targetUCB.py is a simple class implementing of the Target-UCB bandit algorithm, as introduced in the AAAI-19 technical paper Leveraging Observations in Bandits: Between Risks and Benefits.
To import the Target-UCB class, simply use the command "from targetUCB import TUCB".
An example of how to construct and test a clique of 4 Target-UCB agents is also provided. Executing the targetUCB.py file as a script will run this clique for 100 episodes on a two-armed bandit problem and display the cumulative regret of all 4 agents.
The supplemental.pdf file is the supplemental material for the AAAI-19 technical paper. It provides proofs, additional experiments and methodological details about the human bandit experiments.