This repository hosts the code used for the experiments featured in the paper:

“Online Experimental Design With Estimation–Regret Trade-off Under Network Interference.”

---

Experiment Scripts

• exp_101unit_sto_UCB.py
  Models a 101‑node star graph to run multi‑armed bandit trials that account for interference across the network.

All scripts benchmark three exploration–exploitation algorithms:

• Uniform: Pure exploration by sampling arms evenly.
• Standard UCB: The classic Upper Confidence Bound approach for minimizing regret.
• UCB_TSN: Our novel procedure that trades off regret reduction against statistical inference accuracy.

---

How to Launch

From your shell, simply invoke:

# Run the star‑network bandit simulation
python exp_101unit_sto_UCB.py