policy.py
    • run_exp.py

Acknowledgements

• Deep Q-Network model has been adapted from Denny Britz's repository and modified for our purpose.
• Policy definitions have been inspired by UC Berkeley CS 294-112 homework definitions.

---

This is still an ongoing project.