This project wanted to look into if Learning Automata is suitable for solving the El Farol Bar
problem. We chose to try out the Tsetlin algorithm for the problem. We also made one algorithm
on our own, but we were not able to fully neither develop it nor evaluate it within the scope of
this project. As a reference we implemented the El Farol Bar problem as described by Arthur.
The Tsetlin implementation has proved to be accurate, efficient, scaleable and not least, easy to
implement and it really outperforms the competition.
It is believed that the Jeannequin (Artur) agent would gain if more effort was put into making
good prediction methods. However, this also the main weakness with this agent. A lot of tuning
and effort is needed to make it perform for a certain data set.