mamab-jit

This repo contains all the code for my master's thesis, "Enhancing Team Performance in Multi-Agent Multi-Armed Bandits through Optimization". All experiments are implemented in Python using NumPy, SciPy, Matplotlib, NetworkX, CVXPY, and Numba. Numba is a just-in-time (JIT) compiler that optimizes the performance of Python and NumPy code by translating it to machine code. The experiments in this repo take advantage of the "no-python" mode in Numba along with parallel processing to speed up experiments. See here for more details on Numba.

Note: This repo contains a lot of old code that was used for testing early ideas, quick experimentation, and graphing. They may contain unfinished code, repetitive code, bugs, or other issues. They are moved to a separate directory old_code for reference.

---

The main code for the experiments in the thesis is in the following files:

• bandit_rating_proto.ipynb
  • Initial experiments to measure and analyze the difficulty of bandits (contains some plots used in the thesis)
• bandit_rating_plots.ipynb
  • Calculate and generate plots for the bandit difficulty rating measure
• competency_graphs.ipynb
  • Analyze performances when agents in a team have different competencies
• coopucb2_competency.py
  • Run CoopUCB2 with teams containing agents with different competencies while playing bandits from various difficulty levels
  • Run using python3 coopucb2_competency.py --network NETWORK_NAME --alg ALG
    • Refer to data/saved_networks for available graph structures/networks
    • ALG can either be coopucb2_og or coopucb2_limited_communication
• graph_optimization.py
  • Contains implementations of the heuristic and optimization methods for edge weight setting of a graph.
• large_nets.ipynb and nets.ipynb
  • Generate and analyze large networks
• long_optimization.ipynb
  • Run the long-term optimization process proposed by the thesis

Setup project

1. Setup a virtual environment in the root directory of the repo
   • python3 -m venv venv
2. Activate the virtual environment
   • source venv/bin/activate
3. Install the required packages
   • pip3 install -r requirements.txt
4. Run the experiments